Ben Sorscher: Data Pruning for Efficient Machine Learning
Eye On A.I.
00:00
The Importance of Deep Learning in Decision Making
The data we feed into these models itself is biased independent of whether we bake any bias into the models ourselves just by collecting examples at random from the web. We're presenting our own biases to these deep models and so they can pick up on prejudices and things like that. When we use these models to make decisions those prejudices which arise from bias in the data are baked into the outputs. Large language models aren't really being used to make decisions right now are they? i think if not our theory or our algorithm something like it probably will change the way we train these large language models in the future. I'm trying to take some of the lessons we learned from these deep neural networks and look for
Transcript
Play full episode
