I can accept calling LLMs “AI” because machine learning is a branch of AI and they are also able to figure things out, so especially when combined with a capable harness, they legitimately are artificially intelligent. However, LLMs aren’t “super intelligence” that vastly surpasses human capability, and instead are basically nifty tools for automating grunt work. Given they seem to be leveling off and we are seeing more incremental improvements instead of exponential gains at this point, “super intelligence” is unlikely to be an apt term for a long time to come, if ever for LLMs.
I can accept calling LLMs “AI” because machine learning is a branch of AI and they are also able to figure things out, so especially when combined with a capable harness, they legitimately are artificially intelligent. However, LLMs aren’t “super intelligence” that vastly surpasses human capability, and instead are basically nifty tools for automating grunt work. Given they seem to be leveling off and we are seeing more incremental improvements instead of exponential gains at this point, “super intelligence” is unlikely to be an apt term for a long time to come, if ever for LLMs.