[herculity275]

I used to think that but newer models' increasing coherence coupled with the astonishing success of RLHF makes me think that LLMs that can effectively intuit "reality" will emerge fairly soon.