The Dual State of Artificial Intelligence

The first FDA trial of deep learning classifier running in production actually saved like 2000 lives a week while I was running. And then we had to shut it down and no one wanted to repay for it because the way we saved those lives was not monetizable. He got quite disillusioned with the medical, medical applications still very motivated and liked him a lot. But we need to have very, very deep domain expertise as well as AI expertise to make progress in that world.