Pages

Saturday 9 December 2017

Mere calculation

Not everyone plays chess so I’ll try to minimise the chess aspects of this post because there is an interesting addition to the previouspost. In one of the games between AlphaZero and Stockfish 8, AlphaZero made what the YouTube narrator calls a deep positional knight sacrifice.

If AlphaZero had been a human player then this knight sacrifice may well have attracted more superlatives - for example it might have been described as brilliantly imaginative. Computer chess can sometimes seem imaginative, brilliant, clever and strategically creative despite every single move being the result of calculation. There are sometimes subtle differences between computer and human chess but usually we cannot tell the difference unless the human makes the kind of mistake computers don’t make.

It is mere calculation, but can seem like brilliant positional insight. Mere calculation – is that all we do too? Not in quite the same way of course, but the question still hangs in the air.

6 comments:

Sam Vega said...

I don't play chess, but I think I get your general point here. If so, I reckon that brilliant positional insight is what people do when they play as well as computers, but without the necessity of the microsecond-long grind to get there. What makes us stand up and cheer human excellence is that it all happens in front of us, rather than being the product of many tedious years of technological development. A computer can do it, but a man did it; and we saw him!

Scrobs. said...

Both Mrs O'Blene and I are Sudoku fanatics. She devours several of them daily, from the easiest to the fiendish, and often employs the written techniques of writing in pencilled candidates when she feels fit.

My approach is only similar in that I only do the writing-in on the really diabolical nastiest ones, and prefer to do 'The Times Fiendish' without resorting to that method. Sometimes a single puzzle can take anything up to five weeks. I've got to No 102 in about a year and a half, so the book is getting a bit 'tired', but so far there are no written candidates in any of them...

Like a chess quandary, I've recently been pondering how various moves pan out for the next seven or eight numbers - again in my head, and there is the technique for solving Sudoku which indeed does just that by trial and error. It's also well known that one can be stuck on a break for several days, and then in a flash, the answer appears, and the whole puzzle dissolves in a few minutes.

So, to complete the analogy (about time Ed.), I am still trying to 'see' a result of five or seven moves ahead without writing them in, and while there's no scientific result, I sometimes feel that the old brain knows, or at least is learning to recognise what's coming next, and responds accordingly!

Chess has so many millions of combinations, whereas Sudoku is limited, but with fewer blatant options on the more difficult puzzles. I used to have a chess programme on an old Amstrad 464 PC, and didn't get very far then either, although I tried!

Phew, I feel an easy crossword coming on..;0)

Sackerson said...

I think there may be a difference in styles of thinking. Presumably the computer looks at all possible positions in detail, as far ahead as it can; beyond a certain point the human mind works on hunches?

A K Haart said...

Sam - I'm sure you are right because we still admire erudite people even if they know almost nothing we couldn't look up for ourselves on Google. Will it last though?

Scrobs - Mrs H does Sudoku but I've no idea what the puzzles are about. The huge advantage chess computers have is their ability to avoid mistakes by looking many moves ahead in spite of those millions of combinations. Like Sudoku the flash of insight takes us so far but there is a limit to our ability to foresee these things and now computers can exceed that limit.

Sackers - yes, as I understand it the computer does look at all possible positions and much of what it does must be brute force, the ability to calculate further.

Yet the computer also looks at patterns, advantageous and disadvantageous and that seems to be what the best human players do to. Humans 'see' the patterns but this may be a memory effect in that the best players remember and recognise a vast number of patterns with their associated advantages and disadvantages.

Scrobs. said...

Mr H, Goosegirl has been trying to post here, but somehow can't make it work (sorry Goosey, I know it does, but I'm dumbstruck...)!

Here's her post, clipped from one of mine, and she can rest easily now...

"Aren't hunches the same as a gut instinct? If so, that could imply an entwining of previous experiences and knowledge together with one's innate persona that would far outweigh any computer because they work on rationality, whereas we work on a different level. As Sam Vega said, many men have achieved things that any computerised system could not have done e.g. Alexander Fleming and his discovery of Penicillin because his lab had been left in an untidy state. I do the Graudian cryptic crosswords to keep my brain fit, and there are times when I instinctively know the answer but can't get it from the clue until I see it the following week; even then I sometimes can't fathom it out".

Phew, what a scorcher...

A K Haart said...

Scrobs / Goosey - people sometimes have problems commenting here but I never seem to find out why. It is necessary to have a Google account or OpenID because otherwise I get spam with dodgy links.

The interest in computer chess is that this particular computer is not a specialist chess system and seems to have taught itself to play to an unbeatable standard within a matter of hours. I wouldn't care to predict any absolute barriers to its future abilities.