麻豆传媒

You believe the philosopher led us astray. In what way?

He鈥檚 the one who argued that the mind couldn鈥檛 be just the brain, couldn鈥檛 be just the body, that it had to be a different kind of substance altogether. It鈥檚 one of the greatest mistakes in the history of thinking. And we have been trying to undo that mistake ever since. Certainly, I have.

Why has it been so hard to overturn this picture of the mind?

There鈥檚 a very powerful and ultimately emotional resistance to the idea that our brains are organs for thinking, and that our experience of mind is simply the workings out of some neural machinery. Darwin showed us how all the wonderful 鈥渋ntelligent design鈥� of the universe, and that includes our brains and minds, comes out of a non-intelligent process: evolution. Some people have a lot of trouble with that. Then Alan Turing came along.

What did Alan Turing add to this idea?

that you could launder all the mentality out of computation and make an entirely mechanical computational device, then build that up into artificial intelligence (AI). Both Darwin and Turing were visionary theorists of bottom-up design. One of the great things about a modern computer is that you know to a moral certainty that there is no extrasensory perception in there 鈥� it鈥檚 just as mechanical as an adding machine.

I call it 鈥溾�� 鈥� that鈥檚 my bumper sticker. Your basic computer doesn鈥檛 understand what arithmetic is or why it鈥檚 useful, but it does it perfectly. And evolution by natural selection is a process that鈥檚 breathtakingly competent, but has no understanding at all. Yet it has given humans the ability to reason, to understand. And now that we have such thinking tools, we are using them to achieve kinds of comprehension that no other species has. But I don鈥檛 think we ended up with the likes of Shakespeare simply through evolution acting on our genes. It took another evolutionary process to join forces with it 鈥� cultural evolution. .

How do you define memes?

It鈥檚 a way of doing something 鈥� including ways of doing something in your mind 鈥� that you weren鈥檛 born knowing. It鈥檚 not in your genes; it鈥檚 something you鈥檝e acquired from society, your social environment, and it鈥檚 contagious.

What is the basic unit of a meme?

Words. I don鈥檛 claim to know how and when human language evolved, but it鈥檚 very clear that when it first started, the hominins that were infected with these 鈥渕ind viruses鈥� didn鈥檛 know what they were doing. They didn鈥檛 know they were talking. They were making sounds to do things that benefited them, in the same way a butterfly uses eye spots on its wings to benefit it but doesn鈥檛 need to know why.

How did these mind viruses get us to here?

Gradually, the protolanguage of these hominins turned into a wonderful medium for manipulating others, and manipulating yourself. And as the products of the cultural evolutionary process have accumulated, they have enabled us to become truly intelligent designers, with forethought, planning and so on. Contrast a termite castle with Antoni Gaud铆鈥檚 wonderful church in Barcelona, . They look similar, but Gaud铆鈥檚 church is a product of intelligent design; it鈥檚 top-down, with a charismatic boss who thought it out in advance. There is no Gaud铆 in the termite castle.

What has to be explained is how, in one species, we鈥檝e gone from termite-style building to Gaud铆-style building, or to Turing-style building or Shakespeare-style building.

And that鈥檚 the transformation we need to unpick?

Yes, and it鈥檚 particularly thorny when you recognise that what we have between our ears is more like a termite colony than you might be happy thinking. The latest count is 86 billion neurons, each more clueless than a termite, with no boss. How on earth do you organise 86 billion neurons into Gaud铆鈥檚 mind or Shakespeare鈥檚 mind? That鈥檚 the puzzle.

As intelligent designers, we build computers. We don鈥檛 claim they are able to comprehend, but will they get there?

It鈥檚 possible, but we鈥檙e not there yet. Take Google Translate. Until just a few years ago, the idea that you could translate from one foreign language to another without comprehending what the words meant was laughable. But today Google Translate does just that, through machine learning and big data. It鈥檚 completely parasitic, though. It works because it has a huge database of successful translation and can sift through many millions of little bits of translation that have been endorsed by human translators. It uses them to guide its own translation. Otherwise it鈥檚 clueless.

It shows that it is possible to do decent-quality translation without any real comprehension at all. Some people view that as sort of the end of the line. But I don鈥檛 think it is.

What do you think is coming next with artificial intelligence?

The new wave is projects to add comprehension to existing machine learning or deep learning systems. There鈥檚 a project at the Massachusetts Institute of Technology that is supposed to permit a self-driving car to tell you why it鈥檚 doing what it鈥檚 doing. And there鈥檚 a new initiative at the US military agency DARPA calling for AI that can interact with humans in a reason-giving way.

I think these projects are much, much harder than the people who are doing them think. In any case, I鈥檓 not sure it鈥檚 a good idea.

Why might it be a bad idea to make AI with the ability to comprehend?

I think it鈥檚 better to use AI as a tool than a colleague. As soon as it is a colleague, we will have the problem we have with colleagues: we won鈥檛 know for sure what they know, and they may not want to tell us. Trust becomes an issue.

Are you worried about machine superintelligence?

Well, I鈥檓 not worried about super-AI that is going to design computers more intelligent than we are and then enslave us. I鈥檓 worried that we鈥檒l overestimate the comprehension of our devices and cede authority to them. It鈥檚 happening already. In certain areas of medical diagnosis and treatment, there are computer systems that can do a better job than any doctor, and doctors who overrule the software could later be deemed irresponsible or liable for malpractice. This means that, with the best of intentions, doctors will start following the advice of AI advisers. This will substantially diminish the role of doctors: if they are going to be turned into glorified doorkeepers with good bedside manners and an ability to push buttons, what about the rest of us?

鈥淗ow on earth do you organise 86 billion clueless neurons into Gaud铆鈥檚 mind?鈥�

In your new book, you also worry about our reliance on technology and that we are becoming 鈥渙vercivilised鈥濃��

I don鈥檛 think people realise how incredibly brittle the technology that we depend on is. If the internet goes down today, we鈥檇 be in a world of chaos. I鈥檓 worried about the first 48 hours, about people just going berserk if they don鈥檛 have the internet. They are not going to have TV, they are not going to have radio stations probably, their cellphones will fail. They are going to be plunged into electronic darkness and scared out of their wits. After all, they have just been plunged back into the 19th century, and frankly they don鈥檛 have the skills to cope.

Would you survive a digital dark age?

I could be a pretty good survivalist. I鈥檓 getting a little old, sore and cranky, and I don鈥檛 know what I鈥檇 do for my protein exactly, because I鈥檝e never hunted. But I鈥檓 pretty good with making fire by flint, at least.

This article appeared in print under the headline 鈥淭hat鈥檚 a termite colony between your ears鈥�