Rise of the Machines with Matt Ginsberg (Re-release)
StarTalk Radio
00:00
Why Machine Learning Is Hard
If we use that model you've just described it talks to me about history now from a player's point of view. You can look back historically and see LeBron James scored way more points than average five years ago, yes that's history but it certainly bears on today. There is so much data in something that happens with machine learning when you have too much data programs start what's called overfitting which means they say wow whether or not the Lakers one last year was correlated with the phase of the moon oh dear that’s just luck. If you have so much data some of it's going to look relevant when it really isn't.
Transcript
Play full episode
