Re: computer-go: FPGA

[Don Dailey <drd@xxxxxxxxxxxxxxxxxxxxx>]

> Hoping that increase in computer-power will automatically give a good
> go-program is a lazy thought. 

No, the goal isn't to depend on the hardware to  play good go, it's to
figure out how to fully utilize whatever hardware  you happen to have.
Is it lazy to wish for this?



> I also think it's not realistic, at least not
> until someone finds a good way to make a good self-learning program. With
> the power of computers today there's still so much possible, so much to do
> to improve a programs play. The main problem is one of man-power. How to
> find enough computer-programming resources to make a better go-program.


This all sounds so familiar.  This  is how it  used to be argued about
computer chess 20 years ago.  It was considered completely unrealistic
to expect faster computers to play  better chess.  It actually came as
a  big surprise to  almost  the  whole computer  chess community  that
faster  computers equalled much better chess.    Many assumed it would
help a little, but no one imagined it would help so much!   

What confused  the  matter so  much  in  those days was  that  no  two
programs were  very much alike.   Most  programs in fact  didn't scale
very well  to increasing  hardware (like  todays go  programs) and the
quality of  the program is what  dominated the results.

I also remember that about the same  time chess computers were getting
close to the expert level, an article was written somewhere, in a very
authoritive way, explaining that computers were  pretty close to their
theoretical  limit of strength and  that extra  computing power was no
longer much of  a factor in the playing  strength.  This was based  on
the  idea that computer  rarely  made tactical  mistakes and  yet they
still made sometimes very  basic  strategic mistakes that couldn't  be
solved by simply looking a few moves deeper.

These arguments sounded logical and intuitive, but in fact were proved
to be complete nonsense!  Computers  kept improving, almost in perfect
proportion  to computing power.  People kept  imagining that there was
some magic barrier  of strength  which  computers would  never  exceed
(because they were  in such awe of  human masters.)  But there  was no
magic barrier and it was found that a  properly written chess programs
scaled extremely well  with   computing power (memory  and   processor
speed.)

I really believe  that the source  of  confusion here is  that we just
haven't learned to write  scalable go programs yet.   To me  this just
means the programs (probably)  have not been  written correctly in the
sense that they do not utilize avaiable  computing resources very well
at all.

I'm not being critical of anyone and  I certainly have  no idea of how
to do this myself either.  But I believe future progress will be based
on someone figuring out how to do this!


Anyway, I agree with  you when you say that  there is so much possible
with  todays  computers and   that   the most  important  resource  is
man-power!


Don