3
Overheard a podcast guest explain AI bias with a dog analogy, clicked for me
I was listening to a tech podcast on my commute last Tuesday and this researcher compared training data bias to teaching a dog only to fetch red balls, then acting surprised when it ignores blue ones. That simple example finally made the whole concept of skewed datasets make sense to me, way more than the dense papers I've tried reading. I've been messing with my own little text generation projects and now I'm double checking what sources my training examples come from, because if the data only shows one kind of output, that's all you'll get. Has anyone else found a dumbed down analogy that actually changed how you approach building or testing AI stuff?
0 comments
Log in to join the discussion
Log In0 Comments
No comments yet
Be the first to share your thoughts on this discussion.