FAIR :: Facebook Artificial Intelligence Research
Muhammad Anser

Thanks for the informative post, but here’s a quibble.

“All of information which isn’t fundamental to noting your question is simply noise, and all the more essentially, a waste of resources to store and analyze.”

This strikes me as a very dangerous mindset. To put it differently, suppose we want to know if X, then Y.

P(Y|X) = very high, but not 1

P(Y|X + something we don’t quite know) = 0

P(something we don’ quite know) = very small.

So “something we don’t quite know” is just noise because we just want Y and X almost always gives us Y, since “something we don’t quite know” is rare? This seems to run counter to the idea of science via statistics: as Fermi put it, “When your findings confirm your hypothesis, you have made a measurement. When they contradict, you have made a discovery.” So we do not want to make discoveries? Every hypothesis is wrong, at least some of the time, and knowing how, when, and where our predictions go awry is how we learn about the limits of what we do and what we do not understand. I’ve always gotten the impression that this is dangerously underappreciated nowadays and your line above made me wonder aloud.

Like what you read? Give Henry Kim a round of applause.

From a quick cheer to a standing ovation, clap to show how much you enjoyed this story.