Q What, what's, uh, because You know, we're very far from, uh, so what's your, I guess, uh, appreciation of the state of the market?
A Yeah, so I think there are some good reasons to be excited about. Um, many of the problems that deep learning is able to solve, uh, solve in the sense that Have reasonable solutions. You know, none of these problems are solved. So they are not, you know, vision is not a solved problem. Language understanding is not a solved problem, but they are good enough to be used in applications. And, uh, that's a position that we were not in before, before the deep learning revolution. Um, the, um, accuracy of speech recognition and accuracy of image classification has, and, and accuracy of translation has gone up Quite a lot compared to what the state of the art was before, ah, 2012. So there are reasons for excitement, but people also got overexcited about some of the results that we were getting from game playing, ah, you know. So all the excitement about deep reinforcement learning was because you could play games with very little knowledge about the game itself. You know, you just give it pixels and, oh, it learns to play the game. But those excitements Turned, turned out to be premature because, um, it's, yeah, game playing is, um, an easier problem compared to real world problems. Those, those techniques do not translate to real world problems. Um, so this applies, and AI goes through this game playing phase multiple times, you know. Games have been the driving force for many AI, u…
AI assessment note: “So there are reasons for excitement, but people also got overexcited about some of the results”