From a discussion on machine learning:

"For any program, there are two possibilities. Either it works exactly as intended, of some bug crops up in the code. In the first case, we probably have nothing to fear. The latter case isn't terribly concerning, either."

WRONG! When it comes to something as powerful as AI, both of these scenarios should be deeply terrifying. For the first, it takes all the usual problems of trust, accountability, and intention, and magnifies them a hundred thousand fold. For the second, it introduces a possibly apocalyptic uncertain consequences. What will it do? What will happen? Fuck if we know. :/

@Angle neural networks are already sufficiently complicated that in some cases we're not entirely sure how they've reached the results they did. And this is only the early stages of the field - the complexity is only going to grow.

It's time to seriously rethink our debugging tools.

@polychrome Maybe we should use machine learning for our debugging tools. What could go wrong with that? :V