"Probabilistic generative models are fun but no amount of probabilistic sequence generation can be a substitute for logical reasoning."

Someone at Bell Labs wrote a program back in the early days of UNIX that generated research papers. AFAICT no one took it seriously, even if it was good enough to fool someone

Bonus points for anyone that finds the reference. It might have been Kernighan

For the money to keep pouring into "AI companies", LLMs must be taken seriously

More than just autocomplete

Autocomplete is useful, no doubt, but at the energy cost and water usage of "AI" maybe not worth it