[neilv]

...and plagiarize like crazy, while lying about it. :)

[dbrgn]

Could it be plausible that ChatGTP processed some text describing this exact game, where it is claimed that the idea is novel (because it was, at that time)? Since ChatGPT does not understand the concept of novelty, it would simply "learn" that the rules are novel, and then repeat the rules, still claiming that they're novel. After all, that's the information it was trained with :)

[ChatGTP]

...maybe this is the key to success and ChatGPT is here to show us the way! :)

Ride those coattails and take other peoples ideas as your own!

[sebzim4500]

Is this actually in the training data though? I couldn't find a textual description of the rules though Google, so I'm not convinced.

[sekai]

Is it plagiarizing if ChatGPT knows about chess rules, analyzes every single chess strategy, then create a similar game?

[yowzadave]

ChatGPT does not know chess rules, nor can it “analyze” a chess strategy. ChatGPT has digested the conversations of many many people who have talked about chess, and can reproduce a transcript that sounds similar to this corpus of conversations.

But it is not synthesizing an understanding of the game of chess.

[skeaker]

Except that there is demonstrable evidence that indicates GPT actually does have some level of understanding via internalized world models (of Othello in this case, not chess, but the idea is the same): https://thegradient.pub/othello/