Are we committing a Feynman error today, with the way we're developing large language models or other forms Artificial Intelligence?