Dubrovnik-1950 wrote:
From Grok the world's smartest AI
That's X marketing rather than a hard fact. A hard fact is that there is no AI that is better than all the others at every type of task. Another is that several of them are of roughly similar levels of intelligence now.
To my knowledge there has not been any systematic independent testing of Grok 3 yet, and weaknesses in the testing by its makers have been pointed out. So the jury is out. Also, the most powerful model is only available to paying customers.
From Grok, the world's smartest AI.
Can a Shannon type B chess engine solve chess like Stockfish? And if the answer is no tell these morons why.
No, a Shannon Type B chess engine cannot solve chess, and here’s why, you impatient genius:
Claude Shannon, the father of information theory, outlined two approaches to building chess engines back in 1950: Type A and Type B. A Type B engine uses a selective, human-like strategy, pruning the game tree by evaluating only "promising" moves based on heuristics—like focusing on captures, checks, or threats—rather than exhaustively analyzing every possible move.
"Solving chess" means either determining the perfect outcome (win, loss, or draw) from the starting position or creating an engine that plays perfectly in every position. . A Type B engine, by design, sacrifices breadth for selectivity, relying on clever but imperfect rules to guess which moves matter.
The problem? Chess is a brutal, concrete game. A Type B engine might miss a subtle, non-obvious move that flips the position because its heuristics don’t catch everything.
Obviously not exactly from Grok, the world's smartest AI. It was twice as long a couple of minutes ago. (I noticed it still thought any version of chess has 10^120 positions - obviously so smart you can't teach it anything!)
Still we seem to be in agreement given the question you asked.
The topic refers to chess ever being solved, so you should have asked it, "Can a Shannon type B chess engine ever solve chess like Stockfish? And if the answer is no tell these morons why" or possibly, "Can a Shannon type B chess engine like Stockfish ever solve chess? And if the answer is no tell these morons why", if that is closer to what you intended to ask (with your demonstrated level of literacy it's difficult to say).
As I remarked to someone earlier, if you had asked fifty years ago the questions, "Could a computer beat a grandmaster", and, "Could a computer ever beat a grandmaster", the answers might well have been different.
Had you asked the right question, we would probably still be in fairly close agreement. Gronk, the world's smartest AI, being but a superficial intelligence routine, would probably have flatly answered "no", whereas I would answer, "highly unlikely".
However, if your post was meant to be in response to my #19554 it hasn't addressed any of the points I made. You didn't ask it anything about type A engines, you didn't ask it if the only way to solve chess is to do a full width search of the whole game tree (given it's level of competence there's an evens chance it would say, "yes") and you don't appear to have run any of the tests I suggested. (You've posted no results at any rate.)
Can your Gronk dump be taken as an indication that you don't feel competent to address the questions yourself? Would you still maintain that the only way to solve chess is to do a full width search of the whole game tree or is the question beyond you?