Charles Isbell makes the case for more ethical AI
The Robot Brains Podcast
00:00
Is There a Better Result?
As machine learning researchers, we're were used to tuning a lot of hyper prameters behind the scenes. But then when we open source or release our code, the people who are going to use it are actually not going to retune all that. So we release something that's real good for the bench mark, but is not necessarily really good for a real world application. And so we fiddle with amma till it's point seven, or alpha with its point o one. It doesn't matter. We don't really know what those things mean, only that it worked for us. The better result is probably something like accuracy. But maybe that isn't the right better result you
Transcript
Play full episode
