Singularity Institute

Marshall Brain talks about the rise of the second intelligent species and the future of humanity.

Marshall Brain (yes, that is his real name!) is the founder and CEO of HowStuffWorks, Inc. A published author with more than a dozen books to his name, including the popular The Teenager's Guide to the Real World, Marshall began HowStuffWorks.com as a hobby"€”a way to learn more about the world and to share that knowledge with online readers. Today, the company has expanded to include magazines, kids clubs, and book. Marshall lives just outside Raleigh, NC, with his wife Leigh, son David, and daughter Irena.

Singularity Institute: Will there be a Singularity?

Marshall Brain: I buy into the argument that we are in the process of creating the second intelligent species in the universe. This will be something that I'll be talking about tomorrow. As far as we know right now, there is one intelligent species in the universe. We have no evidence that there is more than one. That's all we know. We're creating the second intelligent species in the universe, which we predict will eventually be much smarter than us because of the ramp in computational power.

The idea that we can't predict what will happen when there are two intelligent species in the universe, I totally buy into that. The idea that that intelligence may far surpass anything that we can imagine in ourselves, the idea that in 2100, you can buy for $500 a million human intelligences in a box, or one intelligence that is a million times greater than human intelligence, and we have no idea what it is,in a box, and we can't predict what it's going to do, I totally buy that aspect of unpredictability in the Singularity.

SI: Is the Singularity near?

MB: If we just follow a straight Moore's Law curve, then in around 2020, you can buy a computer at your local retailer for $500 that has about a trillion operations per second. In 2040, you can buy about quadrillion operations per second for $500. We believe the human brain has something on the order of a quadrillion operations per second of processing power. So I think sometime in the 2030, 2040, 2050 timeframe, we're able to say, "Here's a computer for about $1000 that has computing power of the human brain." And by then, we've also come up with the algorithms that actually let us do something with all that computing power in an intelligent way. Then if you just follow that curve, in around 2100, say, you have a million human intelligences in a box, or in a speck, [laughs] however big it is, on your desk. It costs $500. We'll see if all that actually comes to pass. But I think around 2040, plus or minus ten years, you have human intelligence on a chip.

SI: Can we create machine consciousness?

MB: The human brain is a static object. It is there to be studied, and we will tease it apart until we understand every single neuron in it and how they all connect. Yes, we will reverse engineer it because it is static. That's one way to approach artificial intelligence.

Another way to approach artificial intelligence is to come up with a set of algorithms that simulate what is happening inside all the neurons without actually replicating all the neurons. Another way is to create a whole bunch of programs, where each one does things very well. The perfect example of that is, we know now how to play chess with a computer. It has nothing to do with the way human beings play chess, and yet it can beat us. Now at this point, even your desktop computer can beat just about every human being.

You can imagine creating an intelligence that is made up of a set of these really beautifully competent individual things. Like there's a module that knows how to walk in a straight line, and another that module that knows how to walk up stairs, and another module that knows how to drive a car, and another module that knows how to find a doorway, ring the doorbell, and go in. If you had a million modules that were supremely competent at doing all these individual things, and then you started hooking them together, it wouldn't be anything like human intelligence, yet it would be superior to human intelligence because all the components would be the best possible to do things.

We might create intelligence or consciousness or something, that is exactly simulating how we do it, or is sort of like how we do it, or has nothing to do with how we do it, or some combination of all three. The only question is, "How long does it take?" It was the same thing with flight. We couldn't fly, and we couldn't fly, and we couldn't fly, and we couldn't fly, and then one day, somebody got up in the air. Then you look at that event and just watch what happened once that first person got in the air. I flew here today in this big aluminum airplane for $200 from all the way across the country. It's insane!

SI: What breakthroughs do we need to achieve in order to have greater than human intelligence?

MB: There are things that we know we don't know. We don't know yet how to do vision at the level that a mouse or a dog or a human can do it. All three of those systems use very different amounts of computational power to do vision. We can't do any of those yet. We are getting better at language processing, understanding words as they come in, then understanding the meaning of those words as people put them together. But we're nowhere near a point where I could just talk to a computer like I'm talking to you, and it would understand my intent and my meaning.

Those are two big things that really hold robots back. If you had the ability to understand language and you can see, then you can replace the entire FedEx, UPS, post office kind of workforce. There are million people who drive the trucks for FedEx and UPS. If you have a robot that can see and understand language, those folks are all out of work because now you have a twenty-four by seven system that costs less than a human being, and off you go.

When does that happen? I have always thought that it would be very funny if vision was figured out in the year 2020, because "2020" has this relationship to vision. We talk about 20/20 vision. If it were to all get figured out around the year 2020, that would be very poetic and interesting. That might be off by five or ten years, but I think in that kind of timeframe, we're going to have the computational power to do vision. That's a big part of our brain. There's a whole module back here that is using... half a quadrillion operations per second? We don't have the power right now to do it, but by 2020 we will.

I would fully expect in that kind of timeframe that we've got the vision thing figured out. The language thing probably gets figured out around the same time. Even if you use Dragon Naturally Speaking or any of the dictation systems now, they're so much better than they were even five years ago that it's startling to use them. Even without training, they can understand most of what you say. They don't understand it - they recognize it, I guess is the right word. We have a little bit of a ways to got there, but we know Google is working on that. We know lots of people are working on how to understand human beings when they speak.

So is it ten years out? Is it twenty? I don't know the exact timeframe, but somewhere in that range, we're going to have language and vision figured out. With those two things figured out, we can do so much stuff that we can't do now, that the consciousness part of it almost becomes irrelevant. You don't need consciousness to do most of the jobs in our economy, you just need competence, the ability to execute.

SI: Will we really discard our bodies in the future?

MB: I think in the simplest sense, there are a set of problems that go with having bodies. One of them is death. I find the idea of death absolutely repulsive. The idea that, pick anybody... Winston Churchill, here's a person. I don't know how old he was when he died. Let's say he died at seventy. He accumulated seventy years of knowledge, thousands of relationships, an unbelievable network of people and ideas. When he died, we lost it all. That is repulsive. It is utterly ridiculous that a whole consciousness should just evaporate like that because you get a cancer or your heart stops beating or whatever.

So you have that problem with your body. The second problem with our bodies is that they're just so terrible as vehicles. I mean, they aren't bad, but if you don't feed your body, it dies; if it gets injured, you can't do things; we've got to use the restroom every two hours; and we sweat. These aren't the greatest vehicles ever designed to hold a consciousness. That's a problem.

The third thing is, we're creating these environments on computers, and just imagine you want to interact with a computer environment. Most people today do that today with two thumbs. They get a game controller and with their two thumbs: they walk, and they aim, and they shoot, and they look around with their thumbs, where our brains are designed to control a set of hundreds of muscles and get really rich sensory information on multiple channels. In a video game right now, you get visual and auditory, but you don't get anything on your touch sensors or on your taste sensors. You can't feel wind. You can't feel acceleration. There are all these other sensory things.

So you've got death, you've got the problems with the vehicle, and you've got the input and output channels of the brain that can't map into virtual environments. Those three things together will cause us, as soon as the technology is available, to want to take our brains out of our bodies, put them someplace safe, someplace that is optimally configured to keep the brain alive for as long as possible, and plug it into virtual environments fully. So that we can use all our muscles to control, and so that use all our sensors to read information, and actually be inside these virtual environments. As soon as that's possible, safe, and cheap, everybody will make the transition because we don't want to die. We just don't.

SI: Will machines surpass humanity in relevance?

MB: I personally think that there's going to be several different things happening simultaneously. One of them is intelligences far greater than human intelligence are developing in hardware and software. Artificial intelligence, whatever you want to call it. At the same time, computer environments will be getting richer and cooler and more interesting.

There's another part where human beings are becoming... more relevant, less relevant? Our relevance in the universe, because this other intelligent species is arising, changes. We don't know how it changes, but something happens.

I personally don't think that we will ever have our physical brains riding around in robot bodies. I'm not sure that we will even care about the real, physical universe as soon as we can plug in and spend all our time in artificial universes, because we can tailor the artificial universe. It will be richer and more interesting and more fun. It won't have all of the problems that go with the physical universe. There's no rain - or you can make it rain when you want, and make it stop when you don't. There's so many things about virtual universes that will be so much better than real universes that we just won't even bother with the real universe anymore.