> What used to take a dedicated research team and millions of dollars of compute (to play professional go), can now be done with a sub $1 gpt4 call (to do something like make strategy slides).
Is that an accurate statement, though? AFAIK, the Go-AI was hyper-specific to Go. I don’t think any LLM would approach the level of skill exhibited by those specialized AIs (much less provide valid moves for an entire game in my experience).
Is that an accurate statement, though? AFAIK, the Go-AI was hyper-specific to Go. I don’t think any LLM would approach the level of skill exhibited by those specialized AIs (much less provide valid moves for an entire game in my experience).