He (and I) might have been wrong on how useful the LLMs could get, but time isn't proving him wrong on his claims that the usefulness doesn't justify the investments around LLMs