The Role of Data Labelling in Modeling

In order to make n l p models work, you need tens of thousands of examples. But nowadays, we have these giant transformer models that are pre trained on all this internet text. And so they can get a pretty good model with like, a relatively few examples. This requires having access to a very large, unlabelled data set which happens to be the same distributions as whatever documents your model is not doing well one.

Transcript

Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.