None of the Large Language Models, or LLMs, can play chess. They get easily confused by a chessboard.

However, GPT-4 had the smarts (with a little bit of handholding on my part) to build code that can properly draw, and properly update, a chessboard, represented using FEN (Forsyth-Edwards Notation).

Not very useful, I know, but it shows the impressive coding capabilities of GPT-4.

Here is the result, in its full glory:

The current string:  


Moves so far