still there is probably room for improvement. human titled players see the whole board at once and use pattern matching to rapidly rule out most moves without calculation. (Reti famously claimed he only calculates one move ahead.). with advances in quantum computing and massive parallelism, compute engines could be designed to do something similar, instead of what they do now-- calculating most moves on the board, essentially brute force.
The future of Stockfish
Sort:
15 hrs ago
0
#2
stockfish updates now just fully utilize advances in computational density. the engine itself doesn't improve all that much. only its analytical depth.
still there is probably room for improvement. human titled players see the whole board at once and use pattern matching to rapidly rule out most moves without calculation. (Reti famously claimed he only calculates one move ahead.). with advances in quantum computing and massive parallelism, compute engines could be designed to do something similar, instead of what they do now-- calculating most moves on the board, essentially brute force.
still there is probably room for improvement. human titled players see the whole board at once and use pattern matching to rapidly rule out most moves without calculation. (Reti famously claimed he only calculates one move ahead.). with advances in quantum computing and massive parallelism, compute engines could be designed to do something similar, instead of what they do now-- calculating most moves on the board, essentially brute force.
#4 quantum computing mostly can't play chess. Quantum computing is usable for quite small range of things. For most of them we still need classic computers, the media just exaggerates a bit
#2*
10 hrs ago
0
#5
#3 I can see you don't know anything about these kind of topics. You can run a chess engine on quantum chips. Imagine running a chess engine on thousands of quantum chips. There is already quantum chip in production.
#5 how confidently. First of all, it is NOT a normal chip. If it's named a chip it doesn't mean it can do same things as normal computer. Do you have basic understanding of what actually quantum computing is? Not media stuff, actuall quantum mechanics principles? Quantum have fundamental flaws which make them useless in most of cases, but they are very powerful in a specific topics like cryptography, simulations of molecular/atom physics and other things.
Fet wrote:
#3 I can see you don't know anything about these kind of topics. You can run a chess engine on quantum chips. Imagine running a chess engine on thousands of quantum chips. There is already quantum chip in production.
You are incorrect. When a quantum computer actually plays chess, I'll be sure to let you know.
#5 you are wrong
7 hrs ago
0
#9
You're all both wrong and right.
Lorudar you're right that quantum computers are not Turing Complete.
However I didn't suggest quantum computers would ever play chess. I suggested that with advances they could be components of a more effective system.
Quantum computers could assist with the pattern recognition aspect of a massively parallel system by working in an 2^4 array to assess all possible configurations simultaneously.
Lorudar you're right that quantum computers are not Turing Complete.
However I didn't suggest quantum computers would ever play chess. I suggested that with advances they could be components of a more effective system.
Quantum computers could assist with the pattern recognition aspect of a massively parallel system by working in an 2^4 array to assess all possible configurations simultaneously.
Oh, sorry I misunderstood you Cythaera. But I still don't understand what are you suggesting with "pattern recognition" in quantum algorithm? And why only 2^4?
5 hrs ago
0
#12
Fet wrote:
#3 I can see you don't know anything about these kind of topics. You can run a chess engine on quantum chips. Imagine running a chess engine on thousands of quantum chips. There is already quantum chip in production.
Most people don't and you don't either but you can start here, if you like.
https://en.wikipedia.org/wiki/BQP
https://en.wikipedia.org/wiki/EXPTIME
Cythaera wrote:
You're all both wrong and right.
Lorudar you're right that quantum computers are not Turing Complete.
However I didn't suggest quantum computers would ever play chess. I suggested that with advances they could be components of a more effective system.
Quantum computers could assist with the pattern recognition aspect of a massively parallel system by working in an 2^4 array to assess all possible configurations simultaneously.
Lorudar you're right that quantum computers are not Turing Complete.
However I didn't suggest quantum computers would ever play chess. I suggested that with advances they could be components of a more effective system.
Quantum computers could assist with the pattern recognition aspect of a massively parallel system by working in an 2^4 array to assess all possible configurations simultaneously.
Lots of people mistakenly equate quantum computing with "massive parallelism", and ergo "pattern recognition" as well (probably an artifact about what used to be said about neural networks back when non-scientists were describing them). This is the fault of "science" articles written by people that do not understand what they are writing about.
GPUs are much faster than quantum computers for pattern recognition, and this is not likely to change anytime soon, because quantum hardware isn’t built for numerical optimization...it’s built for manipulating amplitudes in problems with very specific mathematical symmetries. Code-breaking, great, weather prediction, potentially great...evaluating discrete criteria in chess positions...not so great.
Parallelism is also limited. Quantum computers do encode superpositions of many possible states, but they do not evaluate all possibilities in parallel and read them out. They collapse into a single outcome, not a list of outcomes with valuations.
Where did the 2^4 array come from? Sixteen states is not going to make a dent in much of anything chess-related.
For the past years, we all know that Stockfish has been the uncontested best chess computer on Earth. Some have come close, like Torch or even Mittens, but Stockfish is still the engine that everybody relies on for game analysis.
I know there is absolutely no need to replace it, but has anybody else ever asked themselves the question of what will happen to Stockfish? Will it just get updated constantly until we get to like Stockfish 271 and beyond? Or will a separate team of coders try to build a new tool that could compete with Stockfish? What's your take on this?