A human needs to design the host/guest architecture in emergent inference.
It all goes back to the notion of receding intelligence. Once people know how
a process works, it is no longer "intelligent".
Perhaps the same can be said of people in about twenty years when we have
thoroughly mapped the brain. People will no longer be intelligent when we know
how they work.
~PM.
------------------------------------------------------------------------------------------------------------------------------------------------Confidential
- This message is meant solely for the intended recipient. Please do not copy
or forward this message without the consent of the sender. If you have received
this message in error, please delete the message and notify the sender.
From: ser...@scicontrols.com
To: a...@listbox.com
Subject: RE: [agi] Building high-level features using large scale unsupervised
learning
Date: Sat, 30 Jun 2012 10:26:10 -0500
Piaget Modeler, Thanks for telling me. Just from what you said, I'd reply the
same I replied to Alan. This work seems to be traditional narrow AI, not AGI.
PIAGET MODELER> Breaking channel (modality) dependence just means creating
rules for intermodel associations. That also can be programmed. SERGIO> Who
makes the rules? A human analyst? You are using human intelligence to create
man-made rules. Programming them makes no difference. Ray Solomonoff has said:
"A heuristic programmer would try to discover how he himself would solve (some
problem) - then write a program to simulate himself." The only visible
intelligence is the analyst's, which she used to create the program. But the
program does not know how or what created it, and contains no intelligence.
Proof: how do you make changes to the program? You call the analyst, right? By
contrast, in emergent inference, there is a host/guest architecture. The host
has no program, and no human analyst is necessary, so it does not rely on human
intelligence at all. If the host runs on a computer, then there must be a
program of course, but it only manages I/O and minimizes the functional. The
program is always the same, never changed, no matter what the associations are,
or what the problem is. Just like our brains. The guest is the knowledge
acquired from the world. It makes no difference how it is acquired, by sensors
or from a teacher. The host processes the guest and creates the response
algorithm (perhaps the program we were talking about), without any supervision.
PIAGET MODELER> In the PAM-P2 system, there are Associator processes which
create associations. The only human necessary is the designer. SERGIO>
Creating associations ("binding the qualia") is a century-old problem. Many of
the greatest of the 20th century (Turing and Church among them) were involved
with it. It is still very active, and there must be other "associators" around
(just guessing). The problem is that you need intelligence to make
associations, not the other way around. Intelligence comes first, then the
associations. Proof: why do you need a human designer for your Associator?
Answer: because you need the designer's intelligence, and you don't have any
intelligence in the program. PIAGET MODELER> PAM-P2 is under construction.
Stay tuned.SERGIO> Thanks. Sergio From: Piaget Modeler
[mailto:piagetmode...@hotmail.com]
Sent: Friday, June 29, 2012 7:09 PM
To: AGI
Subject: RE: [agi] Building high-level features using large scale unsupervised
learning Sergio, In the PAM-P2 system, there are Associator processes which
create associations. The only human necessary is the designer. Breaking
channel (modality) dependence just means creating rules for intermodel
associations. That also can be programmed. PAM-P2 is under construction. Stay
tuned.
------------------------------------------------------------------------------------------------------------------------------------------------Confidential
- This message is meant solely for the intended recipient. Please do not copy
or forward this message without the consent of the sender. If you have received
this message in error, please delete the message and notify the sender. > From:
ser...@scicontrols.com
> To: a...@listbox.com
> Subject: RE: [agi] Building high-level features using large scale
> unsupervised learning
> Date: Fri, 29 Jun 2012 10:52:11 -0500
>
> Alan,
>
> You make the same traditional mistake that many make. You offer a
> substantial scientific contribution obtained by humans using their human
> intelligence, but never explain the process in their brains that made it
> possible for them to do all that. Jeff Hawkins (On Intelligence) does the
> same. His entire book revolves around certain "invariant representations"
> but he never explains where those come from. Your proposal is the usual
> mixture of man-made intelligence and promises. It does not contribute to
> artificial intelligence.
>
> The keywords are "association" and "emerge." You write:
>
> "The next challenge is that you need to break channel dependence and
> introduce associations between patterns ie with faces and the various
> representations of the word "face". I suspect that once channel dependence
> is fixed, then, at some high level in the network, these associations will
> emerge on their own. "
>
> "Introduce associations" means, in your mind, that you or some other human
> will be using their human intelligence, not artificial intelligence, to find
> those associations and introduce them in the program. "Emerge" indicates you
> realize that something is still missing and blame something else that you
> can't explain. You are not alone.
>
>
> Sergio
>
>
> -----Original Message-----
> From: Alan Grimes [mailto:agri...@speakeasy.net]
> Sent: Friday, June 29, 2012 8:20 AM
> To: AGI
> Subject: Re: [agi] Building high-level features using large scale
> unsupervised learning
>
> Ben Goertzel wrote:
> > How exactly do you suggest to bridge the functionality gap between
> > visual pattern recognition and all the other things human beings do?
>
> =)
>
> Setting aside problems noted as still being unsolved, here's a crude sketch
> of how the system can be organized. For the sake of brevity, only the
> cortical-thalamic-cortical system will be considered.
>
> The first thing to note is that this is an unsupervised pattern learner.
> That should be pretty amazing all by itself. The second thing to note is
> that all it deals with are vectors of numbers. There is no reason on earth
> that it can't be made to work with any conceivable stimulus that can be
> encoded as a vector of numbers. There are some serious channel dependence
> problems, previously noted, but the basic process is present.
>
> The third thing to note is that they could run their matrix stack in reverse
> and "imagine" what a face looks like. This is critical, especially for motor
> control! =P
>
> This is your basic algorithm. The next challenge is that you need to break
> channel dependence and introduce associations between patterns ie with faces
> and the various representations of the word "face". I suspect that once
> channel dependence is fixed, then, at some high level in the network, these
> associations will emerge on their own.
>
> The next issue is topology. You could organize the topology like the human
> brain and, in theory, it should be human equivalent. Motor control is
> implemented just like perception. It builds up complex sequences of actions
> from simple sequences of actions exactly as complex perceptions are built up
> from simple perceptions. To do something, you just run the stack in reverse,
> as mentioned above. Combined with channel dependence and free association,
> you obtain arbitrary sequences of planned actions.
> Actions that are fully learned become habitual (simply initiate the top
> level abstraction). Other actions require an iterative system-wide process
> for planning, but most of the mechanisms are already present.
>
> You obtain episodic memory by having a pipeline that associates concurrent
> perceptions, which appears to be what the hypocampus does.
>
> To obtain super-human intelligence, you need to make the topology of the
> system adaptive, or even accessible to the system itself. Ideally, you want
> a highly redundant, highly distributed, highly parallel and highly efficient
> architecture. This architecture does have a second class of scalability
> issues, each matrix, at each level of abstraction is of fixed size, There
> needs to be a process that simplifies and consolidates knowledge to a more
> ideal representation. At that point you're off the edge of the
> (metaphorical) napkin I sketched this all out on. =P
>
> About 80% of everything else you need is already available off the shelf,
> the other 20% might have some important, perhaps even difficult, challenges
> but then we're talking about emotions and motivation instead of
> intelligence.
>
> --
> E T F
> N H E
> D E D
>
> Powers are not rights.
>
>
>
>
>
> -------------------------------------------
> AGI
> Archives: https://www.listbox.com/member/archive/303/=now
> RSS Feed: https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57
> Modify Your Subscription:
> https://www.listbox.com/member/?&;
> d2
> Powered by Listbox: http://www.listbox.com
>
>
>
>
>
> -------------------------------------------
> AGI
> Archives: https://www.listbox.com/member/archive/303/=now
> RSS Feed: https://www.listbox.com/member/archive/rss/303/19999924-5cfde295
> Modify Your Subscription: https://www.listbox.com/member/?&;
> Powered by Listbox: http://www.listbox.comAGI | Archives | Modify Your
> Subscription
AGI | Archives
| Modify
Your Subscription
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com
