Chess and the art of defeating a computer programmer

[TheExcession]

I've decided to pop this post into the science forum not the sport, lounge or computer forums.

Has anyone else taken a look at this?

Roger Penrose poses a Chess situation which Humans can solve but computers can not

The three black bishops taint the scenario (in my mind) but I guess it might have been possible to run two pawns to the back line and claim a bishop.

White to play, what would you do?

[purplepolarbear]

Keep the king on a white square - the bishops won't be able to check it if it is. Don't move any of the pawns. You'll be able to keep the king on a white square for 50 moves and claim a draw under the 50 move rule (where a pawn hasn't been moved and no piece captured for 50 moves).

[fido]

Has anyone tried running this on a computer - not sure I believe the proposition that a computer cannot work out a solution for black. If the algorithm cannot force a win, then it will look for a draw i.e. move the king to a better position or attack the bishops - which would achieve the same result as a human player than can see the end-game immediately?

[Simpo Two]

The Turing Test but with chess pieces?

[CrutyRammers]

fido said:

Has anyone tried running this on a computer - not sure I believe the proposition that a computer cannot work out a solution for black. If the algorithm cannot force a win, then it will look for a draw i.e. move the king to a better position or attack the bishops - which would achieve the same result as a human player than can see the end-game immediately?

Yeah I think the reporting is rather gash. Chess computers are not (generally) AI, they are just predicting and scoring all possible moves close in, then look ahead down the most promising sequences as far as possible. I can see that they're trying to work out how a human might just "get it" in a sudden thought, which is quite different to what a computer does, and using this problem to elicit that difference. But I don't buy that it can't get a draw just by looking ahead and avoiding moves which lead to a loss.

[foreright]

Sorry, ape rose may be a brilliant scientist / programmer but he clearly doesn't play chess. There are literally thousands of this type of problem that the "computer cannot solve", may story due to the fact that the computer has a finite "horizon" it can't see past. This has nothing to do with AI and everything to do with an inherent limitation in the way that computers "think" about chess. I don't think it would be that difficult to add a heuristic that detects a draw in this instance but it's such an edge case that it's not worth the effort.

IMO of course.

[PurpleAki]

May be of interest to some

Storyville, Kasparov and the Machine:

www.bbc.co.uk/iplayer/episode/p03rq51h via @bbciplayer

[TheExcession]

PurpleAki said:

May be of interest to some

Storyville, Kasparov and the Machine:

www.bbc.co.uk/iplayer/episode/p03rq51h via @bbciplayer

Is that the one where he plays big blue? IIRC he accuses the team of giving the computer hints as he considered there was no way it should/could gave played certain moves.

[PurpleAki]

Yeah. It's excellent viewing. Others on the programme tend to agree with him. Plus some secretive behaviour by IBM.

[Fugazi]

PurpleAki said:

May be of interest to some

Storyville, Kasparov and the Machine:

www.bbc.co.uk/iplayer/episode/p03rq51h via @bbciplayer

Thanks for that, will bookmark it to watch over the weekend!