AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
How Do You Know What Your Model Is About?
People are starting to be affected by the problems that we were foreseeing, you know, a few years ago. Theire sort of turning to see, well, what can be done? What are we thinking? And it turns out that that one of the key things to be able to do is to document. Cause as you document, you can trace where issues are coming from and address them. One in the take layes from that paper was that lang models are too big when the data they're using is not documentable. It's because you'll start having really harmful behaviour from language models that you don't know where it came from.