 So this is Alan Turing. Everyone knows that Alan Turing was one of the people who founded computation and he first asked this question about whether machines could think like people. But what people don't know is that in that very same paper, Turing said that the real question was not whether computers could be as smart as the people in this room, it was whether computers could be as smart as a two-year-old. And he thought that the way to design intelligent computers was to design a child computer, to make them as smart as children, not necessarily as smart as adults. And at Berkeley, that's one of the things we've been trying to do is figure out how our children as smart as they are, what can they do and how could we use that kind of learning and intelligence in designing computers. And what that's taught us is that even the youngest babies are much smarter than the smartest computers. The big advances that we've made in computers in AI recently have depended on learning. And here's an example. Computers have learned how to do very sophisticated statistical learning from enormous samples of data. And as a result of that, all these genius brilliance and data, what they can do is say kitty cat when they see a picture of a cat, which is something that every one year old can do. Now, here's another thing that computers can do and babies can do. A more powerful way of learning than simply picking out statistics from the data is to develop a hypothesis and test it against the data. And we started to develop ways to get computers to do that too. And we've shown that even three and four year olds can think up an abstract hypothesis and then test it against the data. So those are things that computers can, where computers are sort of starting to approach the capacities of very young babies. But there are three other areas in which babies still far surpass computers. One of them is in creativity. So anyone who's seen a three year old knows that three year olds think up crazy ideas all the time. Crazy imaginary companions, crazy pretend play. And that's not just crazy. It's one thing to actually take a hypothesis and test it. It's another thing to think up a brand new hypothesis that nobody has thought of before. And what we've been showing in my lab is that not only are very young children very good at doing that, they're actually better at doing it than adults. So we did an experiment, for example, where we showed both four year olds and Berkeley undergraduates a machine that works in an unusual, strange way and just gave them data and let them try and figure it out. And the four year olds are consistently better than the Berkeley undergraduates. This is probably not good if I'm trying to promote Berkeley here, but the same thing is true for MIT and Harvard undergraduates. So the children are actually better at finding a new solution, doing something out of the box, doing something unusual, something creative, even than regular adults, let alone computers. And we can even see this when we look at the design of the baby brain. So you can see that between zero and five, we're making many, many, many more new connections than we do for the whole rest of our life. And what we do as adults is to just take the connections that work well and keep them up. So it's that early period in which there's this burst of creativity that we don't see, we're not even approximating in computers. Here's another thing that babies do much better than computers. And that's to explore. So one of the things that we know is that things like experimentation are one of the most important things we can do in science. Babies are doing incredible experimentation. We call it getting into everything. And one of the things that my colleagues and I have shown is that when babies are getting into everything, when they're playing, say, looking at a toy like this, they're actually doing very systematic moves not only to be able to interpret data, but to go out and get exactly the data that's the right kind of data to solve a problem and to think of a new solution. A third thing that babies can do that computers can't is learn from other people. And that's one of the most distinctive things about us as human beings. So we know that even newborn babies already imitate the expressions that they see on the faces of others. So newborn babies are already designed to be learning from other people. And by the time babies are nine months old, they're not only learning how to stick out their tongues, they can do something like learn how to make a machine go just by watching what another person does. And it isn't just that they can learn these things by watching, they can assess how good a teacher the other person is. So they'll learn differently from someone who they think is an expert and knows what they're doing. Compared to someone who's just a blowhard who talks a big game, but it isn't actually doesn't actually know what they're doing. They can assess how ignorant someone is, how knowledgeable they are, whether they're trustworthy or whether they're not trustworthy. And again, that kind of social learning, this is another experiment we've been doing in my lab, is something that computers aren't really even in the ballpark of doing. And as Ken said, it's that kind of ability to interact with other people. That's the really distinctive human thing. So maybe sometime in the future, computers will be able to be smart. But at the moment, they're not even as smart as a two-year-old.