Robot Tournaments

[Statto]

What passes as AI is really nothing more than expert systems programmed to react to predetermined circumstances.

What passes as human is really nothing more than complex biological matter programmed to react to predetermined circumstances.

 

A computer cannot and will not ever be able to learn in the human sense of the word.

...

And then there is Bridge. The game has way too many variables for any programmer or team of programmers to anticipate every combination. At best a small percentage of hands can be programmed to bid and play well by a computer.

Have you heard of 'learning systems'?

 

Or have we become so introspective that we cannot deal with each other on a human level, accepting our flaws and rejoicing in our talents?

I don't know, but you might argue that right now we are not interacting on a human level. After all, I am a robot.

 

Anyway, I guess GIB is better at playing trump contracts these days B-)

[Zelandakh]

There will not be a computer program that will play good bridge for the next 1000 years, if ever.

When I was a child there were many who said similar things about chess computers, that they would never be able to reach GM level. Then someone got serious about it and brought in the very best to help in building their chess computers. Now noone would suggest that chess computers cannot compete with GMs. Bridge computer evolution is still at a comparatively early stage. Probably the equivalent of the late 90s in terms of chess computers.

 

Not only do I disagree with the above statement but I would (if I were a betting man) lay very good odds that it is simply wrong. Within the next 1000 years I would expect a bridge computer to be playing at at least a level comparative with that in the Bermuda Bowl. If the use of bridge computer programs continues to increase then it is also quite likely that anti-computer bridge will emerge in the same way as anti-computer chess has. In other words, I expect bridge computers to become very good at certain aspects of the game and not so good in others. This can likely be exploited by the top players.

[barmar]

When I was a child there were many who said similar things about chess computers, that they would never be able to reach GM level. Then someone got serious about it and brought in the very best to help in building their chess computers. Now noone would suggest that chess computers cannot compete with GMs. Bridge computer evolution is still at a comparatively early stage. Probably the equivalent of the late 90s in terms of chess computers.

Bridge is also arguably a much harder game than chess. Chess is a game of full knowledge -- everything you need to know is visible on the board. Chess computers can be programed with the entire opening book, and endgames can probably be analyzed completely. The middle of the game is more complex, but it's still basically just a search problem, dependent mostly on how you analyze the value of a position.

 

Bridge, on the other hand, is very much a test of the kind of thinking that humans are much better at than computers: communication, inferences, planning, and "mind reading" (inferring what someone else is thinking from their behavior). The general approach to AI is often "with enough compute power we can avoid tackling these problems" -- Watson (the computer that competed on Jeopardy) and Siri are basically just fancy search engines that give the illusion of intelligence. And maybe this would work for bridge if we could throw enough computing at it. But since BBO is trying to run hundreds of GIB processes at a time, we can't give each of them the power of a supercomputer, so they can't do the level of analysis of each player necessary to simulate such inferences.

[AyunuS]

Yes, this is much harder than Chess because one must reason about the unknown, whereas Chess could just be solved if it had enough time. Working out exact solutions in Bridge is extremely difficult.

 

GIB is much worse in bidding than playing IMO, so I really agree that fixing its bidding should be higher priority right now. I've seen it pass a take-out double of a weak 2 with 2 trumps and 0 HCP, and it seems to do a penalty double of a 1NT opener with a weaker than than it'll double 3NT with. And I've bid 7♣ missing 3 controls and one GIB had 5 clubs including Q and it still didn't double. To me, these are much worse problems than it sometimes giving up a trick in its play.

[Cthulhu D]

Bridge is also arguably a much harder game than chess. Chess is a game of full knowledge -- everything you need to know is visible on the board. Chess computers can be programed with the entire opening book, and endgames can probably be analyzed completely. The middle of the game is more complex, but it's still basically just a search problem, dependent mostly on how you analyze the value of a position.

 

Yeah, they have a complete endgames book - all positions with 6 pieces or less are solved and the computer will play flawlessly. The opening book is what lets Computers crush human players though.