Computer engine aesthetics?

So one often times argued point against engines is their lack of creativity. Humans make mistakes, but humans also sometimes find very wonderful moves that are heralded as being extremely creative thinking and in a way, therefore, beautiful. Computers do not generally play for aesthetics, they play simply on tactical and strategic algorithms designed to find the most efficient way to win.

However, sometimes I find that the computer, in their efficiency, can coincidentally create a thing of beauty! For instance, this game I just played: http://en.lichess.org/2V3nR5nl/black#0

I missed a mating sequence at move 31. Check out Stockfish's suggested mating sequence. It is a thing of beauty (especially note its suggestion of 34. ...Qd3+)! In an over the board game, a player finding this sequence, playing the stunning Qd3+! would be praised for brilliance! Am I wrong to think that Stockfish came up with a very elegant mating sequence here in its algorithmic efficiency?

Yes, it can work out that way, but generally only when the engine's responses correctly matches the skill level of the relevant players. If it doesn't do this (pretty much all the time) it will effectively assume that people will play the best counter moves that it has found. Consequently many games are not as efficient as they might be as the engine pads out a move sequence with threats only it sees.

A good example is my second game of the 90+30 tournament. The more thorough Stockfish analysis I ran afterwards told me that my 13th move was an error (it doesn't say that here on lichess.org, but only due to lack of time, the analysis I ran off site went a lot longer and deeper). It then recommended a series of moves which took the game to move 38, had me lose all my pieces except my Queen and still wasn't finished.

In the actual game the result was a checkmate at move 17. You can see it here:



I believe that game may be one of the most eloquent I've ever played. It certainly made a very good candidate for me to try my hand at annotating a game.

Not that engines aren't useful, they are, but it's important to understand precisely what they are in order to both get the most out of them and see their limitations.

The game you posted is indeed a bit elegant. By move 11 your opponent has given you complete control of the board. How nice! But what you say about your move at 13; if your opponent hadn't made two consecutively bad moves, it would not have resulted in a mate so soon. In fact the lichess engine DID rate that move poorly, it dropped your advantage by nearly half a point. However, it only annotates moves that raise/lower your advantage by 1 or more points.

Regardless, though, what I meant by efficiency is the fact that they plug-n-chug the most direct way to play a position assuming direct play from the opponent as well. In other words, a computer never looks at a position and has a "gut" instinct to make a brilliant exchange sacrifice, or a beautiful tactical play that results in an aesthetically pleasing finish. So, in that regard, what I was saying is that they are generally seen to be unable to be creative.

I was pointing out the fact that on rare occasions, the engine finds a sequence that could be seen as eloquent or beautiful...almost human. Beauty arising from form...very interesting concept to me, living in Japan.

#1: To find and play 34...Qd3+ over the board, your opponent first must be willing to play 34. Qc3. I doubt any human would make that move OTB instead of just resigning.

Haha! Fair point.

Arguement is not fair because computers/chess engines can not make any arguments supporting their own way of looking effeciency as creativity. They are designed by humans so they naturally are dumb. Machines can't think so there is no possibility of "Creative thinking" as they are designed by humans to be good at something that they themselves are not good - efficiency.

Computers/chess playing softwares generally don't play for aesthetics because they are not designed to play in such a way. If you wish you can modify code and design them in such a way that it makes moves which are more elegant/aesthetically pleasing in your eyes, as beauty is a realative thing. For stockfish king's gambit is same as caro-kann untill evaluation remains in control say -+ 0.5 or half a pawn.

Still moves made by most chess playing engines are not 100% efficient either because algorithms designed by humans are not 100% efficient. And all chess engines are doing is just mathematical calculations based on these algorithms. Endgame tablebases are perfect example of 100% effecient chess engine with 7 or less pieces(currently available), and no engine that plays opening or middle game will be able to achieve the same efficiency as entire energy in universe is not enough to fire such a computer to play 100% efficient move in the middle game.

Frankly speaking I don't find solution/sequence give by stockfish after move 31 in this game - lichess.org/2V3nR5nl/black#0 - really very amazing or elegant, but as I said it depends how you define beauty and elegance. Mating sequence in this game suggested by stockfish was not very elegant in its own eyes and it was just more efficient than other alternate moves that it calculated using algorithm that itself is not very efficient or elegant.

<Not that engines aren't useful, they are, but it's important to understand precisely what they are in order to both get the most out of them and see their limitations.>
You are right Hasimir but no human brain or computing machine is good enough to achieve what you are trying to achieve. Their limitations are impossible for humans(chess players not programmers) to find because they are more efficient than humans. In order to answer - 'what they are?' - I would say they are algorithms that can calculate as instructed move/s tree to a certain depth with the help of available processing power.

It is not impossible in theory to design a chess playing software with a "gut" instinct to make a brilliant exchange sacrifice, or a beautiful tactical play (they already are better at tactics compared to humans) that results in an aesthetically pleasing finish. Though for them engines it will be just another kind of alogirthm to calculate a finish that is more aesthetically pleasing to humans. Here too we need to have more than one such engine according to different personality types as same move is pleasing to one eye and may burn eyeballs of another type.

<I was pointing out the fact that on rare occasions, the engine finds a sequence that could be seen as eloquent or beautiful...almost human. Beauty arising from form...very interesting concept to me, living in Japan.>
It happens sometimes but it is just a co-incidence as in other fileds too this happens. Sometimes scientific theories inspire painters to paint something based on it and other way around. Even non-working wall clock shows right time twice a day that doesn't mean it is working at that particular point of time. So untill machines can have conscisouness and can start thinking on their own you can enjoy such rare occasions of eloquent and beauty as just happy coincidences.