DeepMind researcher argues LLMs lack the mechanism to spark scientific revolutions — world models might

A Google DeepMind position paper argues language models can't originate scientific revolutions the way world models potentially could.