Interesting Concept

[sojumasterGA]

This is an interesting exercise in thought.
The problem with LLMs is that they are not chess (or game) engines. LLMs are basically Google on steroids by using natural language.
I have submitted manual prompts with ChatGPT in regards to chess problems and it is funny seeing that it tries to come up with a solution even though it contains illegal moves. The reason that ChatGPT is able to "solve" some problems is because a lot of them are already posted on the internet with solutions.

Try some more outragous problems like below and it painfully/ridiculous it is watching them try to solve it.
8/1p2k2P/5NK1/8/8/8/8/8 w - - 0 1 "This is a serieshelpstalemate. Black plays 9 consecutive moves, without reply from White, after which White stalemates in one! Both players cooperate."