[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [computer-go] Modern brute force search
At 04:33 PM 11/7/04, you wrote:
... how to program an infinite speed computer to play chess. ...
The obvious way would be to search to the end of the game ...
...an "omniscient" player ...d never lose a game even to "god."
But would that be the best way to play against a human?
sure if you knew the strategy.
Perhaps the game theoretic value of chess is draw, ...
But it would be easy to imagine an omnicient computer applying
heuristics ...
i would think there would be no need to.
I admit I'm not an expert on GO by any stretch,
i am an amateur 1 dan, an old programmer (but not an ai type) and no expert
on chess.
but I have been
speculating that an "omniscient" computer GO player would be far ahead
of human masters.
sure, in an even game, but i can not fathom that the value of a game with a
large handicap is anywhere near positive for white. a one stone handicap is
a *big* advantage. the difference in prefessional dan levels is 1/3 of a
handicap stone. this means that there is a difference of only about 4
stones between the weakest and strongest pro's. (1 dan pro = 6 dan amateur
approximately)
I suspect that an omniscient GO player could be
programmed to play in the same way strong players beat weaker players
with handicaps, as you say, "leaving everything unsettled for as long
as possible."
sure, it could be, but why bother? the pro's will always leave as much as
possible unsettled. i try to do this in my games.
also, a typical asking move is to move at 4-2 (attaching under a 3-4 stone
that is a corner enclosure with a stone at 3-5) and then make a move that
is a distance (manhatten) of say 9 away.
It sounds like GO isn't as complicated as I thought it was, due to the
fact that many of you believe masters are pretty close to optimal play.
iirc, the pro's that play shogi say it's like playing 6 games at once.
For instance, Jonathon Schaeffer ... "checkers" ...
noticed that no matter how deep the program looked, it seemed to
benefit from looking even farther. ...
It seems this trivial game was deeper and more profound than anyone
believed.
I see chess as being much deeper and much more complicated. I see GO
as being much more profound than chess. Am I wrong about this?
don't know, if my recollection of what the pro's think is correct then it
is by a factor of 6 (but this is for a human).
my very limited understanding of chess program is that they do not "know"
how to play chess. i realize that no progam does "know" any thing, but i
mean this in the sense that they do not play like a human would.
in the beginning and middle of the game, the branching factor is so bad
that the the only hope i can imagine is some combination of heuristics and
patterns. and i can not see a program being as good as today's 9-dan's.
even given a speed improvement of many orders of magnitude.
i have played for a long time (about 30 years). i use my experience and a
set of heuristics and principles that i got from Yu-Lin Yang (an excellent
teacher). i figure with these heuristics and a whole bunch of other stuff a
program that can beat me is possible (with a *lot* of work) on todays
hardware. the programs out there are rated at around 3-4 kyu. but they are
not that strong as i can give them 6 stones and win without thinking to
much. so they are probably around 8 kyu (using many faces 11.0 for
example), dave may have a stronger version for tournaments, but i doubt
that is more than one stone stronger.
If it's the case that GO is simpler than even Chess, I would retract
my speculation on this. I know that it is often held that GO is
simpler than chess for humans, just harder for computers.
the rules are simpler, but the play is much more complicated. what is
important varies over time. capturing 1 stone can be worth 1/3 of a point
(a simple ko) or as much as 30 points (ponuki in the center).
But I don't
agree with this. A strong chess player who is also a realtively good
GO player and an expert at many games (Larry Kaufman) once told me
that there are many more levels of skill in GO than in chess.
yes. players start around 20-30 kyu and it's one handicap stone per level.
i would guess that chess is mostly reading. you can not read in the
beginning of a go game. reading in the middle is *very* hard. sakata was
said to be a master at this.
also, many different skills are required. the pro's can read out the end
game. but the beginning is more right brain (this move feels right). some
people are good at fighting, some are good with moyo's, some know many
joseki, some know many fusekis, some know may tesuji's, some are good/with
against small group to small group games, so style and skill set can figure
prominently. another skill is timing. when do i make the asking move? when
do i make the invasion? (one move before he can make it impossible or make
you suffer greater consequences elsewhere)
one can clearly build in fuseki and joseki knowledge, and ignoring how to
choose a joseki (a non trivial task). even if you get to the early mid-game
with things still even, i still don't see a program doing that well against
say an american 3 dan let alone the worst pro.
In
other words, as he explained, the median GO play has less of a chance
beating a top GO player than a median chess player has of beating the
top chess player.
the median go player may be as strong as 6-kyu. he has no chance against
any 1 dan, let alone the worst pro.
I have been told that it takes longer to "master" GO than chess,
although I don't really know what this means.
me neither, but i suspect that go takes longer for most people as there are
many skills and concepts to learn.
I know it's possible
for a very young person to be world chess champion. How old are the
world champion GO players?
the best pro's start learning around 8-10 and they are better than me
(amateur 1 dan) in a year or so. some pro's win some top tournaments in
their early twenties. don' t know what the average age is.
So my view (and I'm sticking with it) is that we may someday find out
that great GO players are not as great as we think they are, but it's
just my opinion. I certainly don't believe they are playing anywhere
near optimal, although it probably seems that way to us mere mortals.
- Don
we may, but it might take quantum computers to do it.
my suspicion is that they are near optimal for a 19x19 board. they tried
some experiments with a 21x21, but i don't know the results.
on very small boards, it looks like playing in the center is a win. if this
generalizes that could be interesting. but most people do not play that way
(they go corner, side, middle, or they build a moyo, or they start a large
scale fight early). any program that is any good is going to have to know
what style of game it is in in a very small number of moves
thanks
X-Sender: rtayek@xxxxxxxxxxxxxxxxx
Date: Sun, 07 Nov 2004 15:42:21 -0800
From: Ray Tayek <rtayek@xxxxxxxxxxxxxxxxx>
X-Scanned-By: MIMEDefang 2.42
At 10:51 AM 11/7/04, you wrote:
>...
>And that is a shame. I strongly feel that even when humans are clearly
>surpassed, we are just getting started. Chess is an incredibly deep
game
>and the very best humans are very weak compared to the "ultimate player."
>
>Just my thoughts. All of this is even more true of Go. There exists
>(in principle) a future GO player than can make the very best Go
>player (of today or yesterday) look like a baby.
i don't think so. if by "look like a baby", you mean able to give a large
handicap (say 6 stones), then no way.
i don't see any evidence that human players can become that much stronger
than todays professional 9-dan's and i doubt that a program will beat
them
for a *very* long time (think about "asking" moves, or watch a 9 dan pro
give an amautuer 7 kyu player a nine stone handicap - the pro will leave
everything as unsettled as possible for as long as possible). ...
---
ray tayek <http://tayek.com/>, co-chair <http://www.ocjug.org/>, actively
seeking telecommuting work.
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/