That was pretty much always the only potential path forward for LLM type AIs. It’s an extension of the same machine learning technology we’ve been building up since the 50s.

Everyone trying to approximate an AGI with it has been wasting their time and money.