[ExI] spike and chess
interzone at gmail.com
Thu May 16 19:39:53 UTC 2019
I wouldn't call it creative. It is not generating something sui generis.
While the results are impressive in terms of play, it is just optimizing
for a global minima in the total space of available moves. It is a brute
force technique based on learning existing winning play patterns, and
optimizing for it with math. Its moves may look like the work of a mad
genius from outside, but at its core, it's really just a few equations that
have been optimized for using training data.
I would not define anything that we have seen out of deep learning so far
as intelligent. The AI opponent doesn't have any meta view into the
game. It's given a goal and optimizes for it using linear
alegbra/calculus/monte carlo techniques.
This is a good overview that they were kind enough to share.
This gets into the details:
On Thu, May 16, 2019 at 3:01 PM William Flynn Wallace <foozler83 at gmail.com>
> chess programs got to where they were just as creative, imaginative,
> artistic and all that as the best humans, in fact… better. So… apparently
> chess doesn’t require intelligence. spike
> This is preposterous - Why can't what the chess computer is doing be
> intelligent? Because you define that what it is doing not intelligence.
> It seems really out of step when you consider that chess geniuses are
> generally pretty high IQ people to start with. I think you just don't want
> to call it intelligent and are wiggling around out of calling it that. For
> my money, creative, inventive, excellent recall all point to superior
> Here's an idea: restrict the chess computer to the memory level of a
> great chess player - even the field. Recalling millions of games is an
> unfair advantage, it seems to me.
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat