A fascinating new paper from scientists at the AI research nonprofit LAION finds that even the most sophisticated large language models (LLMs) are frequently stumped by the same simple logic ...