> I wonder what this says about AI in the future, since you can imagine that as AI becomes more powerful, it starts to make inferences that aren't explicit, and

> I wonder what this says about AI in the future, since you can imagine that as AI becomes more powerful, it starts to make inferences that aren't explicit, and discovers connections that aren't obvious. That's why we should invest more in AI interpretability. I heard some famous researchers saying that it's not important, but for these reasons it actually is.