Comment by perdomon

It seems like we're hitting a solid plateau of LLM performance with only slight changes each generation. The jumps between versions are getting smaller. When will the AI bubble pop?