AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Language Models Infer Causation From Correlation
The study basically found that LMS are not great at inferring causation. They're much better at correlations, which kind of makes sense. And I think that this actually hints that in fact, even causal reasoning may may be overcome through scale. It's possible that just as you scale these systems up more, you know, through all these objections, scale will win through nonetheless.