Subtitles and Transcript
0:13 I'm going to talk a little bit about where technology's going. And often technology comes to us, we're surprised by what it brings. But there's actually a large aspect of technology that's much more predictable, and that's because technological systems of all sorts have leanings, they have urgencies, they have tendencies. And those tendencies are derived from the very nature of the physics, chemistry of wires and switches and electrons, and they will make reoccurring patterns again and again. And so those patterns produce these tendencies, these leanings.
0:53 You can almost think of it as sort of like gravity. Imagine raindrops falling into a valley. The actual path of a raindrop as it goes down the valley is unpredictable. We cannot see where it's going, but the general direction is very inevitable: it's downward. And so these baked-in tendencies and urgencies in technological systems give us a sense of where things are going at the large form. So in a large sense, I would say that telephones were inevitable, but the iPhone was not. The Internet was inevitable, but Twitter was not.
1:32 So we have many ongoing tendencies right now, and I think one of the chief among them is this tendency to make things smarter and smarter. I call it cognifying — cognification — also known as artificial intelligence, or AI. And I think that's going to be one of the most influential developments and trends and directions and drives in our society in the next 20 years.
1:59 So, of course, it's already here. We already have AI, and often it works in the background, in the back offices of hospitals, where it's used to diagnose X-rays better than a human doctor. It's in legal offices, where it's used to go through legal evidence better than a human paralawyer. It's used to fly the plane that you came here with. Human pilots only flew it seven to eight minutes, the rest of the time the AI was driving. And of course, in Netflix and Amazon, it's in the background, making those recommendations. That's what we have today.
2:33 And we have an example, of course, in a more front-facing aspect of it, with the win of the AlphaGo, who beat the world's greatest Go champion. But it's more than that. If you play a video game, you're playing against an AI. But recently, Google taught their AI to actually learn how to play video games. Again, teaching video games was already done, but learning how to play a video game is another step. That's artificial smartness. What we're doing is taking this artificial smartness and we're making it smarter and smarter.
3:17 There are three aspects to this general trend that I think are underappreciated; I think we would understand AI a lot better if we understood these three things. I think these things also would help us embrace AI, because it's only by embracing it that we actually can steer it. We can actually steer the specifics by embracing the larger trend.
3:38 So let me talk about those three different aspects. The first one is: our own intelligence has a very poor understanding of what intelligence is. We tend to think of intelligence as a single dimension, that it's kind of like a note that gets louder and louder. It starts like with IQ measurement. It starts with maybe a simple low IQ in a rat or mouse, and maybe there's more in a chimpanzee, and then maybe there's more in a stupid person, and then maybe an average person like myself, and then maybe a genius. And this single IQ intelligence is getting greater and greater. That's completely wrong. That's not what intelligence is — not what human intelligence is, anyway. It's much more like a symphony of different notes, and each of these notes is played on a different instrument of cognition.
4:26 There are many types of intelligences in our own minds. We have deductive reasoning, we have emotional intelligence, we have spatial intelligence; we have maybe 100 different types that are all grouped together, and they vary in different strengths with different people. And of course, if we go to animals, they also have another basket — another symphony of different kinds of intelligences, and sometimes those same instruments are the same that we have. They can think in the same way, but they may have a different arrangement, and maybe they're higher in some cases than humans, like long-term memory in a squirrel is actually phenomenal, so it can remember where it buried its nuts. But in other cases they may be lower.
5:09 When we go to make machines, we're going to engineer them in the same way, where we'll make some of those types of smartness much greater than ours, and many of them won't be anywhere near ours, because they're not needed. So we're going to take these things, these artificial clusters, and we'll be adding more varieties of artificial cognition to our AIs. We're going to make them very, very specific.
5:37 So your calculator is smarter than you are in arithmetic already; your GPS is smarter than you are in spatial navigation; Google, Bing, are smarter than you are in long-term memory. And we're going to take, again, these kinds of different types of thinking and we'll put them into, like, a car. The reason why we want to put them in a car so the car drives, is because it's not driving like a human. It's not thinking like us. That's the whole feature of it. It's not being distracted, it's not worrying about whether it left the stove on, or whether it should have majored in finance. It's just driving.
6:17 Just driving, OK? And we actually might even come to advertise these as "consciousness-free." They're without consciousness, they're not concerned about those things, they're not distracted.
6:28 So in general, what we're trying to do is make as many different types of thinking as we can. We're going to populate the space of all the different possible types, or species, of thinking. And there actually may be some problems that are so difficult in business and science that our own type of human thinking may not be able to solve them alone. We may need a two-step program, which is to invent new kinds of thinking that we can work alongside of to solve these really large problems, say, like dark energy or quantum gravity.
7:07 What we're doing is making alien intelligences. You might even think of this as, sort of, artificial aliens in some senses. And they're going to help us think different, because thinking different is the engine of creation and wealth and new economy.
7:24 The second aspect of this is that we are going to use AI to basically make a second Industrial Revolution. The first Industrial Revolution was based on the fact that we invented something I would call artificial power. Previous to that, during the Agricultural Revolution, everything that was made had to be made with human muscle or animal power. That was the only way to get anything done. The great innovation during the Industrial Revolution was, we harnessed steam power, fossil fuels, to make this artificial power that we could use to do anything we wanted to do. So today when you drive down the highway, you are, with a flick of the switch, commanding 250 horses — 250 horsepower — which we can use to build skyscrapers, to build cities, to build roads, to make factories that would churn out lines of chairs or refrigerators way beyond our own power. And that artificial power can also be distributed on wires on a grid to every home, factory, farmstead, and anybody could buy that artificial power, just by plugging something in.
8:38 So this was a source of innovation as well, because a farmer could take a manual hand pump, and they could add this artificial power, this electricity, and he'd have an electric pump. And you multiply that by thousands or tens of thousands of times, and that formula was what brought us the Industrial Revolution. All the things that we see, all this progress that we now enjoy, has come from the fact that we've done that.
9:01 We're going to do the same thing now with AI. We're going to distribute that on a grid, and now you can take that electric pump. You can add some artificial intelligence, and now you have a smart pump. And that, multiplied by a million times, is going to be this second Industrial Revolution. So now the car is going down the highway, it's 250 horsepower, but in addition, it's 250 minds. That's the auto-driven car. It's like a new commodity; it's a new utility. The AI is going to flow across the grid — the cloud — in the same way electricity did.
9:33 So everything that we had electrified, we're now going to cognify. And I would suggest, then, that the formula for the next 10,000 start-ups is very, very simple, which is to take x and add AI. That is the formula, that's what we're going to be doing. And that is the way in which we're going to make this second Industrial Revolution. And by the way — right now, this minute, you can log on to Google and you can purchase AI for six cents, 100 hits. That's available right now.
10:05 So the third aspect of this is that when we take this AI and embody it, we get robots. And robots are going to be bots, they're going to be doing many of the tasks that we have already done. A job is just a bunch of tasks, so they're going to redefine our jobs because they're going to do some of those tasks. But they're also going to create whole new categories, a whole new slew of tasks that we didn't know we wanted to do before. They're going to actually engender new kinds of jobs, new kinds of tasks that we want done, just as automation made up a whole bunch of new things that we didn't know we needed before, and now we can't live without them. So they're going to produce even more jobs than they take away, but it's important that a lot of the tasks that we're going to give them are tasks that can be defined in terms of efficiency or productivity. If you can specify a task, either manual or conceptual, that can be specified in terms of efficiency or productivity, that goes to the bots. Productivity is for robots. What we're really good at is basically wasting time.
11:16 We're really good at things that are inefficient. Science is inherently inefficient. It runs on that fact that you have one failure after another. It runs on the fact that you make tests and experiments that don't work, otherwise you're not learning. It runs on the fact that there is not a lot of efficiency in it. Innovation by definition is inefficient, because you make prototypes, because you try stuff that fails, that doesn't work. Exploration is inherently inefficiency. Art is not efficient. Human relationships are not efficient. These are all the kinds of things we're going to gravitate to, because they're not efficient. Efficiency is for robots. We're also going to learn that we're going to work with these AIs because they think differently than us.
12:01 When Deep Blue beat the world's best chess champion, people thought it was the end of chess. But actually, it turns out that today, the best chess champion in the world is not an AI. And it's not a human. It's the team of a human and an AI. The best medical diagnostician is not a doctor, it's not an AI, it's the team. We're going to be working with these AIs, and I think you'll be paid in the future by how well you work with these bots. So that's the third thing, is that they're different, they're utility and they are going to be something we work with rather than against. We're working with these rather than against them.
12:41 So, the future: Where does that take us? I think that 25 years from now, they'll look back and look at our understanding of AI and say, "You didn't have AI. In fact, you didn't even have the Internet yet, compared to what we're going to have 25 years from now." There are no AI experts right now. There's a lot of money going to it, there are billions of dollars being spent on it; it's a huge business, but there are no experts, compared to what we'll know 20 years from now. So we are just at the beginning of the beginning, we're in the first hour of all this. We're in the first hour of the Internet. We're in the first hour of what's coming. The most popular AI product in 20 years from now, that everybody uses, has not been invented yet. That means that you're not late.
13:34 Thank you.