
Intelligence Requires More Than Following Instructions
Post-training improves the accuracy and usefulness of LLMs but does not make them intelligent in any meaningful sense — as the Monty Hall problem showsThe danger is not that computers are smarter than us but that we think they are and thus trust them to make decisions they should not be trusted to make.
Read More ›
