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
