in reply to Re: Artificial Intelligence Programming in Perl in thread Artificial Intelligence Programming in Perl
So I am confused... Do you mean that chess playing programes evaluate all moves or do they only evaluate some? You have made the statement that such heuristics as used in the chess playing computer are either not AI or show AI to be shallow. A difference you gave was that people, supposedly with intelligence of the non-artificial kind, are able to prune out unbeneficial moves quickly. I provided documentation that modern chess playing programs do just that as well as a response that your example difference was unsatisfactory to support your statement.
Okay, we don't "know" how humans make this pre-selection; does it matter? does the mechanism with which the selection is made determine whether it is the exhibition of intelligence? Or is the ability to make the selection and decision the exhibition of intelligence?
If it is the mechanism which is the difference and we don't know what that mechanism is, then how can you state that the mechanism used in AI is shallow or unsatisfactory for the problem set?
Re: Re: Re: Artificial Intelligence Programming in Perl
by ariels (Curate) on Jul 03, 2002 at 12:52 UTC
|
Of course there is a difference between how humans play chess and how computers play chess. Consider: How many positions does Deep Blue examine to decide on its next move? -- A quick search on the web seems to suggest 250 million moves. A second. How many positions does a human grandmaster examine to decide on its next move? 100_000 sounds like a huge overestimate (and would be unsupported by any psychological research). In the 80s, David Levy (I think) investigated human chessplayers; he found that for "real" positions (i.e. those occurring in a real game) masters can produce a good move in much less time that for "unreal" positions. Perhaps humans have some capacity for playing chess beyond alpha-beta pruning with a good ordering heuristic??
This hardly points at some difference in heuristics betwee
Humans aren't slightly better than computers at chess (and much better than computers at go) because they analyze more moves. On the contrary: they're better despite analyzing only an insignificant number of moves!
| [reply] [Watch: Dir/Any] |
Re: Re: Re: Artificial Intelligence Programming in Perl
by Corion (Patriarch) on Jul 03, 2002 at 13:03 UTC
|
The idea is that a human, contrary to the computer, has
the idea of "worthless" moves and a special preselection. So a
grandmaster only considers the "good" moves (of which there
are maybe 5 or 10), while the computer has to scan through
all possible moves and rate them to even find out what "good"
moves are. Vast libraries for the opening help the computer
reduce the initial tree of possible moves down to a set of
"traditional" moves.
Another side point might make this phenomenon more plausible
to you :
If you show chessboards with regular play situations on them for a short time to people,
those who play chess will more accurately reconstruct the board than those who
don't. But if you show them pieces randomly placed on a board, both groups
will show a similar rate of errors. This could be interpreted
that the human does not see the places of all pieces on the board
separately but in relation to the other figures, remembering
patterns he has seen/experienced (in play) before.
A game that solely relies on patterns and their recognition
is Go, a game that is considered very hard to write computer
opponents for.
perl -MHTTP::Daemon -MHTTP::Response -MLWP::Simple -e ' ; # The
$d = new HTTP::Daemon and fork and getprint $d->url and exit;#spider
($c = $d->accept())->get_request(); $c->send_response( new #in the
HTTP::Response(200,$_,$_,qq(Just another Perl hacker\n))); ' # web
| [reply] [Watch: Dir/Any] [d/l] |
|
The idea is that a human, contrary to the computer, has the idea of "worthless" moves and a special preselection
I am going to bow out of this now, because all the responses are now devolving to "humans do something, we don't know what, but it ain't what computers do". This is not proof or reasonable support, but merely the naysaying of something.
| [reply] [Watch: Dir/Any] |
|
What reasonable support do you want? Grandmasters *have*
been questioned and observed. It *is* know what they do
(examine just a few number of moves), we don't know how
exactly they select the moves to examine. What they certainly
do know is that they don't examine all the positions a computer
examines. And there's a simple proof: take a random grandmaster,
playing a random game. Measure the time it takes to make a
move. The average of those times is less than 50 years. qed.
There's no magic, handwaving or naysaying going on.
Just because you don't know everything doesn't mean you cannot
exclude everything. Suppose I know X was in a house. I don't
watch the house, but I do watch the road going south from
the house. Now, X is no longer in the house. According to
your reasoning saying "I don't know where X went, but I know
he didn't go south" is not a proof or reasonable support,
but merely naysaying of something, because I don't know where
X went.
But have you presented any proof or reasonable support that
how computers play chess has any similarity to how humans do?
Abigail
| [reply] [Watch: Dir/Any] |
|
|
|
|
Re: Artificial Intelligence Programming in Perl
by Abigail-II (Bishop) on Jul 03, 2002 at 13:02 UTC
|
They evaluate all moves from the current position, to a
certain depth, then start pruning. It's obvious they cannot
evaluate all moves till the end of the game - that would
mean chess would be "solved", and it isn't. It will technically
unfeasible to solve chess the brute force way for many eons
to come. But before pruning, even at the earliest level, a
chess program must consider all moves from
the current position - after all, pruning is done based on
evaluating positions.
Okay, we don't "know" how humans make this pre-selection; does it matter? does the mechanism with which the selection is
made determine whether it is the exhibition of intelligence? Or is the ability to make the selection and decision the exhibition
of intelligence?
The ability to make selections and decisions doesn't make
intelligence. After all, any program with an if
makes a decision.
(But I don't want to go into the question of "what is intelligence"?
That's a holy way I don't feel stepping into.)
If it is the mechanism which is the difference and we don't know what that mechanism is, then how can you state that the
mechanism used in AI is shallow or unsatisfactory for the problem set?
I never said that the mechanism used in AI is shallow. What
I said was that if you call the rather brute force technique
of chess programs (despite some pruning, the technique is brute
force, as it considers way more paths than are needed for
the "solution") AI, then the term AI becomes shallow. Brute
force techniques are amongs the simplest techniques of solving
problems.
Abigail | [reply] [Watch: Dir/Any] |
|
|