Garry Kasparov lost his 1997 match against Deep Blue, a chess computer developed by IBM, for several reasons:
1. Deep Blue was significantly faster than previous chess computers and was capable of evaluating hundreds of thousands of positions per second, giving it a distinct advantage over human players.
2. The chess machine was designed specifically to play against Kasparov, with the ability to adjust its strategy based on his playing style.
3. Deep Blue used a combination of brute-force calculation and advanced chess knowledge, which allowed it to make highly tactical moves and unexpected sacrifices.
4. The pressure of playing against a machine may have affected Kasparov's performance, as he was known to become highly competitive and emotional during matches.
It is worth noting that Kasparov won the first match against Deep Blue in 1996, but lost the rematch in 1997. The matches between Kasparov and Deep Blue marked a turning point in the history of computer chess, as they demonstrated the significant progress that had been made in developing machines that could compete against the best human players.
ChatGPT is amazing, it seems to have an answer to almost everything. Try it yourself and ask about on how to program NNUEs.
P.S. ChatGPT can give slightly different answers every time you ask the same question. So you might get a slightly different answer.
Last edited by Cardoso on Sun Jan 29, 2023 4:40 pm, edited 1 time in total.
* DB analysed 200 million positions per second, which was exceptional at the time
* I don't remember it being able to "adjust it's strategy based on his playing style": that would imply changing the eval function mid-game.
* I don't remember a whole lot of "unexpected sacrifices"
But somebody who wasn't around at the time could easily be taken in!
A quick opinion on my part: before fact-checking on the internet became common, information at this level of accuracy was normal, and we used to be happy with it most of the time.
Human chess is partly about tactics and strategy, but mostly about memory
I read the adjustment more as a change that could be made between games. You cannot trivially change Leela to adjust its playstyle, but a handcrafted eval like Deep Blue's you could tweak weights to play more in some style or another.
There was this famous Caro-Kann Nxe6 sacrifice from the final match game. I agree that the DB isn't known for it's sacrifices, but one of the main controversies centered around Kasparov thinking DB would not go for the knight sacrifice in that line.
ChatGPTs answers feel a bit handwavy here, but the only part that is definitely just flat out wrong is the number of positions per second it could evaluate.
It's mostly number 4: Emotions and nerves. I remember that match really well. Kasparov won one of the games, lost another, third was an honest draw. But in 4th he offered a draw in a winning position . In 5th he did not find a winning move and that game also was a draw. (The order of those games could be different, but that's the drama). And then finally 6th game - he blundered in the opening ! Used a wrong order of moves in caro-kahn, and allowed a standard sacrifice. But at that time, absolutely all chess community knew, that Kasparov was stronger than Deep Blue 2. He wanted a rematch never got one. And then in few years, chess engines became unreachable to humans.
schahmatist wrote: ↑Sun Jan 29, 2023 6:45 pm
It's mostly number 4: Emotions and nerves. I remember that match really well. Kasparov won one of the games, lost another, third was an honest draw. But in 4th he offered a draw in a winning position . In 5th he did not find a winning move and that game also was a draw. (The order of those games could be different, but that's the drama). And then finally 6th game - he blundered in the opening ! Used a wrong order of moves in caro-kahn, and allowed a standard sacrifice. But at that time, absolutely all chess community knew, that Kasparov was stronger than Deep Blue 2. He wanted a rematch never got one. And then in few years, chess engines became unreachable to humans.
Kasparov did not offer a draw in a winning position.
He did a mistake of resigning in a drawn position but not offering a draw in a winning position.
I do not think that all chess community knew that kasparov was stronger than Deep blue 2.
I did not not know and there were not enough games to know.
towforce wrote: ↑Sun Jan 29, 2023 4:39 pm
Obviously wrong on several points:
And that's one of the problems with ChatGPT. Reasonable sounding text on the surface, but incorrect information when you actually look at the specifics of what it wrote.
towforce wrote: ↑Sun Jan 29, 2023 4:39 pm
Obviously wrong on several points:
And that's one of the problems with ChatGPT. Reasonable sounding text on the surface, but incorrect information when you actually look at the specifics of what it wrote.
I think its strength lies in creating "good enough" art: have it write a nice poem for your wife - it will do a good job for you. I personally wouldn't use it for factual information (this may change in the future).
Human chess is partly about tactics and strategy, but mostly about memory
schahmatist wrote: ↑Sun Jan 29, 2023 6:45 pm
It's mostly number 4: Emotions and nerves. I remember that match really well. Kasparov won one of the games, lost another, third was an honest draw. But in 4th he offered a draw in a winning position . In 5th he did not find a winning move and that game also was a draw. (The order of those games could be different, but that's the drama). And then finally 6th game - he blundered in the opening ! Used a wrong order of moves in caro-kahn, and allowed a standard sacrifice. But at that time, absolutely all chess community knew, that Kasparov was stronger than Deep Blue 2. He wanted a rematch never got one. And then in few years, chess engines became unreachable to humans.
Kasparov did not offer a draw in a winning position.
He did a mistake of resigning in a drawn position but not offering a draw in a winning position.
I do not think that all chess community knew that kasparov was stronger than Deep blue 2.
I did not not know and there were not enough games to know.
Ok, my bad, yes, he resigned in drawn position. But that's actually proves my point even more. Analysis showed that in all except the last one (when he just blundered in the opening) , he was either better or not worse than Deep Blue. You are right, not enough games were played, but those that were played showed that mostly the nerves and psychological pressure proved to be the weak link. I mean technically he lost, nobody argues, but because of blunders ,- his chess skills (both tactical and strategical) were still better at the time.
towforce wrote: ↑Sun Jan 29, 2023 4:39 pm
Obviously wrong on several points:
* DB analysed 200 million positions per second, which was exceptional at the time
* I don't remember it being able to "adjust it's strategy based on his playing style": that would imply changing the eval function mid-game.
* I don't remember a whole lot of "unexpected sacrifices"
But somebody who wasn't around at the time could easily be taken in!
A quick opinion on my part: before fact-checking on the internet became common, information at this level of accuracy was normal, and we used to be happy with it most of the time.
Interesting. Its throwing information with errors at you. It looks through millions of papers and spits out a statment based on what it read.
