Building AI Data Responsibly with Edouard d’Archimbaud - ML 081
Adventures in Machine Learning
00:00
Is Scaling a Good Idea?
When labeling for computer vision, it was about labeling from 10,000 to maybe 1 million images. In huge case, I think it's becoming less and less to use such large data sets. And this is a super good news, because the data is cut for AI. The other thing is that you can use less examples, but you need to have very high quality examples. That one example will have more impact.
Transcript
Play full episode
