I can....parrish wrote:I can't believe Jeopardy! didn't have a category about computers with A.I. (fictional or non) with answers from movies like War Games, 2001 A Space Odyssey, Terminator, Star Trek, etc.



Moderator: Ras
I can....parrish wrote:I can't believe Jeopardy! didn't have a category about computers with A.I. (fictional or non) with answers from movies like War Games, 2001 A Space Odyssey, Terminator, Star Trek, etc.
It seems Kasparov still hasn't given up the idea that humans are superior to computers at chess.Garry Kasparov wrote:(...) chess computers. They play fantastically well in maybe 90% of positions, but there is a selection of positions they do not understand at all. Worse, by definition they do not understand what they do not understand and so cannot avoid them.
The statement you quoted doesn't support that claim, it simply says chess computers don't do well in a selection of positions. Humans probably play fantastically in 10% of positions.rbarreira wrote:It seems Kasparov still hasn't given up the idea that humans are superior to computers at chess.Garry Kasparov wrote:(...) chess computers. They play fantastically well in maybe 90% of positions, but there is a selection of positions they do not understand at all. Worse, by definition they do not understand what they do not understand and so cannot avoid them.
But he also says that computers can't avoid those positions they can't play in. Which immediately implies that humans could win by tricking computers into those unavoidable positions they don't play well, according to Kasparov's assumptions.Leto wrote:The statement you quoted doesn't support that claim, it simply says chess computers don't do well in a selection of positions. Humans probably play fantastically in 10% of positions.rbarreira wrote:It seems Kasparov still hasn't given up the idea that humans are superior to computers at chess.Garry Kasparov wrote:(...) chess computers. They play fantastically well in maybe 90% of positions, but there is a selection of positions they do not understand at all. Worse, by definition they do not understand what they do not understand and so cannot avoid them.
Yeah, the computer has a faster reaction time. How are you going to fix that?karger wrote:This was not a fair game whatsoever .
Unplug the computer, that usually kills its reaction time.Dirt wrote:Yeah, the computer has a faster reaction time. How are you going to fix that?karger wrote:This was not a fair game whatsoever .
In any case, just getting to the point that the reaction times mattered was quite an accomplishment. Congratulations to IBM.
In 40 years you may find that old trick rather hard to pull...JManion wrote:Unplug the computer, that usually kills its reaction time.Dirt wrote:Yeah, the computer has a faster reaction time. How are you going to fix that?karger wrote:This was not a fair game whatsoever .
In any case, just getting to the point that the reaction times mattered was quite an accomplishment. Congratulations to IBM.
I don't fully agree about low probability of a learned position to be hit in search. Let us suppose that the learning engine runs a special book where it always plays the same opening lines... E.g. always english as white, always Caro-kann, Nimzoindian as black. Let us suppose there's only one move for learning engine in that book, while it includes all possible opponent reply. That book would consist of few hundreds variations. The engine would learn the related middlegame position fast enough.Dann Corbit wrote: ....................
Position learning is 100% effective if you hit the same position again (but this has a surprisingly low probability). Suppose (for instance) that you are playing a chess game and encounter this position:
[d]2r3k1/4ppb1/2P5/4P2p/2R3p1/1p6/1B4PP/5K2 w - -
Your chess engine makes a bad move and writes out a record that stores the correct value after the opponent's move. The odds that you are going to play this move again are basically zero, unless it is near the origin (in which case it is probably a book move).
......................
I think that there is a long way to go before computers utilize chess statistics properly in learning and move selection. It is actually something that I am actively working on.
I don't fully agree about low probability of a learned position to be hit in search. Let us suppose that the learning engine runs a special book where it always plays the same opening lines... E.g. always english as white, always Caro-kann, Nimzoindian as black. Let us suppose there's only one move for learning engine in that book, while it includes all possible opponent reply. That book would consist of few hundreds variations. The engine would learn the related middlegame position fast enough.Dann Corbit wrote: ....................
Position learning is 100% effective if you hit the same position again (but this has a surprisingly low probability). Suppose (for instance) that you are playing a chess game and encounter this position:
[d]2r3k1/4ppb1/2P5/4P2p/2R3p1/1p6/1B4PP/5K2 w - -
Your chess engine makes a bad move and writes out a record that stores the correct value after the opponent's move. The odds that you are going to play this move again are basically zero, unless it is near the origin (in which case it is probably a book move).
......................
I think that there is a long way to go before computers utilize chess statistics properly in learning and move selection. It is actually something that I am actively working on.