Re: [agi] The Smushaby of Flatway.
On Jan 8, 2009, at 10:29 AM, Ronald C. Blue wrote: ...Noise is not noise... Speaking of noise, was that ghastly HTML formatting really necessary? It made the email nearly unreadable. J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Hypercomputation and AGI
On Jan 1, 2009, at 2:35 PM, J. Andrew Rogers wrote:
Since "digital" and "analog" are the same thing computationally
("digital" is a subset of "analog"), and non-digital computers have
been generally superior for several decades, this is not relevant.
Gah, that should be *digital* computers have generally been superior
for several decades (the last non-digital hold-outs I am aware of
were designed in the late 1970s).
J. Andrew Rogers
---
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription:
https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b
Powered by Listbox: http://www.listbox.com
Re: [agi] Hypercomputation and AGI
On Dec 30, 2008, at 11:45 AM, Steve Richfield wrote:
Bingo! You have to "tailor" the techniques to the problem - more
than just "solving the equations", but often the representation of
quantities needs to be in some sort of multivalued form.
What I meant is that if the standard algebraic reduction algorithm is
not possible, there are other algorithms you can use to generate a set
of equations that can be solved using algebraic reduction. Humans are
pretty limited in their ability to manually apply the "generate a set
of equations that can be solved" algorithm(s) because there are too
many dimensions, but computers have no problem. I cut my teeth
working on these types of solvers (in FORTRAN, yech).
I wonder if we aren't really talking about analog computation (i.e.
computing with analogues, e.g. molecules) here? Analog computers
have been handily out-computing digital computers for a long time.
Since "digital" and "analog" are the same thing computationally
("digital" is a subset of "analog"), and non-digital computers have
been generally superior for several decades, this is not relevant. The
difference between "digital" and "analog" is the signal-to-noise ratio
(SNR) that has to be maintained by the computer system. You can
simulate with perfect fidelity high SNR computers on low SNR computers
(like digital computers) since they are equivalent, trading SNR for
frequency. If you apply the formula for converting digital bits to
analog SNR (analog SNR = 1.76+6.02*bits), it becomes obvious why
things like thermal noise make it impossible to directly implement
e.g. a modest 32-bit digital processor as a non-digital equivalent.
When most people talk about "analog computation", they are really
talking about real computers (whether they realize it or not), which
are a form of hypercomputer. If it was possible to build such a
computer, it would have some strange consequences for physics that are
not in evidence.
Cheers,
J. Andrew Rogers
---
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription:
https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b
Powered by Listbox: http://www.listbox.com
Re: [agi] Hypercomputation and AGI
On Dec 30, 2008, at 12:51 AM, Steve Richfield wrote: On a side note, there is the "clean" math that people learn on their way to a math PhD, and then there is the "dirty" math that governs physical systems. Dirty math is fraught with all sorts of multi- valued functions, fundamental uncertainties, etc. To work in the world of "dirty" math, you must escape the notation and figure out what the equation is all about, and find some way of representing THAT, which may well not involve simple numbers on the real-number line, or even on the complex number plane. What does "dirty math" really mean? There are engineering disciplines essentially *built* on solving equations with gross internal inconsistencies and unsolvable systems of differential equations. The modern world gets along pretty admirably suffering the very profitable and ubiquitous consequences of their quasi-solutions to those problems. But it is still a lot of hairy notational math and equations, just applied in a different context that has function uncertainty as an assumption. The unsolvability does not lead them to pull numbers out of a hat, they have sound methods for brute-forcing fine approximations across a surprisingly wide range of situations. When the "clean" mathematical methods do not apply, there are other different (not "dirty") mathematical methods that you can use. Indeed, I have sometimes said the only real education I ever got in AI was spending years studying an engineering discipline that is nothing but reducing very complex systems of pervasively polluted data and nonsense equations to precise predictive models where squeezing out an extra 1% accuracy meant huge profit. None of it is directly applicable, the value was internalizing that kind of systems perspective and thinking about every complex systems problem in those terms, with a lot of experience algorithmically producing predictive models from them. It was different but it was still ordinary math, just math appropriate for the particular problem. The only thing you could really say about it was that it produced a lot of great computer scientists and no mathematicians to speak of (an odd bias, that). With this as background, as I see it, hypercomputation is just another attempt to evade dealing with some hard mathematical problems. The definition of "hypercomputation" captures some very specific mathematical concepts that are not captured in other conceptual terms. I do not see what is being evaded, since it is more like pointing out the obvious with respect to certain limits implied by the conventional Turing model. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Hypercomputation and AGI
On Dec 29, 2008, at 1:22 PM, Ben Goertzel wrote: Well, some of the papers in the references of my paper give formal mathematical definitions of hypercomputation, though my paper is brief and conceptual and not of that nature. So although the generic concept may be muddled, there are certainly some fully precise variants of it. My comment was not really against the argument you make in the paper, nor do I disagree with your definition of "hypercomputation". (BTW, run spellcheck.) I was referring to the somewhat anomalous difficulty of deciding whether or not some computational models truly meet that definition as a practical matter. Anyway the argument in my paper is pretty strong and applies to any variant with power beyond that of ordinary Turing machines, it would seem... No disagreement with that, which is why I called it a "meta- comment". :-) Super-recursive algorithms, inductive Turing machines, and related computational models can be made to sit in a somewhat fuzzy place with respect to whether or not they are hypercomputers or normal Turing machines. A Turing machine that asymptotically converges on producing the same result as a hypercomputer is an interesting case insofar as the results they produce may be close enough that you can consider the difference to be below the noise floor, and if they are functionally equivalent using that somewhat unusual definition then you effectively have equivalence to a hypercomputer without the hypercomputer. Not strictly by definition, but within some strictly implied error bound for the purposes of comparing output (which is all we usually care about). The concept of non-isotropic distributions of random numbers has always interested me for much the same reason, since there seems to be a similar concept at work there. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Hypercomputation and AGI
On Dec 29, 2008, at 10:45 AM, Ben Goertzel wrote: I expanded a previous blog entry of mine on hypercomputation and AGI into a conference paper on the topic ... here is a rough draft, on which I'd appreciate commentary from anyone who's knowledgeable on the subject: http://goertzel.org/papers/CognitiveInformaticsHypercomputationPaper.pdf This is a theoretical rather than practical paper, although it does attempt to explore some of the practical implications as well -- e.g., in the hypothesis that intelligence does require hypercomputation, how might one go about creating AGI? I come to a somewhat surprising conclusion, which is that -- even if intelligence fundamentally requires hypercomputation -- it could still be possible to create an AI via making Turing computer programs ... it just wouldn't be possible to do this in a manner guided entirely by science; one would need to use some other sort of guidance too, such as chance, imitation or intuition... As more of a meta-comment, the whole notion of "hypercomputation" seems to be muddled, insofar as super-recursive algorithms may be a limited example of it. I was doing a lot of work with inductive Turing machines several years ago, and most of the differences seemed to be definitional e.g. what constitutes an algorithm or answer. For most practical purposes, the price of implementing them in conventional discrete space is the introduction of some (usually acceptable) error. But if they approximate to the point of functional convergence on a normal Turing machine... As best I have been able to tell, and I have not really been paying attention because the arguments seem to mostly be people talking past each other, is that ITMs raise some interesting philosophical questions regarding hypercomputation. We cannot implement a *strict* hypercomputer, but to what extent does it "count" if we can asymptotically converge on the functional consequences of a hypercomputer using a normal computer? It suspect it will be hard to evict the belief in Penrosian magic from the error bars in any case. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Indexing
On Dec 26, 2008, at 7:40 PM, Jim Bromer wrote: I noticed that neither linked lists nor arrays were particularly efficient for general operations that would include insertions, deletions and searches, which, when you think about it, are pretty much the norm. How often do you need a large data index that only rarely needs to be searched. The irony is that you cannot combine the two forms in a simple manner so that you can have a linked list for fast insertion and deletion and an array for fast searches. There are data structures and algorithms that offer fast insert/delete and fast search, approximately constant computational complexity for both even. It does require slightly more cleverness than a linked list though since your glorified lookup table will require a space- preserving representation. It is much more common and usually simpler to merely use order-preserving representations like the common B+tree variants unless you have vast quantities of data. Brute-force can be exceedingly efficient in small doses. And with indirect indexes (using a handle or an index to an location entry) the data requires frequent compression (to squeeze out the gaps in the data area) if there is a heavy insertion and deletion. An old, solved problem. Well, "solved" in the sense that the tradeoffs and methods for managing this are well-understood. I believe the problem is directly related to agi because data relevant to some particular situation will tend to be distributed in a file so that a lot of relational indexing is needed. Perhaps the most relevant application to AGI is that it would very significantly improve the computational complexity of of representing and manipulating high-dimensionality relationships, particularly in distributed systems. In conventional data-mining and pattern discovery analytics, the lack of scalability of high-dimensionality representations has long been major limitation on what one could do. But for AI, consider algorithms like SIFT, which turn massive aggregates of 2-dimensional representations of 3-dimensional space (i.e. "photos") into a virtual model of the 3D space represented. A neat algorithm, but limited by the fact that the algorithm represents the data in a 128-dimensional space before reducing it to 3- dimensional space, limiting the amount of data you could apply as a practical matter. Since a lot of data can be described as being analogously similar to other kinds of data and since many variations in some particular kind of data might already exist in a database, a great many complicated modifications of concepts could, hypothetically, be done by modifying the indexes alone. In an ideal system, the database relation *is* the index. External indexes are largely a software engineering artifact of only being able to represent one dimension per relation in a scalable manner. J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Universal intelligence test benchmark
On Dec 26, 2008, at 7:24 PM, Philip Hunt wrote: 2008/12/27 J. Andrew Rogers : I think many people greatly underestimate how many gaping algorithm holes there are in computer science for even the most important and mundane tasks. The algorithm coverage of computer science is woefully incomplete, Is it? In all my time as a programmer, it's never occurred to me to think "I wish there was an algorithm to do X". mybe that's just me. And there are vast numbers of useful algorithms that people use every day. Computers are general, so there always exists an obvious algorithm for doing any particular task. Whether or not that obvious algorithm is efficient is quite another thing, since the real costs of various algorithms are far from equivalent even if their functionality is. The Sieve of Eratosthenes will allow you to factor any integer in theory, but for non-trivial integers you will want to use a number field sieve. The limitations of many types of software are fundamentally based in the complexity class of the of the attributes of the algorithms they use. We frequently improperly conflate "theoretically impossible" and "no tractable algorithm currently exists". I wonder (thinking out loud here) are there any statistics for this? For example if you plot the number of such algorithms that've been found over time, what sort of curve would you get? (Of course, you'd have to define "general, elegant algorithm for basic problem", which might be tricky) I am still surprised often enough that it is obvious that there is considerable amounts of innovation still being done. It both amuses and annoys me no end that some common algorithms have design characteristics that reflect long-forgotten assumptions that do not even make sense in the context they are used e.g. compulsive tree balancing behavior of intrinsically unbalanced data structures. In short, we have no idea what important and fundamental algorithms will be discovered from one year to the next that change the boundaries of what is practically possible with computer science. Is this true? It doesn't seem right to me. AIUI the current state of the art in operating systems, compilers, garbage collectors, etc is only slightly more efficient than it was 10 or 20 years ago. (In fact, most practical programs are a good deal less efficient, because faster processors mean they don't have to be). It is easy to forget how many basic algorithms we use ubiquitously are relatively recent. The concurrent B-tree algorithm that is pervasively used in databases, file systems, and just about everything else was published in the 1980s. In fact, most of the algorithms that make up a modern SQL database as we understand them were developed in the 1980s, even though the relational model goes back to the 1960s. I don't think I understand you. To me "indexing" means what the Google search engine or an SQL database does -- but you're using the word with a different meaning aren't you? I mean it exactly like you understand it. Indexed access methods and representations. Sorry, you've lost me again -- I've never heard of the term "hyper-rectangles" in relation to relational databases. Most people haven't, because there are no hyper-rectangles in relational database *implementations* seeing as how there are no useful algorithms for representing them. Nonetheless, the underlying model describes operations using hyper-rectangles in high-dimensional spaces. In an ideal relational implementation there are never external indexes, only data organized in its native high-dimensionality logical space, since external indexes are a de-normalization. It is not because it is theoretically impossible, but because it is only possible if someone discovers a general algorithm for indexing hyper-rectangles -- faking it is not distributable. How do we know that there is such an algorithm? We don't unless someone publishes one, but there is a lot of evidence that seems to imply otherwise and which proves that much of the research that has been done was misdirected. Aesthetically, the current algorithms for doing this are nasty ugly hacks, and that lack of elegance is often an indicator that a better way exists. In the specific case of indexing hyper-rectangles, the first basic algorithm was published in 1971 (IIRC), but was supplanted by a completely different family of algorithm in 1981. Virtually all research has been based on derivatives of the 1981 algorithm, since it appeared to have better properties. Unfortunately, we can now prove that this algorithm class can never yield a general solution and that a solution must look like a variant of the original 1971 algorithm family that has been ignored for a quarter century. Interestingly,
Re: [agi] Universal intelligence test benchmark
On Dec 26, 2008, at 6:18 PM, Ben Goertzel wrote: Most compression tests are like defining intelligence as the ability to catch mice. They measure the ability of compressors to compress specific files. This tends to lead to hacks that are tuned to the benchmarks. For the generic intelligence test, all you know about the source is that it has a Solomonoff distribution (for a particular machine). I don't know how you could make the test any more generic. IMO the test is *too* generic ... I don't think real-world AGI is mainly about being able to recognize totally general patterns in totally general datasets. I suspect that to do that, the best approach is ultimately going to be some AIXItl variant ... meaning it's a problem that's not really solvable using a real-world amount of resources. I suspect that all the AGI system one can really build are SO BAD at this general problem, that it's better to characterize AGI systems An interesting question is which pattern subset if ignored would make the problem tractable. J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Universal intelligence test benchmark
On Dec 26, 2008, at 2:17 PM, Philip Hunt wrote: I'm not dismissive of it either -- once you have algorithms that can be practically realised, then it's possible for progress to be made. But I don't think that a small number of clever algorithms will in itself create intelligence -- if that was possible then the secret to AI would have been discovered by now. I think some people get seduced by the beauty and clarity of maths and want to make their programs like that, but I don't think human intelligence is like that. Never mind discovering "a small number of clever algorithms" for AI, we have not even discovered a great many basic algorithms for routine computer science. I think many people greatly underestimate how many gaping algorithm holes there are in computer science for even the most important and mundane tasks. The algorithm coverage of computer science is woefully incomplete, which is why after a half century people are *still* finding general, elegant algorithms for basic problems, many of which are bloody obvious in hindsight. In short, we have no idea what important and fundamental algorithms will be discovered from one year to the next that change the boundaries of what is practically possible with computer science. For example, there is no general indexing algorithm described in computer science. In fact, the only useful indexing algorithm index points on a line. Not points in arbitrary space, not intervals on lines, not hyper-rectangles in high-dimensionality space, never mind more complex relations. Oddly enough, most computer scientists are ignorant of the fact that no useful indexing algorithm exists for most data representations or that a vast number of software applications are not tractably implementable as a result. The ability to tractably index almost nothing has consequences. Relational database theory describes the manipulation of hyper- rectangles, but we fake it very badly with indexes we actually have algorithms for. Did you ever wonder why no one has built a massively distributed SQL database despite the obvious value? It is not because it is theoretically impossible, but because it is only possible if someone discovers a general algorithm for indexing hyper-rectangles -- faking it is not distributable. Several other big limitations in software are actually based in (the absence of) this algorithm. It is utterly trivial to describe, and there are literally several dozen algorithms that come close, but after 40 years no one has published such an algorithm. When such an algorithm is finally published, it will completely reset everything we think we know about many algorithms and data structures. There is a really large laundry list of undiscovered fundamental algorithms like this that we work around with mediocre alternatives. If you look at most of the limits of software, the vast majority are not theoretical limits but limits based on the fact that there a lot of missing pages in our data structures and algorithm texts. Spatial indexing, for example, currently uses "insanely, infeasibly much computation resource", so no one implements it beyond uselessly trivial systems. But as most people familiar with the minutiae of the related theoretical computer science will tell you, not only is it very probable that a broadly general algorithm exists, but it will almost certainly scale like Google does. We will go from "intractable" to "insanely cheap" in one day. The algorithms around the AIT definition of intelligence look very much like a similar case,a very sparsely studied algorithm space with some rather obvious gaping holes when it comes to the kinds of algorithms that very likely should exist in that space. It would seem premature to write it off solely on the basis of the negligible computer science that has thus far been done in that algorithm space. An AGI written by humans would hopefully be a lot more nicely structured than this, but I think it would still consist of large number of modules, none of which was intelligent in itself. How big would it be? The human genome is 750 MB so intelligence could presumably be coded in less than that. I'd guess an AGI could be written in about a tenth that, say 75 MB. The human genome size has no meaningful relationship to the complexity of coding AGI. And what ever happened to Machine is Software is Data? Ignoring this seems to be a frequent enabler of specious reasoning. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Introducing Steve's "Theory of Everything" in cognition.
On Dec 24, 2008, at 10:33 PM, Steve Richfield wrote: Of course you could simply subtract successive samples from one another - at some considerable risk, since you are now sampling at only half the Nyquist-required speed to make your AGI/NN run at its intended speed. In short, if inputs are not being electronically differentiated, then sampling must proceed at least twice as fast as the NN/AGI cycles. Or... you could be using something like compressive sampling, which safely ignores silly things like the Nyquist limit. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Building a machine that can learn from experience
On Dec 19, 2008, at 7:01 PM, Ben Goertzel wrote: Yes, you can work around it by assuming Occam's Razor as a sort of primal religious principle ... but then you're making a big assumption pulled out of the glorious subjective nothing ... which is fine, but you should acknowledge that's what you're doing... I do acknowledge this readily; I often have to disabuse people of the notion that I believe much of anything in an absolute sense. I'm not a big believer in absolute truth, but if I have to accept some axioms as a creedal minimalist, I will provisionally start with mathematics since everything else useful seems to follow from there and I need to get work done. Provisional in the sense that I only accept those axioms as long as they remain unreasonably effective. If there is some other base assumption that is similarly simple and works better, I would happily entertain it. I just am not aware of one though that may reflect my limited experience. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] AGI Preschool: sketch of an evaluation framework for early stage AGI systems aimed at human-level, roughly humanlike AGI
On Dec 19, 2008, at 6:43 PM, Ben Goertzel wrote: Although, I note, I know a really good baker who makes great cakes in spite of the fact that she does not eat sugar and hence does not ever taste most of the stuff she makes... But she *used to* eat sugar, so to an extent she can go on memory Fortunately, baking is more about process control than flavor control. Unlike normal cooking, which is significantly fine-tuned by taste, the taste of baked goods is pretty invariant. On the other hand, baking requires a lot of attention to detail and process precision that normal cooking does not. Which is why I am merely an adequate baker instead of a great one. :-) Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Building a machine that can learn from experience
On Dec 19, 2008, at 5:35 PM, Ben Goertzel wrote: The problem is that **there is no way for science to ever establish the existence of a nonalgorithmic process**, because science deals only with finite sets of finite-precision measurements. I suppose it would be more accurate to state that every process we can detect is algorithmic within the scope of our ability to measure it. Like with belief in god(s) and similar, the point can then be raised as to why we need to invent non-algorithmic processes when ordinary algorithmic processes are sufficient to explain everything we see. Non-algorithmic processes very conveniently have properties identical to the "supernatural", and so I treat them similarly. This is just another incarnation of the old "unpredictable versus random" discussions. Sure, non-algorithmic processes could be running the mind machinery, but then so could elves, unicorns, the Flying Spaghetti Monster, and many other things that it is not necessary to invoke at this time. Absent the ability to ever detect such things and lacking the necessity of such explanations, I file non-algorithmic processes with vast number of other explanatory memes of woo-ness of which humans are fond. Like the old man once said, "entia non sunt multiplicanda praeter necessitatem". Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Building a machine that can learn from experience
On Dec 19, 2008, at 4:45 PM, Colin Hales wrote: I'm not clear how you came to the conclusion that I was discussing an 'algorithmic system'. You, like the rest of us, are incapable of discussing anything else. Email cannot carry non-algorithmic ideas or concepts. Just because you do not consider your system "algorithmic" does not mean that it is not. Nature is algorithmic, your chip is algorithmic, everything is algorithmic. That which we call a rose by any other name would smell as sweet. If you really understood the implications of your assertion, you would not have wasted your time trying to explain it to us. Seriously, you should think *really hard* about what you have asserted in your last several posts, because the set of assertions you make are transparently internally inconsistent, never mind that you play fast and loose with the definitions of the terms you are using to get around pesky theoretical restrictions. There seems to be sufficient available evidence to doubt that 'cognition is computation'. There seems to be sufficient available evidence to doubt that you understand 'computation' well enough to make this judgement. It is not so much that I understand everything you are talking about, but that the parts are I *do* understand are quite wrong on their own. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Building a machine that can learn from experience
On Dec 19, 2008, at 12:13 PM, Colin Hales wrote: The answer to this is that you can implement it in software. But you won't do that because the result is not an AGI, but an actor with a script. I actually started AGI believing that software would do it. When I got into the details of the issue of qualia (their role and origins) I found that software alone would not do the trick. Nonsense, an algorithmic system is describable entirely based on input and output without any regard for its internal structure. If two blackbox systems produce identical output based on identical input, then they are mathematically equivalent in every meaningful sense even if they have radically different internal construction. You say "actor with a script" as if that means something important, ignoring that every process in our universe is necessarily equivalent to an "actor with a script". Your magical EM chip is, in fact, "an actor with a script". The simplest way to get to the position I inhabit is to consider that the electromagnetic field has access to more information (about the world outside the agent) than that available through peripheral nerve signaling. It's the additional information that is thrown away with a model of the electromagnetic field. This does not even make sense. Either the software model captures the measurable properties of the EM field or it does not, but either way it does not support your proposal. In the former case, the external input and dynamic *must* be measurable and therefore can be reflected in the software model, and in the latter case it is nothing more than handwaving about something you are asserting exists in the complete absence of material evidence. I'm having a hard time accepting that there is something you can specify and measure that magically has no useful software description. That is not even supposed to be possible as a kind of basic mathematical theorem thing. I mean, you are asserting that some very specific inputs to the system are not being modeled, and if you know this then you can very easily add them to the software-modeled system. You have not explained why this is not possible, merely asserted it. J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Building a machine that can learn from experience
On Dec 18, 2008, at 10:09 PM, Colin Hales wrote: I think I covered this in a post a while back but FYI... I am a little 'left-field' in the AGI circuit in that my approach involves literal replication of the electromagnetic field structure of brain material. This is in contrast to a computational model of the electromagnetic field structure. Here is a silly question: If you can specify it well enough to implement the desired result in hardware, why can't you implement it in software? It is equivalent, after all. And if you can't specify the dynamic well enough to implement it virtually, why would there be any reason at all to believe that it will do anything interesting? The hallmark of a viable AGI theory/design is that you can explain why it *must* work in sufficient detail to be implementable in any medium. J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: >> RE: FW: [agi] A paper that actually does solve the problem of consciousness
On Dec 2, 2008, at 8:31 AM, Ed Porter wrote: From my quick read it appears the only meaningful way it suggests a brain might be infinite was that since the brain used analogue values --- such as synaptic weights, or variable time intervals between spikes (and presumably since those analogue values would be determined by so many factors, each of which might modify their values slightly) --- the brain would be capable of computing many values each of which could arguably have infinite gradation in value. So arguably its computations would be infinitely complex, in terms of the number of bits that would be required to describe them exactly. If course, it is not clear the universe itself supports infinitely fine gradation in values, which your paper admits is a questions. The universe has a noise floor (see: Boltzmann, Planck, et al), from which it follows that all "analog" values are equivalent to some trivial number of bits. Since "digital" deals with the case of analog at the low end of signal to noise ratios, "digital" usually denotes a proper subset of "analog", making the equivalence unsurprising. The obvious argument against infinite values is that the laws of thermodynamics would no longer apply if that were the case. Given the weight of the evidence for thermodynamics being valid, it is probably prudent to stick with models that work when restricted to a finite dynamic range for values. The fundamental non-equivalence of digital and analog is one of those hard-to-kill memes that needs to die, along with the fundamental non- equivalence of parallel and serial computation. Persistent buggers, even among people who should know better. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Mushed Up Decision Processes
On Nov 30, 2008, at 7:31 AM, Philip Hunt wrote: 2008/11/30 Ben Goertzel <[EMAIL PROTECTED]>: In general, the standard AI methods can't handle pattern recognition problems requiring finding complex interdependencies among multiple variables that are obscured among scads of other variables The human mind seems to do this via building up intuition via drawing analogies among multiple problems it confronts during its history. Yes, so that people learn one problem, then it helps them to learn other similar ones. Is there any AI software that does this? I'm not aware of any. To do this as a practical matter, you need to address *at least* two well-known hard-but-important unsolved algorithm problems in completely different areas of theoretical computer science that have nothing to do with AI per se. That is no small hurdle, even if you are a bloody genius. That said, I doubt most AI researchers could even tell you what those two big problems are which is, obliquely, the other part of the problem. I have proposed a problem domain called "function predictor" whose purpose is to allow an AI to learn across problem sub-domains, carrying its learning from one domain to another. (See http://www.includipedia.com/wiki/User:Cabalamat/Function_predictor ) In Feder/Merhav/Gutman's 1995 "Reflections on..." followup to their 1992 paper on universal sequence prediction, they make the observation, which can be found at the following link, that it is probably useful to introduce the concept of "prediction error complexity" as an important metric which is similar to what you are talking about in the theoretical abstract: http://www.itsoc.org/review/meir/node5.html Our understanding of this area is better in 2008 than it was in 1995, but this is one of the earliest serious references to the idea in a theoretical way. Somewhat obscure and primitive by current standards, but influential in the AIXI and related flavors of AI theory based on computational information theory. Or at least, I found it very interesting and useful a decade ago. Cheers, J. Andrew Rogers --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244&id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
On Jun 23, 2008, at 7:53 PM, Steve Richfield wrote: Andy, The use of diminutives is considered rude in many parts of anglo- culture if the individual does not use it to identify themselves, though I realize it is common practice in some regions of the US. When in doubt, use the given form. This is a PERFECT post, because it so perfectly illustrates a particular point of detachment from reality that is common among AGIers. In the real world we do certain things to achieve a good result, but when we design politically correct AGIs, we banish the very logic that allows us to function. For example, if you see a black man walking behind you at night, you rightly worry, but if you include that in your AGI design, you would be dismissed as a racist. You have clearly confused me with someone else. Effectively solving VERY VERY difficult problems, like why a particular corporation is failing after other experts have failed, is a multiple-step process that starts with narrowing down the vast field of possibilities. As others have already pointed out here, this is often done in a rather summary and non-probabilistic way. Perhaps all of the really successful programmers that you have known have had long hair, so if the programming is failing and the programmer has short hair, then maybe there is an attitude issue to look into. Of course this does NOT necessarily mean that there is any linkage at all - just another of many points to focus some attention to. Or it could simply mean that the vast majority of programmers and software monkeys are mediocre at best such that the handful of people you will meet with deep talent won't constitute a useful sample size. Hell, even Brooks suggested as much and he was charitable. In all my years in software, I've only met a small number of people who were unambiguously wicked smart when it came to software, and while none of them could be confused with a completely mundane person they also did not have many other traits in common (though I will acknowledge they tend to rational and self-analytical to a degree that is rare in most people though this is not a trait exclusive to these people). Of course, *my* sample size is also small and so it does not count for much. Similarly, over the course of >100 projects... Eh? Over 100 projects? These were either very small projects or you are older than Methuselah. I've worked on a lot of projects, but nowhere near 100 and I was a consultant for many years. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] Approximations of Knowledge
On Jun 22, 2008, at 1:37 PM, Steve Richfield wrote: At the heart of the most troubled projects. I typically find either a born-again Christian or a PhD Chemist. These people make the same bad decisions from faith. The Christian's faith is that God wouldn't lead them SO astray, so abandoning the project would in effect be abandoning their faith in God - which of course leads straight to Hell. The Chemist has heard all of the stories of perseverance leading to breakthrough discoveries, and if you KNOW that the solution is there just waiting to be found, then just keep on plugging away. These both lead to projects that stumble on and on long after any sane person would have found another better way. Christians tend to make good programmers, but really awful project managers. Somewhere in the world, there is a PhD chemist and a born-again Christian on another mailing list "...the project had hit a serious snag, and so the investors brought in a consultant that would explain why the project was broken by defectively reasoning about dubious generalizations he pulled out of his ass..." J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] More brain scanning and language
On Jun 12, 2008, at 9:25 AM, Steve Richfield wrote: My assertion was that once you figure out just what it is that the neurons are doing, that the difference between neural operation and optimal operation will be negligible. This because of the 200 million years they have had to refine their operation. Of course, the other argument was that there was just so much that could be done in wetware. While all computational models are general in theory, they optimize for different kinds of operations in practice such that an algorithm that could be efficiently implemented on one would be nearly intractable on another. We see this kind of impedance matching issue in regular silicon architectures, with different functions/algorithms putting different stresses on the model. I don't doubt that neurons are reasonably optimal implementations of their computing model, but there will be some types of functions that are not very efficient using them. Evolution optimized the architecture for a specific use case given the materials and processes at hand. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More brain scanning and language
On Jun 11, 2008, at 5:56 AM, Mark Waser wrote: It is an open question as to whether or not mathematics will arrive at an elegant solution that out-performs the sub-optimal wetware algorithm. What is the basis for your using the term sub-optimal when the question is still open? If mathematics can't arrive at a solution that out-performs the wetware algorithm, then the wetware isn't suboptimal. Lack of an elegant solution, one that is more efficient than the wetware methods in the broadest general case, does not imply that mathematics does not already describe superior average case methods. Wetware methods are general, but tend toward brute-force search methods that can be improved upon. A number of recent papers suggest that an elegant, general solutions may be possible; it is an active area of DARPA-funded theoretical mathematics research. None of which has anything to do with AI, except to the extent AI may involve efficiently manipulating models of spaces. Sloppy thinking and hidden assumptions as usual . . . . The irony is rich. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More brain scanning and language
On Jun 11, 2008, at 12:05 AM, Vladimir Nesov wrote: And it extends to much more than 3D physical models -- humans are able to adjust dynamic representations on the fly, given additional information about any level of description, propagating consequences to other levels of description and forming a plausible model from heterogeneous hints. I consider this ability to accumulate flexible, incrementally adjustable models, that can incorporate hints from nonatomic analogous models, to be the central capability of human-like intelligence. While I was not aiming at it in particular, it is a manifestation of a very general model of computation based on efficient prediction and induction. The model is informed by measurements of the world, which allows us to predict the next best action in some goal context. You can build very general computational models this way, even (or particularly) when thinking about interactions with 3-dimensional spaces that are not directly tractably decidable. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More brain scanning and language
On Jun 3, 2008, at 8:44 AM, Mike Tintner wrote: Thanks. I must confess to my usual confusion/ignorance here - but perhaps I should really have talked of "solid" rather than "3-D mapping." When you sit in a familiar chair, you have, I presume, a solid mapping (or perhaps the word should be "moulding") - distributed over your body, of how it can and will fit into that chair. And I'm presuming that the maps in the brain may have a similar solid structure. And when you're in a familiar room, you may also have brain maps [or "moulds"] that tell you automatically what is likely to be in front of you, at back, and on each side. Does your sense of "3-D mapping" equate to this? Humans are capable of constructing exquisite 3-dimensional models in their minds. see: blind people. Having that model and computing interactions with that model are two different things. Humans do not actually compute their relation to other objects with high precision, they approximate and iteratively make corrections later. It turns out this may not be such a bad idea, computational topology and geometry is thin on computable high- precision results, but it kind of goes against the grain of computer science. It is not obvious that having that 3-dimensional model and being able to compute extremely complex relationships on the fly are the same problem. We can do the former, both as humans and on computers, but the latter is beyond both humans and computer science. We have a model, but our poorly calibrated interactions with it are constantly moderated by real-world feedback. It is an open question as to whether or not mathematics will arrive at an elegant solution that out-performs the sub-optimal wetware algorithm. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 8, 2008, at 7:27 PM, Richard Loosemore wrote: I directly and exactly *quoted* several passages that you wrote. And completely ignored both the context and intended semantics. Hence why I might be under the impression that there is a reading comprehension issue. But enough of that, let's get to the meat of it: Are you arguing that the function that is a neuron is not an elementary operator for whatever computational model describes the brain? J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 7, 2008, at 5:06 PM, Richard Loosemore wrote: But that is a world away from the idea that neurons, as they are, are as simple as transistors. I do not believe this was a simple misunderstanding on my part: the claim that neurons are as simple as transistors is an unsupportable one. Richard, you reliably ignore what I actually write, selectively parsing it in some bizarre context that I don't recognize. There is a reading comprehension issue, or at the very least you don't follow what I consider to be the dead obvious theoretical implications. Metaphorically, you are arguing that the "latex sheet" model of gravitational curvature is stupid because astronomers have never seen latex in space, and then wonder why the physicists are giving you funny looks. Are you arguing that the function that is a neuron is *not* an elementary operator for whatever computational model it is that describes the brain? J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 7, 2008, at 10:44 AM, Vladimir Nesov wrote: On Sat, Jun 7, 2008 at 8:30 PM, Richard Loosemore <[EMAIL PROTECTED]> wrote: But I have no problem with this at all! :-). This is exactly what I believe, but I was arguing against a different claim! Rogers did actually say that "neurons are simple" and then went on to claim that they were simple because (essentially) you could black-box them with something like a bayesian function. You stepped in and said things that implied you were defending his position, that is all. I certainly am not arguing that neuron functionality will probably be modelled much more simply, in the long run. But that is different. I think you misinterpreted his position also then. I certainly interpreted it to mean something along the lines of what I've just summarized, or even more generally that a design that is even not a neural net can be even more efficient and simple. He is too smart to believe in silliness you argued against. For the record, Vladimir did a pretty good job summarizing my position. The reliability with which some people misinterpreted it, to the point that it almost looked willful, highlighted the futility of my attempting to close that particular comprehension gap. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More brain scanning and language
On Jun 3, 2008, at 6:44 AM, Bob Mottram wrote: 2008/6/3 Mike Tintner <[EMAIL PROTECTED]>: What are the implications for computing - how would it have to change - if the brain uses literal 3D maps - and they turn out to be a necessity? [Computers, I take it, can't currently produce them?] 2D mapping has been achievable for a while, but 3D mapping is a fairly recent phenomena because it's not until recent years that enough processing power has been available to handle this kind of task in anything like real time. To a large extent the DARPA urban challenge was all about 3D mapping and the accompanying sensor technologies needed to support it. DARPA challenges are mostly 2.5D, which is a much simpler problem. On the other hand, 3D mapping is pretty cheap if you have decent algorithms. The sensors are dirt cheap, so it is mostly knowing what to do with the data once you have it. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 1, 2008, at 7:27 PM, Mark Waser wrote: What if the brain truly is a conglomeration of many complex interacting pieces? Are we using the pedestrian sense of "complex" when talking about computational models and AI? Seems like an inappropriate overloading of its more technical and relevant definition. The digits of pi look complex to the naive observer, but they are most assuredly the product of a simple function. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 1, 2008, at 7:27 PM, Mark Waser wrote: Yeah. Those pesky chemicals like adrenaline etc. have absolutely no objective function whatsoever and absolutely zero effect on the functioning of the brain. Reading comprehension is clearly not your strong suit. Describe the function of adrenaline in the context of an abstract computational model. I'm not even arguing such a function does not exist, only that you are incapable of making the case that it has such a function. Don't bother responding. That's a fair point, since you were not actually addressing anything substantive. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 1, 2008, at 5:02 PM, Richard Loosemore wrote: But this statement is such a blatant contradiction of all the known facts about neurons, that I am surprised that abyone would try to defend it. Real neurons are complicated, and their actual functional role in the brain is still quite unknown. So you are asserting (1) that you know very little about neurons *and* (2) that they are fantastically complex devices at a computational model level. Remarkable that you are simultaneously deeply knowledgeable and ignorant at the same time. I see a lot of handwaving and cries of "Complex! Complex!" but I don't see a lot of evidence of that fact in the abstract computational sense. Even if one were to assert Penrosian magic, the result is pretty obviously simple in the theoretical sense, and we are back to algorithmic equivalence. Let's cut through the double-talk. Prove that neurons are not a simple set of functions in a hideously ugly and complex package. Obviously a lot of people think this is a real possibility, why there was a thread posted just this afternoon about how powerful and universal PEC bucket brigade logic is at predicting wetware neural network characteristics. And those were neuroscientists who presumably understand the wetware (though hadn't figured out you can implement that whole schemata in about a hundred lines of Python). You say all neuroscientists know neurons are complex, and yet we just had an article about how simple an effective computational description is that a lot of neuroscientists subscribe to. If that model is "simple", then neurons are ipso facto "simple". The description *is* the complexity from a computational standpoint. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 1, 2008, at 3:03 PM, Mark Waser wrote: I find it very interesting that you can't even answer a straight yes- or-no question without resorting to obscuring BS and inventing strawmen. By "obscuring BS and inventing strawmen" I assume you mean answers that do not fit into your narrow conceptual framework. Are you actually claiming that neurotransmitter levels are irrelevant or are you implementing them? Neurotransmitter levels are irrelevant. The function may or may not be, and the function would be directly implemented in the former case. Are you claiming that leakage along the axons and dendrites is irrelevant or are you modeling it? Axon and dendrite leakage is irrelevant. The function may or may not be, and the function would be directly implemented in the former case. Two simple questions. Two choices for each. Try answering them without the obscuring BS. What is it with you and the false dichotomies? What you fail to state is the reason I would implement any particular characteristic of wetware neurons; what function is being gained by doing so? Hint: if I can implement it in code I can pretty trivially ascertain its function by analysis. This is the whole "cargo cult AI" thing I was talking about. You insist it is valuable to add objectively functionless features, and you have hard time explaining what features are supposedly missing if we *don't* implement things that lack functionality. Sounds like a waste of time to me, unless you think these features do something magical. Step back a second and justify adding gee-gaws to the models first. For the neural characteristics that have a function, we don't need to copy them -- we can implement the functionality directly. Algorithmic equivalence and all that. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] news bit: Is this a unified theory of the brain? Do Bayesian statistics rule the brain?
Quotes like this make me shake my head: Friston’s results have earned praise for bringing together so many disparate strands of neuroscience. “It is quite certainly the most advanced conceptual framework regarding an application of these ideas to brain function in general,” says Wennekers. Marsel Mesulam, a cognitive neurologist from Northwestern University in Chicago, adds: “Friston’s work is pivotal. It resonates entirely with the sort of model that I would like to see emerge.” It is pretty funny to see neuroscientists congratulate themselves for inventing something that was already known in literature that they apparently don't read. Friston's "most advanced conceptual framework" has been around since at least the early 1990s in theoretical computer science and expressly considered in the context of AI and cognitive function. I was personally using predictive error math to reverse engineer neural structure function almost ten years ago (which sounds more useful than it actually is -- that is not the hard part). However, I will grant that nobody was really paying attention to that area of math at the time. And the biology guys have the nerve to say the computer scientists do not pay enough attention to neuroanatomy research. :-) Silly monkeys. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 1, 2008, at 1:44 PM, Mark Waser wrote: So . . . . given that the biological neurons have all this additional complexity that I have listed before, are you going to attempt to implement it or are you going to declare it as unnecessary (with the potential that, if you are wrong, you may doom your AGI effort before you ever get started)? You presume that all this "additional complexity" is actually complex in a meaningful way. Since it is relatively trivial to derive analogous behaviors and structure with understood function in other non-biological models (even if they look biological), I am not sure what to tell you. You are positing unimaginable complexity with a Disneyland of functionality, but you need nothing like that to get the same structure, behavior, and utility out of the system (*cough* Occam *cough*). This appears to be a variant of the "analog is fundamentally different from digital" category error. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 1, 2008, at 12:39 PM, Mark Waser wrote: What do you mean by computationally simple? Meaning there is a trivial set of functions and/or computational model that captures the utility. No need to accommodate patterns below the very high noise floor of wetware or which do not have a material computational purpose (e.g. side effects of biological maintenance). Explain to me how *you* construct a neural network that takes all of this into account. It depends on what you mean by "takes all this into account". Unless you are a biologist of some type, physical fidelity is a complete waste of time but you seem to leaning that way. The seconds hand of a mechanical clock may be driven by a complex dynamical system but that does not make it not equivalent in every important way to an utterly trivial solid-state counter. So in short, I would not construct a "neural network that takes all of this into account". I would construct a functionally equivalent computational model that coincidentally converges on an approximation of the structure and behavior of a biological neural network. I'm interested in AGI, not physiology. Obsessing over biological fidelity is the hallmark of cargo cult AI, the fervent hope that with sufficiently elaborate neural network theater the gods will deliver a mind. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 1, 2008, at 12:17 PM, Mark Waser wrote: Neurons are *NOT* simple. There are all sorts of physiological features that affect their behavior, etc. While I totally agree with your point about "Not only do you have to invent several new layers of abstraction, you also have to invent the control structures to manage all those abstractions and layers." -- as far as I'm concerned, ASSERTing clearly incorrect statements like "Neurons *are* simple" totally invalidates your credibility. Neurons are structurally complex but computationally simple within the usual constraints of computational information theory. Only the latter matters since (presumably) no one is attempting to build actual neurons to get the job done. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] Ideological Interactions Need to be Studied
On Jun 1, 2008, at 11:02 AM, Mark Waser wrote: One is elegance. It would be "oh, so nice" to find one idea that would solve the entire problem. After all, everyone knows that the single concept of "neurons" is what our brains are built upon . . . . The problem is that they then take an incredibly simplistic view of what a neuron is and then can't figure out why they can't get it to work or why they have to use radically different simplifications and formulas to make it work in different circumstances. Neurons *are* simple, analogous to a transistor. What they rarely seem to consider is how many different patterns and levels of pattern abstraction are required to make, say, a general purpose CPU design scale. You do not go from the 2,300 transistors of an Intel 4004 (nematode nervous system) to a modern CPU (reptilian nervous system) simply by slapping more transistors onto the 4004 design. Not only do you have to invent several new layers of abstraction, you also have to invent the control structures to manage all those abstractions and layers. All made out of simple transistors. I think the general problem with neural networks is not the concept of the neuron but the notion that you can scale up the utility of a simple neural network simply by slapping more neurons onto it. It would be lovely if it was that simple, but I do not think the evidence supports the notion that the design can be both simple and efficient (in the sense that evolution would find a design to be "efficient"). J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On May 27, 2008, at 7:00 AM, BillK wrote: As I understand it, Netcraft's results are based on web sites, or more precisely, hostnames, rather than actual web servers. This introduces a bias because some servers run a large number of low-volume (or zero volume) web sites. Of course, many sites use reverse proxies or other shenanigans that run many servers through a single IP, which would have the opposite bias. Accurately counting server boxes is difficult. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On May 26, 2008, at 6:46 PM, Mark Waser wrote: I have ~100% market share. Not sure how it is "two-to-one" or "dwindling", though I suppose it has nowhere to go but down. Huh? First *you* give me numbers of less then two to one and then you claim ~100%. How much did you drink at that barbecue? So you did not write: "That you have less than a two-to-one market share and it's dwindling?" I responded to your actual statement, not what you think you said. Your statement was barely coherent, so I interpreted it literally. Ignoring that you were arguing a transparently false dichotomy, I use neither Java nor .NET but I do know my market share (to the extent that it is "mine"). If you do not want to hear the correct answer, do not ask the question. I concern myself with big server apps So why are you even commenting on something obviously so far out of your realm? What part of OpenCog is not a server app? The big server apps in question hit at least 3 of the 4 "The framework includes" bullet points. Obviously it is very much in my realm. Why do you waste anyone's time? Do you find it amusing? It was not amusing until now. Froth away. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
Replying to myself, I'll let Mark have the last word since, after all, it is *his* project and not mine. :-) My only real quibble was with the notion that choosing .NET would not have a material impact on developer participation. I have to go man a barbecue and get some work done now. Cheers, J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On May 26, 2008, at 3:25 PM, Mark Waser wrote: Do you truly believe that search engine hits is proportional to the use of a language or is it just that the valid methods didn't you give the results that you wanted? There is no really authoritative source, that was just one method of many. So let's do job searches in Silicon Valley metro area on Dice and count the hits: Java server = 482 C server = 371 Perl server = 256 C++ server = 250 .NET server = 115 PHP server = 108 Python server = 101 Ruby server = 34 Yes, clearly there is *huge* demand for .NET server environments in Silicon Valley. Again, not scientific but it shows a trend. And while you do not like the TIOBE results, I cannot help but notice that these regional results roughly match theirs. And if we do a similar search in the Washington DC metro, we find that .NET does integer factors better but still fails to beat Java, C, et al. So your impression of .NETs ubiquity in your *own* area, while much better supported, is hardly true in any real sense there either. Not surprisingly, based on the above trend, Silicon Valley searches solely based on operating systems: Unix: 1681 Windows: 956 ...with a huge chunk of those Windows jobs being for writing drivers apparently. Funny that I should be under the impression that the vast majority of server application targets in Silicon Valley are Unix-based. Do you have *any* viable facts to back up your silly operating system assertions? Which "silly" operating system assertions? That most development targets in Silicon Valley are web or Unix? That the data centers here have virtually no Windows servers in them? You assert that is silly, but you have not provided any "viable facts" to support your assertion. Even to the extent that my assertions are anecdotal, I have worked in the data center operations of many, many companies in Silicon Valley including some of the largest ones, and if you peek under the hood is just lots and lots of Linux servers for the most part. But you would not know that sitting in Virginia I suppose. Just because you are not comfortable with Unix does not mean there is not a huge swath of the industry that is very comfortable with it and has it as their sole development target. In fact, in those rare cases where we needed a bit of Windows software written over the years, we contracted out to other locales. The talent-pool in the Valley is definitely Unix- and web-centric, and the ubiquity of the Mac as a development environment reflects that. Do you so desperately need to rationalize the .NET platform for an open source project that you refuse to accept what is plainly obvious to those actually on the ground? If I were you, I would have simply stated that my desire to use a particular narrow platform for a variety of reasons outweighed my need to maximize developer reach and be done with it. If you had stated it like that, I would have no argument. Cheers, J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On May 26, 2008, at 1:16 PM, Jim Bromer wrote: Would you please direct me to open source project web sites that may be of interest to AI projecteers, and a C++ compiler to use with them. I never found any comments on a good compiler to use on a Windows XP system (other than the microsoft compiler of course.) I am also looking for a web site that also has some introductory material on how one goes about working on a listed open source project. The answer is that using the Microsoft compiler is preferable for some purposes, but the GCC stuff works there. Portability is not an issue I usually have to worry about, so I am not exactly an expert on the ins and outs of moving C++ between Unix and Windows. As long as you are not doing a lot of system calls, it is fairly straightforward. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
xtremely productive, maintainable, and scales up well. There are still a few parts of the US where Perl is going strong, but not in the Valley. The TIOBE programming language popularity index shows Java in the #1 spot followed by C in #2. Even Python edges out C#, which comes in at #8. A lot of big companies in Silicon Valley use the three languages mentioned above as their official development languages, so it is not surprising that they are primary choices in Silicon Valley. The important languages I glossed over were actually things like PHP and Ruby, neither of which are C#. I have yet to see anyone attempt to deny my claim about the relative development speed on .Net vs. anything else. Supposed "development speed" does not do you much good without sufficient developers, even if I allowed that it were true. As a practical matter if you can specify a problem I can always find a platform that has better development speed than .NET -- this is really quite irrelevant and other things are more important in the big picture. The language/environment is a secondary concern to the developer pool because you could develop this project in *any* language. The difference in overhead costs intrinsic to the environment are nominal. I don't like Java myself, but I think a better argument can be made for it *in this instance* relative to .NET because language features are not that important at the end of the day. If you were doing a closed shop project then .NET would be very arguably a superior choice. So, why do you believe that all these developers are staying away from the superior choice? Why aren't the smarter ones defecting? Are you sure that they aren't? Are you sure you want that huge developer pool of those who aren't smart enough to defect? Most of your arguments are based on the assumption that most of these developers are Windows developers who just happen to be stuck working on Java. Often they are Unix developers who happen to be stuck working on Java. If they stopped using Java, .NET would not be their logical next choice. There are practical economic reasons Java is used so much, most notably its ubiquity and the fact that it works well on Unix. There is a trend moving away from Java for some web apps, but it is toward languages like Python, Ruby, and PHP, which also run on Unix very well. If you hate Java, there are other environments with a better feature Where did *that* come from. I don't hate Java. It's just seriously sub-optimal so I don't waste my time with it. *shrug* Maybe I was just projecting my distaste for Java. I don't like it, but I understand the rationale behind using it. I dunno, the obsession with a very particular and narrow platform systems misplaced and inappropriate for a project like this. The goal is (hopefully) *not* to select a platform you like and then rationalize every other decision around that. You mean like choosing the platform solely based upon the size of your developer pool and ignoring what *you* acknowledge as superior features on another platform? You agree with all of my technical reasons and then accuse me of rationalization? I agree with your technical arguments re: Java versus .NET in the abstract. My point is that these are minor factors in the overall platform decision. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On May 26, 2008, at 11:53 AM, Bob Mottram wrote: On Linux the performance of 3D distributed particle SLAM (a CPU intensive task) running on the Mono .NET (version 2) runtime is marginally faster than the same code running on Windows using the MS runtime, but only by a few milliseconds. Performance benchmarks are very similar to the same algorithms written in C++ and compiled with gcc. Thankfully, I do not think anyone is really arguing performance differences. It is mostly seems to be about development environments and portability. Every language can get the job done reasonably well these days. For example, while Java itself is an ugly re-imagining of C ++, several other programming languages have been ported to the virtual machine and seamlessly integrated into Java, similar to .NET. It is six of one and a half dozen of the other. The advantages being advertised for C# (i.e. new functional programming features) only apply to .NET 3 or above, which isn't available on GNU/Linux systems and so is of no interest to me at this point. That's the real issue here as I see it. Unix-like systems still dominate the server market broadly, and a lot of people primarily develop for that market. Despite the efforts of Mono, .NET seems to be permanently marginalized as a Windows-only environment. The ubiquity of Windows on the desktop does not translate into ubiquity as a development target. For many types of very technical development that may be very relevant here, Unix is overwhelmingly the real-world development target and where you will want to pull your developers from. Arguments about programming languages are a popular topic on AI forums, but usually generate more heat than light. Yup. Programming language selection is largely irrelevant to producing AI, and is secondary to other practical concerns. If programming language made a material difference, we'd already have AI. Cheers, J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On May 26, 2008, at 8:41 AM, Mark Waser wrote: C# may have advantages over Java, but it doesn't mean that these advantages are particularly relevant for a particular project. Then make project-specific assertions. The fact that functional programming is an integral part of C# is huge for AGI. (Your turn to make a valid point :-) Too bad C# is not an integral part of many developer environments or most developer experience. It is like selecting Objective-C. Functional programming is a really lame hobby horse because it is well- supported in most of the plausible alternative environments (including C++), if not as a direct part of the language then as a trivial add- on. You are trading a minor nuisance for a showstopper. In other words, this is a pretty crap justification for using .NET. Where is the fabulous .NET support and development environment for MacOS and Linux? Selecting C# really is like selecting Objective-C except going the opposite direction (since Objective-C is broadly supported in Unix-like environments, and even on Windows poorly), and Objective-C is actually a pretty nice language with a solid and growing developer pool. Hell, on the MacOS platform Objective-C even comes with a really deep and slick set of frameworks and libraries that allow you to implement many very advanced capabilities effortlessly. Not that I am suggesting actually using Objective-C; it shares the exact same problems as C#, and I would use a similar criticism. There are people on this mailing list using C# and Objective-C for their projects, but they are closed shops and so the selection is more easily rationalized. For open source projects, ideal environments play second fiddle to broad language support. Painless portability is the reason C is often selected over C++ for open source projects -- universality is that important. Cheers, J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On May 26, 2008, at 4:42 AM, Mark Waser wrote: There has been a large upswing in the number of MacOS laptops. At the same time, there has been an equally large reversal of the correlation with Unix-based back-ends. Macs are being picked up because of their engineering, power-up times, ease of use and coolness factor -- and the fact that they Terminal Services to work desktops just as well as any windows machine. Where do you live, if you do not mind me asking? The preference for server environments is very much a local phenomenon. Using California as an example, in Los Angeles there is a strong preference for Windows systems, but in Silicon Valley you will find that Unix is pervasive. .NET may be ubiquitous in abstract because it is associated with Windows, but if you actually look at some rather important tech centers like Silicon Valley, there is not a Windows server in sight. The dominance of Unix-based systems there is so complete that it is not even a contest any more. You are apparently under the impression that this is not true, but if you continue with that assumption you will systematically exclude a vast and very talented developer pool that has zero interest or experience developing in .NET even if they are using a Windows workstation. A lot of business in Europe specifically excludes .NET as a development target for similar pragmatic reasons. And developing .NET is going to suck on a non- Windows workstation, eliminating one of the major advantages you tout. To be honest, I do not know of anyone that uses a Mac that is using it for .NET development -- total impedance mismatch. To use Silicon Valley as an example, C/C++, Java, and Python will give you about 90% coverage of the developer pool. The .NET languages are in the residue. In Bangalore, .NET is a major percentage of the developer pool. Which is most likely to usefully contribute to your project, programming languages aside? It sounds an awful lot like you are simply trying to rationalize your personal preference for programming language/environment. And what is the value proposition of Java over any other language? It has no unique features. It's development is lagging. It's developers are defecting (again, look at the statistics). It's fragmenting just like Unix so it certainly isn't as portable as claimed. The value proposition of Java is a deep pool of technically proficient hackers know it and it works on all the platforms many such people prefer. MacOS has a C/C++, Python, and Java development environment out of the box (among other less common languages), but no .NET. Linux has similar coverage out of the box. By selecting .NET you have tacitly excluded most developers in Silicon Valley, and a huge number in Europe and many other countries. Java casts a much wider net even if it is an inferior environment. The language/environment is a secondary concern to the developer pool because you could develop this project in *any* language. The difference in overhead costs intrinsic to the environment are nominal. I don't like Java myself, but I think a better argument can be made for it *in this instance* relative to .NET because language features are not that important at the end of the day. If you were doing a closed shop project then .NET would be very arguably a superior choice. If you hate Java, there are other environments with a better feature set *and* much broader portability. A popular one is C + Python, which allows you to combine very pretty syntactic sugar with unfettered performance and system access. My point is not that Java is better than .NET, but that .NET is a really poor choice if you are trying to rope in a large developer talent pool. I dunno, the obsession with a very particular and narrow platform systems misplaced and inappropriate for a project like this. The goal is (hopefully) *not* to select a platform you like and then rationalize every other decision around that. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: Competitive message routing protocol (was Re: [agi] Deliberative vs Spatial intelligence)
On May 1, 2008, at 10:06 AM, Matt Mahoney wrote: --- "J. Andrew Rogers" <[EMAIL PROTECTED]> wrote: Your model above tacitly predicates its optimality on a naive MCP strategy, but is not particularly well-suited for it. In short, this means that you are assuming that the aggregate latency function for a transaction over the network is a close proxy for the transaction cost. At one time this might have been a reasonable assumption, but it becomes less true every year. That's true in my thesis but I dropped it in my CMR proposal. Now I assume that peers operate in a hostile environment. A message could be anything. The protocol has to work even over unreliable UDP with forged source IP addresses. The problem is sort of like building a brain out of neurons that are trying to kill each other. (Yes, late. I do not have much free time.) A brain where all the neurons are out to kill each other is a proper metaphor for the design problem. In real protocols, every time someone posited benevolence for some aspect it was promptly exploited. In my thesis, I asked whether it was possible even in theory to build a large scale distributed index. None existed in 1997 and none exists today. The best known examples of internet wide databases were USENET, which uses O(n^2) storage, and DNS, which is O(n) (assuming it grows in depth with constant branching factor, although it doesn't really) but is vulnerable at the root servers. Centralized search engines are also O(n^2) because you need O(n) servers for n clients. This creates an incentive for engines to merge to save resources, resulting in a monopoly. (Who has the resources to compete with Google?) There is an increasingly strong political incentive (between countries) to create distributed indexes, but quite frankly the technology does not exist. This was something I studied in earnest when various governments started demanding such guarantees. To the best of my knowledge, we do not have mathematics that can support the guarantees desired, though decentralized indexes are certainly practical if one ignores certain considerations that are politically important. Something to understand about the big server clusters: as commonly implemented, the online server cluster is independent of the content generation cluster. Queries may be very cheap to serve even if the aggregation and analytics process is expensive. Compute a result once and serve it to the world a thousand times. The real problems occur when the data set is not sufficiently static that this trick is plausible. Fortunately, no one has noticed the man behind the curtain (yet). Losing to Google is predicated on following their path, and they occupy a space where the computer science is transparently inadequate. It does not take much of a qualitative shift in the market to kill a company in that position. There is plenty of vulnerability left in the market. I would argue, from a business perspective, is that most of the value with respect to distribution is in the metadata protocol, virtually all of which are based on naive designs that ignore literature in practice. A really strong metadata protocol that could be standardized would generate a hell of a lot of value. Past that, whoever controls the essential data under that protocol would win, and for better or worse, Google is largely not responding to this. There are many types of data they have no capacity to handle in bulk. This is not so much a criticism of Google but an observation about their actual behavior. Cheers, J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On May 26, 2008, at 5:52 PM, Mark Waser wrote: That you have less than a two-to-one market share and it's dwindling? I have ~100% market share. Not sure how it is "two-to-one" or "dwindling", though I suppose it has nowhere to go but down. That technically .Net has blown past you and the gap only shows signs of widening? I concern myself with big server apps, and I am not sure what this .NET gap is. Which Silicon Valley companies are developing their server infrastructure using .NET? Other than Microsoft (presumably), I cannot think of any. When companies want server bindings and drivers, they ask for C++, Java, and (god help us) PHP. I have never had anyone anywhere in government or industry ask for .NET. I am sure they exist and it will happen eventually but the avalanche of demand is not there, probably because virtually no one uses .NET on Linux. That when Mono reaches the next version, you're going to switch? Seems unlikely, since it does not offer anything of value for anything I might do. C is faster and more scalable for server engines, particularly for server clusters; if you are going to write that much unmanaged code, you might as well bind it to a super-productive language like Python. Is Microsoft porting Visual Studio to Unix/ Linux in the near future? I already get a really fancy Unix development environment from Apple for free, though it does not support .NET. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
Some not-quite-random observations that hopefully injects some moderation: - There are a number of good arguments for using C over C++, not the least of which is that it is dead simple to implement very efficient C bindings into much friendlier languages that hide the fact that it is still mostly C. There are a lot more problems doing this in C++ than C. If you are doing a project end-to-end in one language though, C++ manages the complexity better than C (though I would observe that some very large yet very tidy and understandable code bases are in C e.g. PostgreSQL). In the big picture, this becomes a detail depending on relatively unimportant design choices. - At every technical conference on a variety of non-platform-specific topics I have been to in the last year that was full of people that actually work on code, I and many others have noticed that at least half the people attending were using MacOS laptops. This is very strongly correlated with Unix-based server back-ends, usually Linux out in the real world these days. The great thing about MacOS X as a developer is that it is Unix, and so there is a good impedance match between the developer desktop and the production cluster. Using a .NET technology for anything is tacitly excluding a huge swath of talent and a significant portion of the developer market. This is particularly true if we are talking about server-like or engine-like code, in which .NET is very much a minority player. - Selecting any narrow platform technology (like .NET or Objective-C) only really makes sense if there is no intention of widely disseminating or collaborating with the code. Having nicer libraries or syntactic sugar does not do a hell of a lot of good if you cannot find enough competent developers to make that feature provide return on the investment -- killer libraries and environment save time, not developer talent. This has been often cited as a key failure of the Ruby community that has caused many projects to move away from it: lots of hype and interest but there is a dearth of top-quality developers that actually choose to work with it, making complex projects effectively non-viable for lack of appropriate talent. I honestly do not give a crap about the subject being argued, but if the goal is to have decent environment support *and* cast the widest possible with respect to developer talent, the obvious choice is actually Java. This coming from someone who does not even like Java and thinks .NET is a better designed environment; the differences between environments is noise in the big picture, but the differences in the breadth and depth of developer talent is not. If the object of this project is *not* to engage the maximum amount of developer talent then the point is moot and it is hard to figure out why it is being argued at all. In short, if it is a closed shop project not meant for wide dissemination, then the benefits of .NET significantly outweigh the benefits of C/C++ (unless performance is paramount) and is a defensible choice. If it is intended to be an open source project that maximizes participation, I cannot imagine why anyone would choose .NET over Java or even C unless they were deluded about the distribution of developer talent on the wild and wooly Internet. The right tool for the job, and all that. Cheers, J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: Competitive message routing protocol (was Re: [agi] Deliberative vs Spatial intelligence)
neral (i.e. not restricted to a set of data with carefully tailored characteristics) will distribute beyond a few dozen nodes, and it gets worse fast when you add dimensions. The new solution I mention above is fully general and scales extremely well in a distributed, decentralized environment, which seemed to be the issue that needed solving. The exception to this is if your spatial index is built once and never modified (i.e. read-only) in which case it is possible to have one that is general and which scales to modest size. Its a real mess, and scalability is something like a five- or six-axis space when talking about these kinds of algorithms. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Deliberative vs Spatial intelligence
On Apr 30, 2008, at 11:41 AM, Matt Mahoney wrote: By distributing the problem across the internet. AGI can be divided into lots of specialized experts and a network for getting messages to the right experts. http://www.mattmahoney.net/agi.html There are a few problems with your model that need to be fixed before it is legitimately viable, though you do acknowledge some of them in the paper: 1.) The protocol design is naive and will not scale up to the level you think it will, simplifying away by assumption topology characteristics where deviations from the assumption will have a major impact. There are no general, computable solution to the underlying issues (neither in literature nor in unpublished research that I know of), and you gloss over or do not consider problems that would have a pathological expression if you actually tried to build it. This is an important and active area of mathematics research in a couple different fields. 2.) There is nothing in published literature that will do the kind of indexing you want to do in the spatial domain, but it is possible in theory. For your purposes in the broadest sense, things like kD-trees will drop dead for pretty trivial systems, never mind for something ambitious. On the other hand, generalized distribution with O(n) storage complexity was solved last year which may or may not address your issues. I think the real killer will be #1, as it has the hallmarks of being a Hard Problem. We had a really simple approximation for routing and resource optimization that worked well in the early days of network applications, but modern network application models break the assumptions. Perhaps relevant for you, a lot of the active theoretical research in this area surrounds the problem of massive distributed and decentralized spatial structures, though not for the purposes you are thinking about. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Deliberative vs Spatial intelligence
On Apr 29, 2008, at 1:46 AM, Russell Wallace wrote: Suppose we say there are two types of intelligence (not in any rigorous sense, just in broad classification): Deliberative. Able to prove theorems, solve the Busy Beaver problem for small N, write and prove properties of small functions, construct cellular automata computers for small functions, derive small functions from specifications, notice what it's doing, accept symbolic heuristics to improve its efficiency, think about said heuristics etc. Symbolic intelligence that can, in some crude sense, copy some of the things humans can symbolically do. Spatial. Able to perceive patterns in two or three dimensions. Can be used, with mods, for a robot visual cortex; image recognition; given a series of photographs of a landmark from varying viewpoints, can derive a 3d model and backtrack that to the 2d image visible from any other viewpoint; can pathfind units around a map in a video game; can make much better than random guesses as to likely folds of a new protein chain; can animate a cartoon from the description "cat sits on mat". I will take a third position and point out that there is no real distinction between these two categories, or at least if there is you are doing it wrong. One of the amusing and fruitless patterns of behavior in the AI community is the incessant categorization of various processes into nominally distinct buckets in the absence of a theoretically justifiable reason for doing so. The above is such an example. As a general comment, the computer science literature on the D- oriented side of things is *much* deeper than the S-oriented side of things, and the literature that theoretically integrates the two is thin on the ground indeed. This is probably a reflection of the observation that competency at D is far more widely distributed than S, or at least that far more competent people have worked on D than on S. When I originally switched to the "Spatial" side of things, one of the first things I noticed was that the backing theory and literature was medieval compared to what you call the "Deliberative" side. On the downside that meant that there was not a lot to work from, but on the upside that also meant there was still a fair amount of low-hanging fruit left to be picked. Spatial can scale extremely well in a very general sense, but you'll have to do some original work to get there because your off-the-shelf computer science will leave you wanting. It should actually be pretty obvious, even without really hammering out the theory, how the "Deliberative" part can be trivially expressed in a "Spatial" solution -- the former can be correctly viewed as a narrow instance of the latter. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Language learning
On Apr 22, 2008, at 11:55 PM, YKY (Yan King Yin) wrote: There is no doubt that learning new languages at an older age is much more difficult than younger. I seem to recall that recent research does not support this assertion. Rate of language learning is essentially the same for both adults and children and is a function of the amount of time spent trying to learn it. The apparent absolute differences in rate of learning turned out to be attributable to adults spending a smaller percentage of their time learning a new language than children on average, which gave the false impression that adults learn languages more slowly. I am too lazy to dig up cites at the moment, but I definitely remember discussions of this research in the not too distant past. Cheers, J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Language learning
On Apr 22, 2008, at 7:17 AM, Mark Waser wrote: In my experience it is not so much that they sound the same but that we don't know how to say them (in terms of mouth mechanics) such that we can isolate the difference between sounds that would have been in the range of a single phoneme in English. No. We have a Thai exchange student this year. There are words that she swears are different that sound to me (and the rest of the family) to be exactly the same. Precisely my point. They sound exactly the same until you understand the mechanics of the sound generation, at which point you have a frame of reference for recognizing the differences. The differences are there, you are just not using them as a means of discernment because you have no knowledge of which differences are important for discernment. This is why it is futile and silly to use sound examples to teach someone a difference that we have already established they cannot isolate. On the other hand, the phoneme generation mechanics are relatively unambiguous. I could never hear many sounds until I figured out what they were doing to create the sound that was different from how I created the sound. Once I figured that out, it became relatively easy to hear the difference because I knew what to listen for. Austroasiatic languages (like Thai) tend to be particularly difficult for native English speakers because they tend to rely heavily on complex usage of all the possible bits that English speakers do not. However, having delved fairly deeply in one such language myself, it is easier than it seems at first once you figure it out. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] WHAT ARE THE MISSING CONCEPTUAL PIECES IN AGI? --- recent input and responses
On Apr 21, 2008, at 6:53 PM, Richard Loosemore wrote: I have been trying to understand the relationship between theoretical models of thought (both natural and artificial) since at least 1980, and one thing I have noticed is that people devise theoretical structures that are based on the assumption that intelligence is not complex but then they use these structures in such a way that the resulting system is almost always complex. This is easily explained by the obvious fact that the definition of "complex" varies considerably across relevant populations, exacerbated in the case of AGI -- where it is arguably a germane element -- because many (most?) researchers are using "complex" in a colloquial (read: meaningless) sense rather than one of its more rigorously defined senses, of which there are a few interesting ones. Most arguments and disagreements over "complexity" are fundamentally about the strict definition of the term, or the complete absence thereof. The arguments tend to evaporate if everyone is forced to unambiguously define such terms, but where is the fun in that. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Re: Language learning
On Apr 21, 2008, at 12:53 PM, Matt Mahoney wrote: Like English speakers learning Hindu cannot learn to speak the 3 different versions of the 'k' sound because they sound the same. In my experience it is not so much that they sound the same but that we don't know how to say them (in terms of mouth mechanics) such that we can isolate the difference between sounds that would have been in the range of a single phoneme in English. I had that problem learning other 'k' sounds (not Hindi though). I figured that out when trying to teach people the different sounds in the range of 't' and 'th' that have languages that contain only one (or languages which have one that is neither English 't' nor 'th', like some Asian languages). My problem learning new sounds was not from an inability to hear the difference but finding someone who could explain what the difference was, which can really only be usefully described by the difference in mechanics of generating the sound (most people attempt to explain by example, which is clearly useless). It has less to do with not being able to hear the difference and more to do with not knowing which differences are important and which are noise. The part of the brain responsible for auditory phoneme recognition becomes read-only by age 6. So we all speak foreign languages learned at later ages with an accent. This appears trivially falsifiable. While I know it seems true in many cases, I know of a few people who came to the US from Asia in the mid-teens who speak perfect accent-free American English that was learned when they moved to the US. Quite a phoneme change from tonal East Asian languages, and you would never know they were not born here. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] database access fast enough?
On Apr 17, 2008, at 3:32 PM, YKY (Yan King Yin) wrote: Disk access rate is ~10 times faster than ethernet access rate. IMO, if RAM is not enough the next thing to turn to should be the harddisk. Eh? Ethernet latency is sub-millisecond, and in a highly tuned system approaches the 10 microsecond range for something local. Much, much faster than disk if the remote node has your data in RAM and is relatively local. Note that "relatively local" can mean geographically regional. The round-trip RAM access time from my machine to a machine on the other side of town is a fraction of millisecond over the Internet connection (not hypothetical, actually measured at ~400 microseconds). I wish disk access was even remotely that good. And this was with inexpensive Gigabit Ethernet. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] database access fast enough?
On Apr 17, 2008, at 12:26 PM, Mark Waser wrote: Actually, it's far worse than that. For serious systems, most of the heavy lifting is done inside the database with stored procedures which are not standard AT ALL. SQL is reasonably easy to port. Stored procedures that do a lot of work are not. The standard is SQL/PSM, which looks similar to Oracle's PL/SQL (and PostgreSQL's pl/pgsql). As a practical matter, support is not consistent enough or widespread enough for it to be entirely usable for purposes of portability though it is getting better. To be fair, full SQL/PSM support will not be core in PostgreSQL until the next release. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] database access fast enough?
On Apr 17, 2008, at 12:20 PM, Mark Waser wrote: It has always been posssible to tweak any of the databases to the other's transactional model. Eh? Choices in concurrency control and scheduling run very deep in a database engine, with ramifications that cascade through every other part of the system. Equivalent transaction isolation levels can behave very different in practice depending on the internal transaction representation and management model. You cannot turn off these side-effects, and you cannot "tweak" a non-MVCC-ish model to behave like an MVCC-ish model at runtime in any way that matters. Second of all, it was not a weakness -- it was a deliberate choice of optimization -- it was a choice of OLAP over OLTP (and, let's be honest, for most databases on limited memory machines with low OLTP requirements, this was the correct choice until ballooning memories made the reverse true). The rise of the Internet, with its massive OLTP load characteristic, kind of settled the issue. It is true though that Oracle-like OLTP monsters have significantly higher resource overhead for storing the same set of records. These days it is concurrency bottlenecks that will kill you. So, is your claim that Oracle distributes better than Microsoft? If so, why? Very mature implementation of the concepts, and almost every conceivable mechanism and model for doing it is hidden under the hood. Remember, they started introducing the relevant concepts ages ago in Oracle 7, though in practice it was mostly unusable until relatively recently. Consequently, their implementation is easily the most general in that it works moderately well across the broadest number of use cases because they've been tweaking that aspect for years. Other commercial implementations tend to only work for a much narrower set of use cases. In short, Oracle has a long head start. There are new transactional architectures in academia that should work better in a modern distributed environment than any of the current commercial adaptations of classical architectures to distributed environments. And PostgreSQL will probably implement them long before Oracle or MS. Ironically, a specific design decision that has created a fair amount of argument for years makes PostgreSQL the engine starting from the closest design point. PostgreSQL does not support threading and only uses a single process per query execution, originally for portability and data safety reasons -- the extreme hackability would be difficult to do otherwise. This made certain types of trivial parallelism for OLAP difficult. On the other hand, it has had distributed lock functionality for a number of versions now. If you look at newer models explicitly designed to make transactional database scale better across distributed systems, you find that they are built on a design requirement of single processes per resource, strict access serialization, no local parallelism, and distributed locks. Which is not that far removed from where PostgreSQL is today, if you remove massive local concurrency support and its high overhead. There are a number of outfits (see www.greenplum.com for a very advanced implementation) that have hacked PostgreSQL to scale across very large clusters for OLAP by essentially making the necessary tweaks to approximate these types of models. The next step would be to rip out a lot of expensive bits based on classical design assumptions that make distributed write loads scale poorly. In a sense, a design choice that has traditionally put some limits on scaling PostgreSQL for OLAP put it in exactly the right place to make implementation of next-generation architectures as natural of an evolution as can be expected in this case. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] database access fast enough?
PostgreSQL server simply by copy the data and rebinding a WINS name or an IP address, I would be in hog heaven even if support wasn't absolutely guaranteed since I could always switch back. Given that there's a huge transition cost (changing scripts, procedures, etc.), I can't get *ANY* agreement for the thought of switching (and I'm sure that there are *MANY* more in my circumstances). The only corporate database that relatively easily ports back and forth with PostgreSQL is Oracle. Nonetheless, a number of people have ported applications to PostgreSQL from MS-SQL with good results; questions about porting nuances come up regularly on the PostgreSQL mailing lists. Beyond your basic ANSI compliance, database portability only sort of exists. Inevitably people use non-standard platform features that expose the specific capabilities of the engine being used to maximize performance. As a practical matter, you pick a database platform and stick with it as long as is reasonably possible. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] database access fast enough?
On Apr 17, 2008, at 6:07 AM, Mark Waser wrote: I have to laugh at your total avoidance of Microsoft SQL Server which is arguably faster and better scaling for truly mixed use than everything except possibly Oracle on ordinary hardware; which is much easier to use than Oracle; and which is the easiest to actually put *GOOD* code in the database engine itself (particularly when compared to Oracle's *REALLY* poor java imitation). Discussing SQL Server does not generalize well in that they reimplement the core engine design with almost every release once they realize they hosed the design with the last release. For example, up until SQL Server 2005 the transaction engine was weak such that PostgreSQL could spank it in transaction throughput -- in 2005 they switched to a transaction model more like PostgreSQL and Oracle and gained some parity. SQL Server still does not really distribute all that easily, unlike Oracle or PostgreSQL. SQL Server versions before the current two year old one were pretty much dogs in a lot of ways. The most recent version is as you state a pretty solid database engine. Oracle is a major pain in the ass to use but does scale well, though for many OLTP loads it is barely faster than PostgreSQL these days. If putting your code in the engine is the goal, PostgreSQL wins by a country mile. The entire engine from front to back is deeply hackable with very clean APIs and you can even safely bind binary code into the engine at runtime. That the transaction engine scales quite well is just a bonus. People have already written hooks for a dozen languages into it. I've written performance-sensitive customizations of PostgreSQL in the past, and for purposes like that it can often be much faster than the commercial alternatives, as the alternatives tend to be relatively feature poor and shallow when it comes to engine customization. Making deep and very flexible customization a safe core feature was a design decision tradeoff in PostgreSQL that is somewhat unique to it. You can do a lot of really cool software implementation tricks with it that Oracle and SQL Server do not do. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] database access fast enough?
On Apr 17, 2008, at 2:50 AM, YKY (Yan King Yin) wrote: ARC (Adaptive Cache Replacement) seems to be one of the most popular methods, and it's based on keeping track of "frequently used" and "recently used". Unfortunately, for AGI / inference purposes, those may not be the right optimization objectives. It is a cache replacement algorithm, what would be a "right optimization objective" for such an algorithm? There is a lot of cleverness in the use of the cache to maximize cache efficiency beyond the cache replacement algorithm -- it is one of the most heavily engineered parts of a database engine. As an FYI, ARC is patented by IBM. PostgreSQL uses a different but similar algorithm that is indistinguishable from ARC in benchmarks (having implemented ARC briefly, not realizing that it was patented). The requirement of inference is that we need to access a lot of *different* nodes, but the same nodes may not be required many times. Perhaps what we need is to *bundle* up nodes that are associated with each other, so we can read a whole block of nodes with 1 disk access. This requires a very special type of storage organization -- it seems that existing DBMSs don't have it =( Again, most good database engines can do this, as it is a standard access pattern for databases, and most databases can solve this problem multiple ways. As an example, clustering and index- organization features in databases address your issue here. It is pretty difficult to generate an access pattern use case that they cannot be optimized for with a good database engine. They are very densely engineered pieces of software, designed to be very fast while scaling well in multiple dimensions and adapting to varying workloads. On the other hand, if your use case is simple enough you can gain some significant speed for modest effort by writing your own engine that is purpose-built to be optimized for your needs. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] database access fast enough?
On Apr 16, 2008, at 9:51 PM, YKY (Yan King Yin) wrote: Typically we need to retrieve many nodes from the DB to do inference. The nodes may be scattered around the DB. So it may require *many* disk accesses. My impression is that most DBMS are optimized for complex queries but not for large numbers of simple retrievals -- am I correct about this? No, you are not correct about this. All good database engines use a combination of clever adaptive cache replacement algorithms (read: keeps stuff you are most likely to access next in RAM) and cost-based optimization (read: optimizes performance by adaptively selecting query execution algorithms based on measured resource access costs) to optimize performance across a broad range of use cases. For highly regular access patterns (read: similar query types and complexity), the engine will converge on very efficient access patterns and resource management that match this usage. For irregular access patterns, it will attempt to dynamically select the best options given recent access history and resource cost statistics -- not always the best result (on occasion hand optimization could do better), but more likely to produce good results than simpler rule-based optimization on average. Note that by "good database engine" I am talking engines that actually support these kinds of tightly integrated and adaptive management features: Oracle, DB2, PostgreSQL, et al. This does *not* include MySQL, which is a naive and relatively non-adaptive engine, and which scales much worse and is generally slower than PostgreSQL anyway if you are looking for a free open source solution. I would also point out that different engines are optimized for different use cases. For example, while Oracle and PostgreSQL share the same transaction model, Oracle design decisions optimized for massive numbers of small concurrent update transactions and PostgreSQL design decisions optimized for massive numbers of small concurrent insert/delete transaction. Databases based on other transaction models, such as IBM's DB2, sacrifice extreme write concurrency for superior read-only performance. There are unavoidable tradeoffs with such things, so the market has a diverse ecology of engines that have chosen a different set of tradeoffs and buyers should be aware of what these tradeoffs are if scalable performance is a criteria. J. Andrew Rogers --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
[agi] META: email format (was Why Hugo de Garis is WRONG!)
Hi Mark, Could you *please* not send HTML email? Ignoring that it is generally considered poor netiquette, and for good reason, it frequently gets turned into barely readable hash by even the most modern email clients. I am using Mail.app 2.0 on OSX 10.5 which handles rendering better than most, and most HTML email is *still* generally rendered as far uglier and less readable than plaintext email. Given that HTML email does not add anything substantive could we please stick to plaintext for the sake of communication? Thanks, J. Andrew Rogers On Mar 26, 2008, at 11:37 AM, Mark Waser wrote: Before swatting at one of those pesky flies that come out as the days lengthen and the temperature rises, one should probably think twice. A University of Missouri researcher has found, through the study of Drosophila (a type of fruit fly), that by manipulating levels of certain compounds associated with the "circuitry" of the brain, key genes related to memory can be isolated and tested. The results of the study may benefit human patients suffering from Parkinson's disease and could eventually lead to discoveries in the treatment of depression. http://www.machineslikeus.com/cms/news/flys-small-brain-may-benefit-humans Mark Vision/Slogan -- Friendliness: The Ice-9 of Ethics and Ultimate in Self-Interest agi | Archives | Modify Your Subscription --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244&id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
Re: [agi] RE:P2P and/or communal AGI development [WAS Hacker intelligence level...]
On Dec 3, 2007, at 12:52 PM, John G. Rose wrote: For some lucky cable folks the BW is getting ready to increase soon: http://arstechnica.com/news.ars/post/20071130-docsis-3-0-possible-100mbps-sp eeds-coming-to-some-comcast-users-in-2008.html I'm yet to fully understand the limitations of a P2P based AGI design or the augmentational ability of a public P2P network on a private P2P network constructed for AGI. I would count out P2P AGI so quickly. Distributed algorithms tend to be far more sensitivity to latency than bandwidth, except to the extent that low bandwidth induces latency. As a practical matter, the latency floor of P2P is so high that most algorithms would run far faster on a small number of local machines than a large number of geographically distributed machines. There is a reason people interested in high-performance computing tend to spend more on their interconnect than their compute nodes. J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=71602529-7ec12e
Re: [agi] Where are the women?
nd if what you assert is true, where are all the female COBOL programmers? wait, why do I have to manipulate complex multi-dimensional graphs in my head? I'm a programmer and I've never done that before. I'd be interested in knowing why you think this skill is important, but I can guarantee you many programmers never do it. Uh, what kind of programming do you do that you would assume that almost the entire software universe is working in some kind of linear scripting environment? Any code design of any significant complexity is generally *always* a complex multi-dimensional graph, particularly true if scalability matters. It is not something you do consciously, but any programmer that works on sufficiently complex software systems is doing it. If you are gluing all the fancy complex code engines together with a few glue bits (a lot of web apps are like this) then I suppose it is possible to avoid it because some other programmer did the heavy lifting. Which is a good thing because most programmers cannot be trusted to produce theoretically complex software elements -- an observation, not a criticism. Communication is necessary for programmers? I'd say useful, but not necessary. What on earth do you think code is? The only difference between code and people-talk is that code requires precision and non-ambiguity since incorrect results are generally considered unacceptable. Why do you think it is innate? Because I've never seen anyone learn it, ever; experience changes a lot, but the ability to handle complex abstract models doesn't seem to. I've known many software engineers with careers that span decades and bucketloads of experience that really don't grok graphs beyond a certain complexity -- it is a bit like you reach a certain description threshold where pushing more bits into the model makes other bits fall out. That threshold varies from individual to individual, and it is difficult to not notice that the correlation between really bright software designers and people who are quite apparently able to atypically work with complex models in their heads. I've worked on more than one software project where there were members of the team that quite obviously never grokked the dynamic characteristics of a system even after many months of intimate experience with it, whereas others grokked it quickly. It had nothing to do with education or experience or even desire to learn in many cases. There is a lot of anecdotal and some literature evidence for this even if you restrict yourself to the pool of pasty white male software geeks. It is also probably why software has the unique feature that half the really brilliant people working in it do not come from a traditional CS background; it was not education per se that made them great. The noted correlations with neurological structures is likely not coincidental either. Mostly though, I've never seen anyone learn to develop the ability to deal with very complex models that otherwise had all the necessary background. Individual ability to deal with abstract complexity has always been remarkably constant over the years in my experience. I have seen people that clearly had the ability from very early on develop the knowledge to put it to work in the software domain; this ability manifests long before it can be productively applied. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=69942428-66d159
Re: [agi] Where are the women?
On Nov 28, 2007, at 10:32 AM, Ed Porter wrote: I talked to her about the gap between women and men in science, and she claimed under her stewardship her junior high schools got a grant to promote the teaching of math to girls, and, in stark contrast to the previous condition, after several years the girls were outperforming the boys substantially on math aptitude tests. Up through junior high, it is *typical* for females to do better than males on math aptitude tests, so the apparent implication that this is not the case is misplaced. If the females were underperforming males in that age range then it probably indicates a problem with the curriculum. Male aptitude catches up and passes females starting around the beginning of high school, a pattern that holds across countries and cultures. I would also point out that at the younger ages you are primarily learning arithmetic, which has characteristics that are not representative of mathematics at large, which may have something to do with it. If you measured math aptitude in 6th grade and at the math department of a university, you would be measuring very different things. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=69621909-5c091d
Re: [agi] Where are the women?
On Nov 28, 2007, at 9:18 AM, Robin Gane-McCalla wrote: The interesting thing about CS and AI is that they are man-defined fields whereas physics, chemistry, biology etc are defined by nature. Only to the extent that mathematics is "man-defined", but then physics et al are built entirely on mathematics so I'm not sure where you are going with this. Computer science, and by extension AI, is not a field coalesced out of an arbitrary set of brain farts. Perhaps the simple fact that almost all programming languages and concepts in AI were designed by white males (and a geeky subculture of white males at that) is the main factor that has limited the entrance of women and other minorities rather than other cultural differences. The only substantive cultural bias in programming languages is the pervasive use of English language keywords, which hasn't seemed to slow down pasty white males who do not speak English a whit. There are only a handful of abstract concepts that underly all programming languages, and if you understand those abstract concepts then the construction details of the programming language are largely immaterial. How, precisely, would a female minority design a lambda calculus programming language that would be radically different from the myriad of such languages invented by pasty white male geeks? Programming languages are derived from mathematical models, with some application-oriented syntactic sugar to make common operations simpler. They are precise and highly regular constructs whose only "cultural bias" is that they disallow ambiguity as a basic feature that follows from their mathematical derivation. Being able to manipulate complex multi-dimensional graphs in your head and communicate without ambiguity are the only background skills required to be a good software geek; the latter is learnable, but I suspect the former is largely innate and even most white males are relatively poor at it. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=69512470-7e910a
Re: Re[6]: [agi] Danger of getting what we want [from AGI]
On Nov 27, 2007, at 7:21 PM, Matt Mahoney wrote: As a counterexample, evolution is already smarter than the human brain. It just takes more computing power. Evolution has figured out how to make humans out of simple chemicals. "figured out"? So if we implemented a planet kill, this "evolution" fellow would whip up another batch of humans posthaste? Lotto balls must be brilliant economists; they make multi-millionaires with impressive regularity. J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=69298173-46dfaf
Re: [agi] Funding AGI research
up with the ideas? And if so, why was it so hard to convince everyone else? No one is making the claim that there is no market for AGI today that I know of. If someone had an AGI as thoroughly designed and spec-ed as Babbage or Leibniz, they would have little problem selling it, but the reality is that we do not have an AGI market full of Babbage and Leibniz, we have an AGI market for wannabes that aspire to being Babbage or Leibniz. That is a distinction with a difference, and the cases are not analogous. Babbage and Leibniz competently designed things for which their was no market. A market exists for AGI, there simply have been no Babbage's around to meet that market. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=66475721-fa9393
Re: [agi] Funding AGI research
ocess works. You don't have to look viable to yourself, you have to look viable to everyone else. Proving your viability is *your* job, and it is no one else's fault if you are incapable of doing it. I empathize, but I figured out that you can accomplish far more by playing the game according to the actual rules than by inventing your own rules and expecting everyone to accept them. If you are smart enough to create AGI, you are smart enough to game the rules of the real world to your advantage without too much effort; don't fight it, use it. It is a waste of effort discussing how things "ought" to be. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=66464336-304149
Re: [agi] Funding AGI research
On Nov 18, 2007, at 10:45 AM, Benjamin Goertzel wrote: AGI is always going to be viewed as a major technology risk, unless one comes into the fundraising process with an extremely strong prototype (and maybe even then). With a strong prototype, you can get enough of the right people on- board such that the perception of technology risk can be greatly mitigated. It is theater, but it does make a useful measure by proxy of the technology for investors who cannot make a really thorough evaluation of the technology themselves. Mitigaging the people-risk requires getting experienced businesspeople on board, which is generically difficult for an AGI company because of the bad reputation AI has. Yes, and a lot of investors use this as a filter for a technology venture. If you cannot find a competent business person you can sell the technology venture to, it is interpreted to mean that there is probably no practical business there. AI is bad in this way, both because it has a well-deserved poor reputation *and* it is so difficult to evaluate from the standpoint of someone who is not deeply technical. Mitigating the market risk means finding a market niche where incremental work toward AGI is of dramatically more economic value than narrow-AI technology. I think this is really the hard part. Yes, and this remains true even if you have concrete, demonstrable proof of solving the general case. If they perceive an incremental path that can generate revenue, that's the path they want you to take even if you could ultimately make more money faster by jumping straight to the end point. It is the way these things work. The particular risk they are mitigating here is that of poor execution, which is significant no matter how killer the technology. The less execution required, the lower the odds you'll do it poorly. As I've said before, I am bullish on virtual worlds and gaming as an area where early-stage AGI tech can have dramatically more economic value than cleverly crafted narrow-AI. Humanoid robotics is clearly another such area, but a trickier area to get started in right now. But I'm not saying these are the only examples. Virtual worlds are an interesting sector because they touch a lot of areas where current computer science does not have a good off-the- shelf solution. It is an environment that makes many inadequacies obvious that software designers have been very good at masking. I have some significant involvement in the virtual world space myself; while it does not interest me per se, there are a number of interesting business opportunities surrounding it. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=66371996-f0ab47
Re: [agi] Funding AGI research
On Nov 18, 2007, at 10:41 AM, Richard Loosemore wrote: An investor will want to know what creative ideas you have that *directly* start to solve that problem. These are available! Both Ben and I have detailed plans. Neither of us say "just trust me". Wait, what? The "that problem" in this case is not AI, from a venture finance standpoint. Understand that you are essentially selling "a non-demonstrable idea on how to do research that may ultimately allow us to solve a problem", which is not the same thing as "solving the problem". You are looking for research money, not venture money. One is allowed to have some amount of uncertainty in the business side of a venture because such things are always a bit non- deterministic. You can come up with an exceptionally detailed business plan for how you are going to become the next Sausage King of Chicago, but you never really know how the market game will unfold in practice. Technology, on the other hand, can be very strictly evaluated in considerable detail such that there is little or no risk that it will turn out to be infeasible; it may not be economical or practical, but it will technically work. Furthermore, an acceptably detailed description such that you are not tacitly stating "just trust me" is indistinguishable from a prototype for most purposes. Unless the documentation demonstrates conclusively why the technology *must* work as intended, it is a "just trust me" proposition. And in some cases you can find investors who will find this to be an acceptable proposition if you have the rest of your game in order. It sounds to me like you are either making a stronger assertion about your design and documentation than I see Ben usually make, or you are implicitly saying "just trust me" to investors and do not realize it. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=66353418-b25d1e
Re: [agi] Funding AGI research
On Nov 18, 2007, at 3:40 AM, Bob Mottram wrote: I've heard people on AI forums make this claim many times over the last 15 years - something like "I have discovered the secret of AI !... but I'm not going to tell you what it is unless you give me a lot of money". I think the thing which makes the difference between regular charlatanry and an investable project is whether or not you can show something which might indicate that it's really feasible - even if that something is less than a fully working prototype. The charlatan of course will always flatly refuse to reveal the smallest detail. Yet crazy "blue sky" ideas get funded semi-regularly by knowledgeable investors. Some pan out, most don't. One does not necessarily have to prove a prototype to fund a "blue sky" venture. Indeed, if you have a prototype it is no longer "blue sky". There is at least one other type of asset that can be a sufficient condition to get conventional venture funding, though it may take nearly as much work: reputation. Individuals with a credible reputation for being capable of feats of technological or business wizardry can often raise money entirely on spec because their credibility and reputation makes a result plausible. You still need a thorough business plan, but you don't have to prove the theoretical nature of the product as your mere involvement mostly covers that bit of due diligence since you are presumably more capable of that evaluation than anyone else in the room *and* your competence at that evaluation is trusted based on past performance. Note that having a credible reputation in "AI research" is usually not sufficient on its own since that whole field has the patina of low credibility, you have to have done something concrete in a more real field; Jeff Hawkins, for better or worse, is an example of someone involved in AI who can carry himself on reputation regardless of proven technical competence in that field. But again, useful reputation in this regard is rarely inexpensive. There are multiple paths to AGI venture funding, and individual situations will vary. This is not just a problem for AI research, you often have to bring reputation and/or a thorough description in other venture areas as well. The reality is that (virtually) no AI research meets the basic level of description and/or credibility that is routinely required in other technology ventures. Any decent AI venture will be able to meet these due diligence thresholds; the howls of protest to the contrary are indistinguishable from those of crackpots and incompetents in every other venture field. AI researchers don't get singled out for being AI research per se, they simply don't rise to the basic level of due diligence required in the venture funding world, even for "blue sky" ventures. I would make the observation that this is eminently fixable if an AI venture is worth a damn, and some people do raise money when they approach it in a proper venture-funding context. A2I2 would be an example of an AI venture that has been relatively successful in this regard *because* Peter Voss understands the mechanics of venture funding as a practical matter; there is a lot more to it than thinking you have a super-duper AI idea, and competence in execution matters at least as much thinking you have an idea. Proactively minimizing risk in as many areas as possible make a venture much more salable, but most AI ventures tend to be very apparently risky at many levels that have no relation to the AI research per se and the inability of these ventures to minimize all that unnecessary risk is a giant mark against them. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=66345307-107ff1
Re: Bogus Neuroscience [WAS Re: [agi] Human memory and number of synapses]
On Oct 21, 2007, at 6:37 PM, Richard Loosemore wrote: It took me at least five years of struggle to get to the point where I could start to have the confidence to call a spade a spade It still looks like a shovel to me. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=56165909-1c84fb
Re: [agi] More public awarenesss that AGI is coming fast
On Oct 18, 2007, at 11:32 PM, John G. Rose wrote: It's really hard to sell if the deliverable time frame exceeds 3 to 4 years. Why does an AGI deliverable require more than 3-4 years? You better have a good answer for that, or no one will fund you. Most people *don't* have a good answer for that. Ya personally I don't like the VC model it's better to deliver concrete results even if they are in baby steps. BUT if the guy next to you is getting massive investment THEN what route do you take? Definitely boils down to who's got the right software IMO. There is enough VC money for everyone with a decent business model. Honestly, most AGI is not a decent business model. Otherwise Mentifex would be smothered in cash. It might even keep him quiet. Good AGI with adequate reputation would have no problem getting funded. Indeed, mediocre AGI with adequate reputation routinely gets funded. The problem is an absence of both reputation and business credibility. There are a lot of people with crap reputation that can still turn a dollar. The reality remains that virtually all AGI projects are objectively crap investments, and consequently it is an uphill battle to find people willing to make such investments. A track record is huge; proving that you have delivered on insanity in the past will have VCs lining up to invest in your insanity in the future. But you have to have delivered at some point. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=55274752-0062d6
Re: [agi] More public awarenesss that AGI is coming fast
On Oct 18, 2007, at 11:00 PM, Benjamin Goertzel wrote: I think that AGI for agent control in virtual worlds is not so hopeless in terms of appealing to VC's ... there's a real market there, and there's clearly a situation where more and more powerful AGI can yield more and more profits... The problem is that VCs want nominally provable profits. AGI does not fall under that classification unless it is so good that you do not need VCs. Mind you, I have no problems with VCs and find them easy to work with, but they are usually good for a certain type of business. Angels are better for spec work, but less reliable and the dollars are smaller. I have talked to a number of VC's in recent months -- and by and large they want to pigeonhole us as a company that forever will be focused on whatever our first product is gonna be (If your first product is for instance an animal in virtual worlds then -- bingo! -- you're a virtual animal company!!) Heh. Very true. One of the things that has reduced my annoyance over the years is finding VCs that have sufficient vision that I can work with them. There are not many of them, but they are not entirely rare either. The problem is largely if they think the venture is common. You have to be something pretty special for a mainstream VC to invest in AGI, which means you have pulled a rabbit out of the hat at least once. And "special" means having a conventional track record. It has been well worth my investment in ruthless conventional business to develop the contacts required to fund more exotic ventures no questions asked. It was a long way around, but not much has changed in the interim. If people make enough money, they'll invest in crazy ventures. VC's in nearly all cases don't have a long time horizon, so to find an AGI opportunity that synergizes with their needs requires a good bit of luck... If the AGI is any good, wouldn't the time horizon be short...? Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=55274070-173c5f
Re: [agi] More public awarenesss that AGI is coming fast
On Oct 18, 2007, at 10:40 PM, John G. Rose wrote: Well after living in Seattle during the dot com craze the hype was just absolutely out of control. Yet people did get funded. Was it all worth it? Hell yeah but the hangover was pretty bad :) AGI IS hypeable but people have to make a conscious decision on whether to do so or not as without any deliverables it's going to look real bad when the investors pull out. AGI is poorly suited for venture capital in every case I can think of. Ignoring everything else, it tends to leave the venture constantly begging for capital which has serious consequences on performance and reputation. It is a Catch-22, though perhaps well- deserved. In short, traditional venture capital is a poor finance model for AGI. Which does not suggest other finance models. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=55269651-199505
Re: Economic libertarianism [was Re: The first-to-market effect [WAS Re: [agi] Religion-free technical content]
On Oct 10, 2007, at 2:26 AM, Robert Wensman wrote: Yes, of course, the Really Big Fish that is democracy. No, you got this quite wrong. The Really Big Fish is institution responsible for governance (usually the "government"); "democracy" is merely a fuzzy category of rule set used in governance. I am starting to get quite puzzled by all Americans (I don't know if you are American though, but I want to express this anyway) who express severe distrust in government. Because if you distrust all forms of government, what you really distrust is democracy itself. This bias is for good reason; there are well described pathological minima that are essentially unavoidable in a democracy. The American government was explicitly designed as a constitutional republic (not a democracy) to avoid these pathologies. In the 20th century the American constitution was changed to make it more like a democracy, and the expected pathologies have materialized. If you do not understand this, then the rest of your reasoning is likely misplaced. Much of American libertarian political thought is based on a desire to go back to a strict constitutional republic rather than the current quasi-democracy, in large part to fix the very real problems that quasi-democracy created. Many of the "bad" things the Federal government is currently accused of were enabled by democracy and would have been impractical or illegal under a strict constitutional republic. Here you basically compare democracy to... whom? The devil!? Perhaps I should refrain from using literate metaphors in the future, since you apparently did not understand it. My recommendation is to put some faith in the will of the people! When you walk on the street and look around you, those are your fellow citizen you should feel at least some kind of trust in. They are not out to get you! I'm sure they are all lovely people for the most part, but their poorly reasoned good intentions will destroy us all. The problem is not that people are evil, the problem is that humans at large are hopelessly ignorant, short-sighted, and irrational even when trying to do good and without regard for clearly derivable consequences. Actually, I believe that the relative stupidity of the population could act as a kind of protection against manipulation. Non sequitur. Also, the history shows that intelligence is no guarantee for power. The Russian revolution and the genocide in Cambodia illustrates effectively how intelligent people were slaughtered by apparently less intelligent people, and later how they were controlled to the extreme for decades. You are improperly conflating intelligence and rationality. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=51970341-6a9d1c
Re: Economic libertarianism [was Re: The first-to-market effect [WAS Re: [agi] Religion-free technical content]
On Oct 9, 2007, at 4:27 AM, Robert Wensman wrote: This is of course just an illustration and by no means a proof that the same thing would occur in a laissez-faire/libertarianism economy. Libertarians commonly put blame for monopolies on government involvement, and I guess some would object that I unfairly compares fish that eat each other with a non-violent economy. But lets just say I do not share their relaxed attitude towards the potential threat of monopoly, and a bigger fish eating a smaller fish do have some similarity to a bigger company acquiring a smaller one. The only solution to this problem I ever see suggested is to intentionally create a Really Big Fish called the government that can effortlessly eat every fish in the pond but promises not to -- to prevent the creation of Really Big Fish. That is quite the Faustian bargain to protect yourself from the lesser demons. Generally though, the point that you fail to see is that an AGI can just as easily subvert *any* power structure, whether the environment is a libertarian free market or an autocratic communist state. The problem has nothing to do with the governance of the economy but the fact that the AGI is the single most intelligent actor in the economy however you may arrange it. You can rearrange and change the rules as you wish, but any economy where transactions are something other than completely random is an economy that can be completely dominated by AGI in short order. The game is exactly the same either way, and more rigid economies have much simpler patterns that make them easier to manipulate. Regulating economies to prevent super-intelligent actors from doing bad things is rearranging the deck chairs on the Titanic. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=51651108-c1aa2b
Re: [agi] Open AGI Consortium
On Jun 5, 2007, at 10:01 AM, Mark Waser wrote: There is nothing necessary to hold up in court. The trustees/"trustworthy owners" are taking the action. The fact that their decision was based upon the ramblings of an AGI is entirely irrelevant as far as the legal system is concerned. There is, of course, the danger of trustee defection but I don't believe that you can legally stop that short of declaring the AGI a person and making the trustees unnecessary (and I'm not holding my breath). The entire point of the trustees is to provide the correct legal cover for the AGI. That sounds like a contributor lawsuit waiting to happen outside of the contributors contractually agreeing to have zero rights, and who would want to sign such a contract? Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Open AGI Consortium
On Jun 5, 2007, at 6:32 AM, Mark Waser wrote: This is the kind of "control freak" tendency that makes many startup ventures untenable; if you cannot give up some control (and I will grant such tendencies are not natural), you might not be the best person to be running such a startup venture. Yup, my suggestion of giving control to five or six "trustworthy owners" is definitely the epitome of "control freak".:-) Why all the emotion? No emotion, just a practical observation. Lacking real capital of some type in the sense most investors would recognize, you do not usually have the luxury of control. Or more accurately, you can have control to the exclusion of others participating. Lack of capital (of all types) is deadly to a startup venture, and that very much appears to be the case here. Blue sky ventures and "maintaining control" are pretty much in opposition to each other if you do not want to marginalize your funding opportunities. The lack of intrinsic capital is going to make things tough, because the only real currency you have *is* control. No, the real currency that I want to have is an awesome talent pool and some good demonstrable progress before we look for additional funding. Declaring your talent pool as "awesome" does not make it meaningfully awesome to the rest of the world ipso facto. Same goes for "demonstrable progress" short of a Killer Demo(tm). To the rest of investor universe, this looks a lot like yet another attempt at a poorly organized and value poor AI venture. I would make the additional observation that you do not bootstrap a seed with five or six peers, more like two to three and maybe four. Far too much dilution of focus and vision otherwise which will bleed energy. What distinguishes this venture from the hundreds of other ones that are frankly indistinguishable from yours? It is a bit like true religion. Everyone says their wonky AI venture is the One True Venture, but if you do not have religion -- and investors generally do not -- they all look pretty much the same. The paucity of credibility that afflicts those hundreds of other AI ventures appears to afflict yours just as much. What is that killer thing that you can convincingly demonstrate you have that no one else can? Without that, your chances are poor on many different levels. I'm trying to find your unique angle here, but have come up empty so far. Yes, that is going to reduce my funding opportunities -- but it's a requirement that I'm not willing to concede and I will black-ball any "trustworthy owner" candidates who show *any* signs of being willing to concede it. I'm not trying to stop you, I'm merely pointing out that it will very significantly reduce your opportunities and probably far more than you are anticipating. Either way, it won't be *my* problem. :-) I'm just trying to give you some practical perspective on the venture thing, both generally and as it pertains to AI. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Open AGI Consortium
On Jun 4, 2007, at 8:07 AM, Mark Waser wrote: (Depending on your specific type of interest in a company, an argument can be made that warrants can be more valuable than equity.) Warrants have the same control problems as options do -- magnified by the fact that they are transferable. They are definitely not what I would call acceptable for this purpose. Eh? What is the problem with them being transferable? Of what value are these instruments to anyone if they are not ultimately transferable? This is the kind of "control freak" tendency that makes many startup ventures untenable; if you cannot give up some control (and I will grant such tendencies are not natural), you might not be the best person to be running such a startup venture. If I was a VC looking at your company -- not a foreign role for me -- the fixation on that aspect would raise red flags. Blue sky ventures and "maintaining control" are pretty much in opposition to each other if you do not want to marginalize your funding opportunities. The lack of intrinsic capital is going to make things tough, because the only real currency you have *is* control. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Open AGI Consortium
On Jun 4, 2007, at 4:35 AM, Mark Waser wrote: This kinds of things are pretty strictly regulated now, and waiting until the end to contract a stake to your contributors would be a disaster for them in terms of both their return and/or tax liability, If you're waiting until the end to distribute shares/equity, the immediate tax liability is nasty because it is counted as a sudden transfer of value. The return, however, if the shares/equity were sold immediately is exactly the same as if they owned it all along. If, however, ongoing profits are simply distributed (instead of equity), there is no problematical sudden transfer of value. And realistically, there aren't going to be profits pre-AGI. Depending on how the nominal value is disbursed, the true financial value can vary significantly. Other than outright equity, Instruments like profit distribution are about the worst in this regard, instruments like warrants are among the best (but you can't give those to just anyone), and most other instruments fall somewhere in the middle. The difference is significant: the real return between the best and worst can easily be 2x. (Depending on your specific type of interest in a company, an argument can be made that warrants can be more valuable than equity.) The closest *decent* way to do what you want to do is to contract options upfront with modifying conditions and qualifications based on future performance. Do you believe that you could successfully do that? Would you be willing to write up an initial shot at it? Since many startups in Silicon Valley do exactly this, I would say that it is quite doable. It is less flexible and accurate than waiting until the end to make determinations of value, but it is a fair proxy and both parties have to agree to it anyway. If structured well, bits can frequently be negotiated off-contract later if conditions change. It is how startups deal with things like high rates of churn. Personally, I do find the current state of regulation to be irritatingly inflexible. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Open AGI Consortium
On Jun 3, 2007, at 5:52 PM, Mark Waser wrote: >> So, the share allocation is left undetermined, to be determined by the AGI someday? That's what I'm saying currently. The reality is that my project actually has a clear intermediate product that would cleanly allow all current contributors to determine an intermediate distribution -- but I'm really not ready to discuss (or more importantly defend) that yet so it's better to just take it as -- Yes, it will be the AGI. You may be assuming flexibility in the securities and tax regulations than actually exists now. They've tightened things up quite a bit over the last ten years. Equity and pseudo-equity (like incentive stock options -- ISOs) should be contracted at the earliest possible time, and before either financial or delivery milestones if at all possible, if you care about the value you will actually be delivering to your contributors. Furthermore, you cannot grant equity instruments to just anyone, and pseudo-equity instruments like ISOs have a ton of rules that limit their ability to return fair value to your contributors. And then there is the what-if of dissolution, acquisition, etc in which a pre-AGI determination of equity ownership needs to be figured out -- the way you've set it up, the contributors would be entitled to squat. This kinds of things are pretty strictly regulated now, and waiting until the end to contract a stake to your contributors would be a disaster for them in terms of both their return and/or tax liability, never mind the unpleasant scenarios that can occur. I cannot imagine that a savvy person would accept deferred contracting of options and equity. It would be one of the worst possible equity stake schemes I have seen. The closest *decent* way to do what you want to do is to contract options upfront with modifying conditions and qualifications based on future performance. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Open AGI Consortium
On Jun 3, 2007, at 6:20 PM, Benjamin Goertzel wrote: So you are going to make a special set of corporate bylaws that disentangle shares from control? Hmmm... Something like: the initial "trustworthy owners" are given temporary trusteeship over the shares, but are then bound to distribute them according to the wishes of the AGI once the AGI passes some threshold level of intelligence?? Disentangling shares from control in a way that is actually bulletproof and/or legally viable is difficult and relatively expensive. The laws and regulations are generally written specifically to make that a pain for anything resembling a for-profit entity. It requires a high degree of trust between multiple parties to make it fly without having an unambiguous controlling financial interest. One State in the United States (all corporate law is state law for most purposes) explicitly allows the creation of non-economic interests in limited liability constructs: Nevada. As far as I know it is unique to that State, but it allows one to completely separate control from equity. This only applies to LLCs rather than Corporations out of practical necessity, I believe due to securities regulations, but it allows 100% of the control to be granted to a party that has no financial interest in the organization and which has no obligations and receives no profits. It is obvious this class of entity was designed to allow the creation of a controlling interest that lacks de facto exposure because the mechanism of control has no intrinsic financial value, unlike control that is tied to equity of some type. Incidentally, control of equity must ultimately resolve to a Natural Person. Your AGI will have no legal ownership of anything. But I guess you can worry about that later... Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Open AGI Consortium
On Jun 3, 2007, at 3:13 PM, YKY (Yan King Yin) wrote: The problem is that I still want to get rich, and to make XYZ a non- profit would be dishonest and may result in some awkward contradictions later. (Unless my personality changes... which is also possible). To put it really simply, your venture is no different than dozens, nay, hundreds of other ones. This is a very well vetted area and just about every possible organizational possibility has been tried numerous times in many contexts with varying levels of success. Rather than grasping for a new way to do things that you find aesthetically pleasing, you would probably be better off specifying what the necessary endpoint is and then pick one of the many extant proven structures for achieving those endpoints if the people involved are up to the task and study why those structures worked and others failed. What you have proposed is a blue sky startup with the negatives compounded by a lack of legitimate experience at pulling such things off. In many ways, you have a naive perspective of the significant constraints on implementation this creates. The risk profile is extremely high, which means that your venture is worth approximately nothing to anyone, which the all the economic consequences implied. Which in short means that you would retain almost no leverage over the project even if you did manage to organize it. You sorely lack capital, whether intellectual, reputation, or cold hard cash -- the stuff ventures are built on. And capital begets capital, so there is a virtuous cycle. That does not mean your project is impossible, but it is implausible. You need to spend more time working on accumulating the necessary capital. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Open AGI Consortium
On Jun 2, 2007, at 10:37 AM, Mark Waser wrote: If the corporation does have an influx of cash (due to an intermediate success), a consensus of active contributors would have to decide how much to share out and how much to retain as seed money (and I would push real hard for the majority, if not all, of it to be retained as seed money -- unless it were the result of a single or small number of contributors who needed to be rewarded with a substantial chunk). If the corporation has an influx of cash due to an investor or benefactor, it would all be kept as seed money to hire individuals (whose contributions would be recognized at a reduced rate due to their paid status). It is worth pointing out that compensation, equity issues, and oversight are highly regulated. About half of the organizational and compensation ideas I've seen proposed would require an army of lawyers to arrange, have serious consequences that have apparently been overlooked, or would simply be illegal under current law. There a complex tax issues that have to be understood as well. Things used to be more flexible, but they've been tightening the screws on creative organization for years in order to "do something" about perceived business malfeasance. The result is that there are complex rules and hoops you have to jump through that get worse every year, some highly restrictive, if you want to legally organize and operate a venture. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 11:35 PM, Samantha Atkins wrote: There are experienced engineers and then there are experienced engineers A few are 10x to 100x more productive than the average experienced engineer. Since in the real world time to implementation is the difference between success and failure it is not exactly true that experienced software engineers are mere commodities of no great importance. I suspect you know that. I will readily concede that. This is a different aspect of it. The variance in the efficacy of software engineers is stunningly high. Unfortunately, for most projects I cannot assume I have a super- engineer on board, though I would love to have one (or a dozen). Most algorithm design work these days is done with the abstract system design context in mind out of necessity. It is often that context which breaks conventional algorithms, so there is less "systems engineering" to it when finished than you might expect. Yes and no. Some of those abstract system models are quite difficult to implement in reality with sufficient scalability, dependability and other desirable motherhoods. Well, that was kind of my point. If the desired characteristics are explicitly specified, the solutions will be much tighter. In practice, I get these tight specifications regarding scalability and failure modes. So all those software project cost overruns come from what exactly? Heh. In practice, they come from project managers and/or clients unwilling to accept realistic estimates. That is one of the absurd things about software projects; they would rather you lie out your ass upfront and bleed them on the back-end than put too large an estimate at the beginning. I'm not saying it makes sense, but it is definitely a common pattern. Small estimate + huge follow-on is easier to sell than a huge estimate + small follow-on. Human nature it seems. Can you get well-bounded costs on entire systems? Not really. How come? Eh? Worst case scenario, assuming some semblance of competence, is a small integer factor discrepancy and usually less. The really hairy problems are those that have been spec-ed with no obvious solution. That's when most consulting firms pray that the brains in their outfit will deliver them. Which was me, at one time. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 2:13 PM, Russell Wallace wrote: Given a precise specification, the cost of converting it into code is reasonably predictable, yes. The more difficult and unpredictable part is coming up with the spec in the first place. I'm not talking about writing version 3 of your in-house payroll program using the same tools you used for versions 1 and 2; sure, that doesn't take a big leap into the unknown - it doesn't require new algorithms either. I'm talking about creating products that didn't exist before. Designing and building new systems with novel ideas from scratch, systems the likes of which have not been built before, is not new to me. I've done it for Global 100 companies, and numerous smaller ones. I am perfectly familiar with the problem space on the systems engineering side. In some of the more interesting cases, it was taking risks with several hundred million dollars in transactions based on exotic analytics, in spaces and using techniques that had never been attempted. I have, on occasion, dabbled in mundane systems but most have involved odd problem spaces. Fundamentally new apps are not something I have not experienced before. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 10:10 PM, Samantha Atkins wrote: Well, in my graduate database implementation class we had to design parts of a relational database from scratch. My design for handling the B+-tree concurrency was almost exactly like the Lehman- Yao algorithm. So it isn't all that obscure. I see there algorithm was published in 1981. My class was in the fall of 1980. Yet another place where not knowing how the academic game is played was a bit of a handicap. I just figured stuff out as I needed it for the pure joy of it. Yeah, but you are brighter than most, and genuinely interested in the problems. I routinely run into people nominally with CompSci degrees who are dumb as a bag of hammers when it comes to theory or problem solving. Lehman-Yao is not complex, in fact it is quite simple and elegant, but it requires grokking the nature of the problem such that the solutions are obvious. It is a minority in CS that do all the heavy lifting of being able to solve real algorithm problems. It is not a light discipline, and most are not suited for it. Most engineers are not all that good at mathematical reasoning. They aren't so great at stepping back from the details and seeing the working abstractions and patterns behind the details in reasonably full generality and formally capturing and manipulating those patterns. Without this ability they often produce sub- optimal, brittle results that cannot be easily adapted to somewhat different but quite related cases. While I am better at this than many working software engineers I learned a long time ago to respect the more mathematical, abstract and theoretical work in Computer Science. So I mine those papers for ideas, abstractions and approaches I missed and was often too head down to see. I would pay a lot for comp sci papers from a few years from now much less fifty years out! Someone after my own heart. :-) Regards, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 1:52 PM, Russell Wallace wrote: So you think the people who created products like Windows, Excel and Firefox shouldn't be writing software? That's the sort of thing I'm talking about, not little utilities or version 3 of your company's payroll program using the same tools you used for versions 1 and 2. This is the frigging AGI list we're having this conversation on, after all, not the "writing small in-house scripting stuff" list. The thing is, designing a product like Windows, Excel, and Firefox are to a significant extent pretty different than an AGI. It seems pretty obvious to me that an AGI will be a lot more like a systems engine design than a feature-centric user application. Windows would straddle this a bit since it does have an operating system kernel somewhere underneath that mess. I would expect the AGI userland to be pretty thin (at least at first) and fairly arbitrary. Lots of tightly specified magic that does not require that much code (relative to something like Firefox or Excel) underneath a very compact API, analogous to modern database engines (except that modern database engines have massive accreted userlands and relatively tiny kernels). Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 1:45 PM, Benjamin Goertzel wrote: Actually, it is quite possible to patent something purely protectively -- i.e. get the patent but then give everyone in the world the right to freely use the idea ;-) ... the point being to stop anyone else from fallaciously patenting it... It would be a lot cheaper just to publish it. Patents, done correctly, are pretty expensive. J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 1:12 PM, YKY (Yan King Yin) wrote: Also can you tell me what you're referring to by "well-documented social harm"? Patents are a market regulating mechanism, and therefore reduce the efficiency of the market in theory -- it increases the cost to the consumer for the patented technology. The other side of that equation is that patents nominally increase innovation which returns value to the consumer in other less obvious but important ways that are difficult to quantify. Whether patent-like regulations are harmful or helpful depends on whether the direct impact in market price are greater or lesser than the indirect impact of innovation rates and characteristics. It is a complex calculus and the direct and indirect impacts vary widely as a function of numerous environmental and demographic factors, which could lead one to conclude that they should actively managed in the same way the Federal Reserve manages the money supply. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 11:40 AM, Russell Wallace wrote: A week of effort will get you a piece of test code that runs in a harness to prove the algorithm works. In other words, it will get you nothing whatsoever that is of any use by itself. Creating software that does something useful typically takes much more than six months of effort, and I assure you, it is not work that a monkey could do. The prototype functions as a template that can be utilized in building the final product, and I would hardly call demonstrating something previously not possible in computer science "nothing whatsoever". It is what separates yet another boring business app from novel new app spaces. It requires nothing more than an experienced software engineer to get a production implementation. The point is that this part is pure commodity, actually solving algorithm problems is not. You cannot pay X dollars to Y computer scientists and get a result in Z months. For this reason virtually all of the economic value is in the algorithms and not the implementation. Most algorithm design work these days is done with the abstract system design context in mind out of necessity. It is often that context which breaks conventional algorithms, so there is less "systems engineering" to it when finished than you might expect. And approximately zero would be spent on its discovery as a percentage of total effort. It's all about systems engineering. Sorry, but the amount of time spent does not determine the value of that time. That idea passed its Sell By date in the late 19th century. Labor has no intrinsic value. In the case of a software application enabled by a new algorithm, virtually *all* of the value is in the algorithm. The value is not in the time spent solving the problem but knowing *how* to solve the problem, since the value of the labor would be zero without it. Hell, we have classic allegories in the high-tech business that are all about that very point. If we are allowed to dismiss those parts of reality that we wish to ignore by calling them "window dressing" and "irrelevant", then algorithm research is irrelevant window dressing, so let's forget about it. Nonsense. One is fungible, the other is not. That is distinction with a very important economic difference. Algorithm research has an unbounded and unpredictable cost, systems engineering costs are generally quite predictable. I can go to any competent software engineer and get a production implementation of an algorithm with well-bounded costs. If I need a new algorithm, many computer scientists will never deliver anything useful and it could take anywhere from a month to a decade to an eternity to actually deliver that new algorithm even if they are capable in theory. The comparative risk between algorithm R&D and implementation of an algorithm that already exists is separated by an astronomical gap, and "risk" plays a major role in economics. Predicting the cost and delivery date of something that requires inventing technology that has never existed and is known to be difficult at a minimum is a fool's game. If we must acknowledge reality, then the truth is that an algorithm by itself is completely useless to anyone. The useful product is, to use Fred Brooks' term, the programming systems product, and that is where almost all of the effort goes. Furthermore, for any given way of solving a problem, even the first product typically has serious limitations. The "systems product" is a commodity product, fungible, the presumed algorithm is not. Apples and oranges. Talking about "effort" is lovely and all, but that is only one facet of the economic calculus. If you are developing an application that requires solving algorithm problems to be feasible, virtually all of the risk is in trying to solve those algorithm problems and the capital invested toward that end. Systems engineering is merely overhead and carries little real risk sans new algorithm development. Full life-cycle engineering of large-ish software systems is not something unfamiliar to me, and one of the reasons I left it was *because* it was boringly predictable. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 10:25 AM, Russell Wallace wrote: I'm talking about the process of going from a stack of CS papers to a working, useful program. I'm pointing out that most of the difficulty lies in that process, not in generating the CS papers. Designing a useful new algorithm may take six months of research and development, but an implementation of that algorithm will take something on the order of a week of effort. There is nothing hard about implementation, a monkey could do it given adequate instruction. There is no shortcut to actually having those instructions. The non-spatial database problem is sufficiently nontrivial in implementation that entire corporate lifespans, man-centuries of effort, get spent on implementing database systems (_not_, by and large, on figuring out the algorithms and data structures). Why do you think the spatial database problem will be so much easier? You are conflating unrelated things, and making a rather significant number of assumptions about how an algorithm might be used. Approximately *zero* implementation effort would be spent on the implementation of a spatial algorithm as a percentage of total effort. A new algorithm is nothing more than an enabling technology with inconsequential implementation time in most cases. All that other window dressing is application specific and quite irrelevant. My window dressing has no bearing on anyone else's window dressing, and is a completely separate calculus. Having the algorithm defines what is possible and what is not, having an implementation of that algorithm is a minor expense and often not even reusable in any case. The economics of many applications pivot entirely on whether or not that algorithm exists, because if it does the expense of implementation becomes entirely inconsequential and if it does not no amount of implementation investment will overcome the lack of it. You seem to be denying this reality. If this algorithm was patented, the market would be defined by the ability to do the R&D, if it was not the market would be defined by the ability to pay the very modest implementation expense and organizations that had large amounts of pre-existing window dressing at their disposal would have a significant advantage because they've already finished most of their investment in turning out a useful product. For example, dropping new index types into existing transactional database engines usually takes effort measured in man-months. The vast majority of all people that might want to use an algorithm (e.g. scalable spatial indexing) will never have any use at all for your implementation but will have a huge amount of use for the underlying algorithm that makes implementation possible at all. You put way too much value on algorithm implementation. The basic Google software implementation could be built from scratch today for less than a million dollars -- chump change -- and is therefore obviously not the gating factor in building another Google. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On Jun 1, 2007, at 7:48 AM, Russell Wallace wrote: Rewind a little, to a very basic data-processing problem: sorting a simple array. Imagine a programmer who knows no computer science trying to tackle this: likely he'll come up with selection or insertion or perhaps bubble sort, which will be painfully slow on large arrays. Show him a paper with the merge sort algorithm; how much of the job have you done for him? Most of it. Anyone competent in a general-purpose programming language can write a fully adequate implementation of merge sort quite easily. Sure. Some algorithms are considerably less obvious than others. Forward some decades to the problem of writing a conventional relational database indexed only by scalar data. Show the programmers (notice the plural - we're now at the stage where teams are typically involved) a stack of computer science papers, any papers you like. How much of the job have you done for them? Very little. B-trees are all very fine, but "now I know B-trees" doesn't actually help all that much. The hard part is in the implementation details, in software engineering not computer science. You call them "implementation details", but the reason we do not actually use vanilla B-Trees is because they have a few pathological characteristics e.g. poor concurrency. After B-Trees were invented it took another ten years before someone figured out the trick to make them support high-concurrency (Lehman & Yao, 1981). The trick is obvious and simple in retrospect, but it nonetheless took a decade for anyone to figure it out. On the other hand, it took no time at all to go from B-Trees to B+Trees. In computer science today, high-concurrency B+Tree implementations are among the more ubiquitous constructs, probably far more ubiquitous than a vanilla B-Tree. These contain two significant improvements over B-Trees: the B+Tree data structure and the Lehman- Yao concurrency algorithm. Based on the evidence at hand and the nature of the problems, I think an argument could be made that given the B-Tree as a starting point the B+Tree was obvious and might be considered an "implementation detail" but the Lehman-Yao concurrency algorithm was not. I put it to you that a spatial database is like this except even more so. I predict that even if you could photocopy a stack of computer science papers from the year 2057 and put it on the desks of a team attempting to write a spatial database, you would have done only a small fraction of the job for them - most of the effort remains in the engineering. This argument is neither here nor there. Do you need CS papers from 2057 today because the problem is not an "implementation detail" today? You are still using "implementation detail" in a vague and poorly defined way. If you assume the algorithm problem has been solved, then of course everything is reduced to implementation detail. The spatial database problem is trivial in implementation. The problem is that we need to select a space decomposition algorithm that works well buried in the innards of an otherwise unremarkable design problem. You may call the design and selection of that algorithm an "implementation detail" but it is the fundamental limitation and there is no off-the-shelf algorithms that we can plug- in there and get good results, despite considerable effort to find such an algorithm. Implementation is irrelevant, it is a design space limited entirely by a lack of specific algorithm know-how. No amount of implementation twiddling will generate the required characteristics. And if we *did* have such an algorithm, the spatial database engine would design itself since most other aspects are thoroughly solved algorithm problems. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
within normal load parameters for a conventional database). Ignore the other major problem (distributability) for now. Spatial data structures/algorithms that scale in size and concurrency are pretty fundamental things. It is analogous to not having B-Trees for dimensionless data. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
Re: [agi] Opensource Business Model
On May 31, 2007, at 5:21 PM, Samantha Atkins wrote: Actually patents are commonly filed to be as broad as possible. So a very specific way of doing X will be filed as a patent on X. Also some things are so obvious that they are very likely to be invented over and over again. The 1-Click patent held by Amazon is a good case in point. Why should everyone have to license or not use something so obvious? These are frivolous patents, and there are as many for hardware as for software. This is a separate problem from software algorithm patents in theory. Incompetent management on many levels does not equate to the underlying idea being poor in principle. There is also the small matter of prior art. I did a LOT of work in distributed objects and object persistence in the mid 80s. But at the time software patents were just not done, at least not by my company. About eight years ago I looked up patents in this area to see that Sun and IBM had a number in these areas that my work in the 80s certainly was relevant to and much earlier. But since the company and I did not keep sufficient records and since I cannot afford to challenge them myself the current practice would restrict me in some cases from using what I myself invented long ago. That is not healthy. Sure, this happens and it has happened to me. The software industry does not have mature mechanisms for dealing with patent issues (other industries handle the mess better), aggravated by the poor job the patent office does. This is a completely separate issue from software patents in theory. If we made it a practice to dissolve things merely on the basis that they are currently being run incompetently then you can start with the government at large. I agree that there is a lot of incompetence in the management of patents (it extends much further than software patents) but I find it interesting that a lot of people want to fix it by eliminating patents rather than dealing with the underlying problem and without consideration for whether or not patents are a good idea. As I stated previously I am not averse to eliminating patents, but it would be short-sighted idiocy to only apply it to software algorithm patents as though they are special or to do so because dealing with the incompetence is too hard without regard for whether or not there is value in principle. I do not agree that all patentable things are equal. I believe that software algorithms are much more fine grained and inter- related and independently discoverable than say newly machine inventions. You believe that, but where is the evidence to support that assertion? How familiar are you with some of the numerous other heavily patented fields? Thousands of fine-grained and inter-related patents is not a feature unique to software algorithms, and people in those other fields will often whine about the same issues. This does not address the issue of non-frivolous algorithm patents. If an algorithm is developed that enables capabilities not previously described in computer science literature, how is anyone being hindered by it being patented? Obviously they were getting along just fine before it was invented never mind patented. Do you think public access to new algorithms would naturally follow from disallowing them from being patented? History suggests not. Fortunately, many new algorithms are not patented, so that is not even a real concern in some cases. Again, this gets back to my point about the arguments against non- frivolous software patents being deeply disingenuous. The claim is that people "lose" by not having unfettered access to something that did not previously exist and which someone expended non-trivial effort in developing. It is a bait-and-switch many times, where a frivolous patent is used as an excuse why everyone should be allowed to use algorithm patents that are non-frivolous. The whole public discussion of software patents is surrounded by anecdote, shoddy reasoning, and transparently greedy motivations, but no one seems interested in dealing with the individual fundamental questions that define the situation. The unwillingness and/or inability of anyone to establish the facts and definitions surrounding software patents in any kind of rigorous manner does not lend credibility to the arguments made, whether for or against. Cheers, J. Andrew Rogers - This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
