AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Different Mechanisms by Which Deep Learning Can Make Good Predictions
It's not hard to find examples where you do get good predictions based on things that you don't actually want so I'll give you a simple example there's a well-known machine learning data set called ImageNet which was very important in the development of machine vision. But now if you want to look a bit closer and say okay how is the system making such good predictions, it may be using correlations in the data that are not really what you want it to be doing. For example the difference between um the difference between a dog and a husky I kind of have these are the specific There's something similar like the end of the day if not this example what's the difference?