The emotional machine

What are you hoping for with her?

I'm not exactly aiming high. One thing that is really important for intelligence is learning. If a system doesn't learn, it can't be considered intelligent because its behavior has been programmed into it. When you think about it, most people think that artificial intelligence just kind of arrives fully formed. But even Einstein took five years to tie his shoelaces. If we're going to make truly intelligent machines, they have to start as babies, knowing nothing and learning where their body starts and ends and who Mommy is and what being a naughty girl is about. All those things that children learn in their first year that we pay almost no attention to. I'm aiming to get something like a 4-month-old baby, which doesn't sound very impressive but it's amazing what a 4-month-old baby knows.

How long have you been working on her?

Two years.

Is anyone else working on similar projects?

Vaguely similar, yes. There are plenty of people building robots, but I'm not aware of anyone trying to do it the way I'm doing it. The theories I'm trying to test out are new. This approach to robotics is certainly unusual. But there are tens of thousands of people interested in how the brain works and how to make machines more like it.

Can you conceive of any way that this could go wrong? The whole "if an amiable, intelligent machine fell into the hands of an evil person" idea? In the book, you mention how we might use this for military purposes.

Bad people can do bad things with anything. That's not a reason to stop doing it. There are military applications for spoons. We'll have to worry ourselves about the ethics of making spoons. It's a separate issue whether they should be used by the military. A few months ago I was on the radio, on the BBC, talking about this subject with somebody who was arguing that we shouldn't do AI because it has military applications. He said, "One day there will be intelligent military machines." And I said, "We already have intelligent military machines. They're called pilots." I regretted saying that because that day happened to be Sept. 11. But it is true: we already have intelligent machines and they're people. If we make intelligent machines in the future, they'll only be variations on that. Another life form. We already share the planet with 10 million other life forms and don't think twice about it. People might argue that these life forms will have super-titanium bodies and turn into these huge and unstoppable machines. But that's going to be true for human beings as well. The more time goes on, the more like machines we get and the more integrated with technology we become.

Are you saying that intelligent machines would be indistinguishable from human beings?

Indistinguishable from other living things, yes. No different from us than a leopard or an elephant is different from us. If you look at the variety of living things on the planet right now -- a cherry tree or a mushroom or bacterium -- there's this huge repertoire of living things. Large numbers of those things we've managed to kill off. Maybe one day we'll create a few replacements.

You're an atheist, but you claim that you're on a spiritual journey. What does this all say about the soul?

I don't know the answer to that yet. Again, it's the problem of thinking of the soul as some kind of stuff. The way we view the world is wrong. We haven't seen the right way of looking at the universe yet and when we do, then some of these spiritual things will start to make more sense.

You say that someone passes away and you call up their memory in your brain -- you simulate them -- is that in some way their soul?

I don't know. There could be some truth in it. If that is their soul, then there's not one soul per person. You might have many people who think you up. You're in many places at once. That's one thing we've learned about organization and information -- you can copy it. We like to think that we're one person when we're alive, but even that's not true. Our brains are made out of 100 billion working parts, none of which are in charge, none of which understand the existence of the other 100 billion parts. You can do bizarre things like cut someone's brain in half -- people who have severe epilepsy have the corpus callosum cut in between the two halves. In a way, they turn into two people then, each having control of half a body, often with different tastes in clothes.

Is the idea of the soul what's interested you from the beginning?

The way I look at it -- and I borrow this idea from Richard Dawkins -- the universe has been around for 15 billion years, during which time I haven't existed. According to astrophysicists we have another 15 billion years to go, during almost all of which time, I won't exist. Then there's this tiny flicker in the middle when I find myself here. I want to know what that means. Where did I come from? Where am I going? What am I doing here? And quickly, I haven't got long.

• View blog reactions