AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Can Language Models Reason Causally?
Language models learn what causes what, call it say common sense, causal knowledge. It's learning from kind of statistical regularities and statements of causal relationships around training data. That doesn't necessarily say that it's doing causal reasoning, just I could have just found a diag somewhere, to remember the diag to some degree. So my personal prediction for 2023 is that we're going to have to get a lot more precise about what we mean by reasoning.