Institute For The Future

The Centaur Revolution

Your future will depend on how well you work with AIs

The common concept of intelligence is linear. It moves from a mouse to a monkey to a dumb person to a smart person, like a sound that grows louder as you move along the dimension. This is completely wrong.

Intelligence is a symphony of different cognitive instruments, with each instrument producing a different kind of sound, a different type of thinking. The result is a mixture of ways of thinking that produces a very complex thing that we call intelligence. Different people have slightly different mixtures.

One of the misconceptions we have about ourselves is the belief that we have a general-purpose intelligence. We don’t. It’s a specific mixture of different kinds of intelligences—deductive reasoning, symbolic reasoning, recall, emotional intelligence, and many others we don’t know anything about yet—that’s evolved over billions of years for our survival.

Animals have a similar kind of aggregation, which has evolved for their unique survival needs. In many cases they have some of the same instruments we have. In other cases some of their instruments are even louder, superior to ours. Squirrels have amazing spatial memory to recall where they buried nuts years ago, exceeding our ability.

Your Calculator Is Smarter Than You Are

When we make Artificial Intelligences, we’re engineering them to accentuate certain kinds of thinking. Right now they’re simple, with just a couple of kinds of thinking. But in some ways, they’re superior to our abilities. Your calculator is smarter than you are in arithmetic. Your GPS is smarter than you are in spatial navigation. Any search engine is much smarter than you are in recall. These are very, very narrow AIs, but the important thing is that they don’t think like humans, and we don’t need or even want them to. The reason we’re creating self-driving cars using AI is precisely because we don’t want it to think like a human, or drive like a human. It’s not worried about whether it left the stove on as it goes down to the street. It’s not worried about whether or not it should have majored in finance. It’s just driving. It’s been engineered in a very specific way.

The whole point of AI is that it doesn’t think like us. Evolution has taken biological life only so far in making different kinds of minds. We’re going to use technology to extend and fill the space of possible ways of thinking. And, as you know, in a global economy thinking different is the primary way to generate wealth.

First Came the Power Grid. Here Comes the AI Grid

The prosperity we have right now is based on artificial power. We use machines that run on fossil fuels to make things like skyscrapers, dams, roads, factories—things we couldn’t make with our own muscle power and animals, at least not at the same scale, speed, or quality. Today, when you drive your car, you’re employing 250 horsepower, which you can turn on or off with a switch very cheaply. That’s the power of artificial power.

We distribute artificial power in a great electrical grid. Anybody can buy power. You just plug in, and you do what you want with it. It’s a source of great innovation. For example, 150 years ago a farmer looked at his hand-pump and said, “I can add electricity to this, and make a powered pump.” Do that many, many times and you get the industrial revolution.

Now we’re at the second phase—we’re adding minds to things. We’ll take the electric pump and add AI. Now we have a smart pump. We’re sending out minds as a service, as a cheap commodity on the grid, as an AI utility that anybody can purchase. You can buy as much as you want, and add the resource of artificial minds to whatever you have. The car still has 250 horsepower, but now we’re going to add 250 minds on top of that and call it the self-driving car.

My formula for the next 10,000 startups is take X, add AI. Find something—the more unusual, the more unexpected, the more counterintuitive, the better. The AI is cheap. It’s the interface you’re adding to AI that makes it valuable. It’s the branding you add, it’s the story. It’s like trying to sell water. You have to add something extra to it.

Today, Google offers a service that lets you ask it questions about the content of images. You ask it “What color is the ball? What does the girl have in her hand?” etc., and it gives you answers in a conversational way. Google sells this AI at 6 cents per 100 hits, and you can add it to whatever you want. Say that you want to sell the kinds of things that appear on TV shows, like a dress. You can use this kind of engine to find that specific dress and then you can bring it to your audience and say, “You can buy it here.” That’s the kind of thing you can do right now.

Productivity Is for Robots. Inefficiency Is for Humans

AI will go through three stages. The first is thinking of it as being alien, separate from what we do. The second is using AI as a utility, like electricity. The third is taking AIs and putting them into bodies that we call robots.

Most jobs are a combination of different tasks. Any part of a job—physical or mental—that can be measured by the criteria of efficiency or productivity, will be done by bots. A lot of the jobs humans do aren’t going to go away. They’re going to be redefined and altered by the fact that we’re going to work with AIs.

But if productivity goes to the bots, what’s left for us? Inefficiency. What humans do well is waste time. Take science, which is, by definition, inherently inefficient because it requires a lot of time generating hypotheses and designing experiments that usually don’t work. But that’s the only way you learn anything. Or take innovation, which is inherently inefficient because you’re initially making things that don’t work. Exploration is inherently inefficient. Art is inherently inefficient. Human interactions are inherently inefficient, but we’re good at them. We’re attracted to roles and tasks that aren’t very productive or can’t be measured in terms of productivity.

The best chess player in the world today is not an AI. It’s not a human. It’s a human plus AI, which they call a centaur. The best medical diagnostician in the world is not Watson Health, the medical AI. It’s not a human doctor. It’s doctors plus AI. Centaurs work because they’re a complimentary team. We’re going to see more of that, where the AI does things that can be measured in productivity and we do things that are measured in exploration, interaction, and experience.

There may be scientific problems that we have today—quantum gravity, dark energy—that we won’t be able to solve with our intelligence alone, so we’ll invent different kinds of thinking that we’re incapable of. Together we’ll solve really difficult problems, and you’re going to be paid by how well you work with AIs.

We’re at the Beginning of the Beginning of the Beginning of the Internet

Even though AI is probably the most important thing going on right now, if we look 25 years into the future, we probably see AI as the most important invention. So what is the next big thing? Whatever it is simply hasn’t been invented. But it will likely be enabled by AI. We’re still at the beginning of the beginning of the beginning. Twenty-five years from now, people will look back and say, “You didn’t have the Internet. You thought you had the Internet, but you didn’t really have it yet. If only I could have been alive back then, before all the things that we have now. You could just take X and add AI. That’s all you had to do!”

Right now is the best time in the world to start things because it’s just the beginning. That means you’re not late.