[computer-go] Re: What is thought

[Eric Baum <ebaum@xxxxxxxxxxx>]

Chrilly,

With all due respect, you ought to look at *What is Thought?* before
flaming it. 

Chrilly> There is a constant in thought-explanation since the
Chrilly> Greek-time. Thought/the human mind is always explained in the
Chrilly> latest technology...  Now the explanation
Chrilly> is DNA, because this is the latest new technology.


I set out to explain how the mind is equivalent to execution of a
computer program, which was more or less proved by Turing.
But Turing's result is mysterious, because how can execution of a computer
program have meaning? How can it be conscious? How can it understand?
*What is Thought?* gives a theory explaining this through an
extrapolation of 20 years of computer science research into Occam's
razor. Basically the point is that if a compact enough program is 
evolved that behaves powerfully enough, this implies the world has
understandable structure and the program understands it. 

Now I did not expect this to be in the DNA, that was forced by the
data. I give arguments from complexity theory, developmental biology, 
evolutionary programming, ethology, psychophysics, simple inspection,
linguistics, and other areas why this is so. The model gives a lot
of insight, agrees with vast amounts of data, and makes falsifiable 
predictions. 

Chrilly> In his book "Behind Deep Blue" Feng Hsu´s opinion of
Chrilly> AI is quite clear: "Bullshit".

*What is Thought?* discusses computer science approaches (like Deep
Blue) as well as AI approaches (more or less attempting to emulate
humans). CS approaches can exploit underlying structure and be
effective in some domains, more effective than humans. By using
evaluation functions and alpha-beta, Deep Blue exploits structure
in chess quite effectively. PARADISE, an AI chess program also discussed,
is far less effective, although very interesting in other ways. 
However as discussed, CS approaches won't 
work for Go, and AI approaches won't understand either. 

One reason AI approaches have trouble is that they are hand coded.
Finding a program that exploits structure is an NP-hard problem, and
humans are no more capable of solving it than they are of solving
hard travelling salesman problems. The only hope, I suspect is
some form of evolutionary computing, and *What is Thought?* describes
new techniques for evolutionary computing that have been remarkably
effective at exploiting structure in some hard domains, although
I make no guarantee they will work for Go with current computational
resources.

Go has an enormous amount of underlying structure. It is basically
defined by 9 rules on an n by n board. If a Go program does not
exploit the underlying structure, it does not understand, and 
will not play effective Go. People exploit the underlying structure
by building programs on top of modules coded into our DNA.

*What is Thought?* will not tell you how to solve Go. If I knew
that, I would have solved it myself. I've worked on the problem.
But I believe *What is Thought?* will give you a lot of insight 
into the nature of the problem, the nature of understanding, 
the nature of human reasoning abilities and how they differ 
from computer science approaches, as well as describing 
important techniques in evolutionary
programming, and addressing mind more generally. 

Also, it gives a pedagogical review of many fundamental concepts
in computer science and computer game playing.

Later today or tomorrow the web cite www.whatisthought.com should
be up, with more information on the book, my previous publications
in game playing, and other information.
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://computer-go.org/mailman/listinfo/computer-go