Despite its Impressive Output, Generative AI Doesn’t Have a Coherent Understanding of the World
Researchers show that even the best-performing large language models don’t form a true model of the world and its rules, and can thus fail unexpectedly on similar tasks.