[ExI] Vint Cerf on AI, at ORNL

[spike at rainier66.com]

>… On Behalf Of William Flynn Wallace via extropy-chat

Subject: Re: [ExI] Vint Cerf on AI, at ORNL

 

>…The AI plays a program that is highly successful at chess.  The AI is told every time it moves in a way that is illegal and of course it notices the moves of the other player.  It is told when failure occurs: piece lost, checkmate…bill w

 

 

There is something here that is remarkable indeed, subtle in a way.

 

AI chess has been attempted by learning algorithms in the past, but the apparent breakthrough came in coding in the ability of the software to make inferences on some of the first principles rather than dictating those to the software based on human experience.

 

We have over 500 years of collective chess experience with the game in its modern form.  An opening book published in 1510 is still in use to this day.  We have guiding principles in chess such as center control, early development of pieces, pawn structure and so on which are programmed into the software to start with.

 

As I understand it, the current AI experiments do not code this information into the software.  They (somehow) let the software figure them out.  From my point of view, that qualifies as (limited) intelligence, for it is making inferences the way humans and other beasts do.  

 

I could dig into the source code of course, and see how it works, at which time I would come to understand how it works and it would no longer be intelligence, but rather just sophisticated code.  We have already defined computer algorithms which can be understood as not intelligence.  It is just code, run by an unconscious dumb machine.  But until I do that, it appears intelligent from my point of view.

 

As we get more sophisticated we run the risk of getting so smart, we figure out how our own brains work, at which time we become stupid by our own definition of smart.

 

Sigh.

 

spike

 

  

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20190905/6d5ff23d/attachment.htm>