Marcus du Sautoy believes that artificial intelligence learns from humans the way a child learns from his parents. He says that whether AI turns good or bad in the long run will depend on the environment humans develop it in.
Does the AI emulate humans the way children learn from their parents?
Exactly. I often give the comparison to, it’s like the relationship between a parent and a child. Of course, the child reflects the genetics of the parent, yet the child is different because they interact with the environment, they interact with other friends, other people, other countries, and that is part of its learning process. For example, Picasso, his genes are the genes of his parents. Yet we don’t say that the art that he produced is the result of his parents. Yes, partly that may have happened, but it’s mostly his environment. So the AI is written, the code, first of all, by the parent, but then the AI interacts with everything else. It becomes its own, has its own identity, separate from the person who coded it.
Yes, you’re saying that it is learning from the environment around it. So, if the environment around it is malicious it can learn that behaviour?
Unfortunately, yes. The nice thing is you can steer, so you can steer, you can try and get an AI to identify malicious behaviour. Then you give it malicious behaviour and it says, okay, I recognise, I now recognise that that’s bad and that’s good. Gradually it starts to get a moral code. You can incentivise it to learn a moral code.
How can you incentivise AI?
It’s like playing a game. So, you say you lose the game if you play… Here’s an example of bad behaviour. It learns that. Here’s an example of good behaviour. Now here’s an example of something new. What do you regard this? Is this closer to something bad or something good? And it makes a guess and it says, no, you got that wrong, that’s bad. And so it will update its code and say, okay, I will get that right next time. And so it’s a learning process. It’s about learning what’s good. And so then it can be used in, for example, in a social context to help other people identify good and bad. So AI is very, very flexible. And it’s about how we use it.
There’s a very interesting novel. For JLF, I interviewed Ian McEwen, who wrote a novel called Machines Like Me. And it’s about a moral dilemma between two humans and an AI. And in the book, Ian explores the fact that the AI actually may be more moral in its decision. It will be thinking about the greater good of society, whilst the humans very often think about individual good. And so maybe AI often is better at making moral decisions.
Can AI develop an ego?
Well, I think AI can definitely demonstrate characteristics of ego. So again, in some ways, what do you mean by an ego? That means almost you’re saying, is AI conscious? Because it’s got to have a sense of self different from other. But often when we say a sense of ego, we mean prioritising the self. So again, a sense of self will come very much later. That’s about consciousness. But is it altruistic or selfish? Or different AI will be different according to its learning process?
Do you watch science fiction movies or series like Doctor Who?
Yes! My children are obsessed with it. It has a lot of things which we think are futuristic or sci-fi. Science fiction is very often good at exploring what’s possible which then becomes science fact. So not everything is possible. The time machine, for example, HG Wells, we don’t think the current physics doesn’t have the opportunity to go back in time. And of course that’s a great theme for a lot of science fiction novels, is going back in time. I mean, Doctor Who is a time traveller, the Time Lord. But I think it’s very interesting that science is about imagination and that can first happen inside something like Doctor Who or a science fiction story. And one of the things you see, we’re already seeing cybernetics taking great advances. So, yeah, scientists’ imagination is quite close to a novelist’s imagination.
But then, if AI or advancements in science can be harmful for the human race in the future, do you think we should curtail this imagination?
I think you can never do that. So, one strategy is to try and put things back in the box. And I don’t think humanity has a very good track record with keeping things in the box. I think the best thing is to let these things out of the box and then for us to understand how do we limit the bad and profit from the good. I don’t think I would want to have not discovered what gets released by splitting an atom. But that can be used to solve the energy crisis and climate change. Or it could be used to blow us all up. But it’s about what we do with that technology. My belief is always we should not ever limit discovery but we should limit what we do with that discovery.
Okay, do you believe that there’s a theory that there’s a left brain and right brain function. The right brain I believe is for creativity or logic. Do you believe maths comes only for those who are more left brain oriented?
So first of all, this science of left brain, right brain has been shown to be incorrect. I mean it’s true that language is more located in the left brain, the Broca area. I’ve made quite a few programs for the BBC about the brain. There’s much more fluidity between the sides of the brain. So, if one side of the brain gets damaged, very often it can recover and replace activity in the other side of the brain. Language seems to be very localized in the left. Now you’re right, there does seem to be more emphasis towards, for example, the imagination and solving new problems seems to be a right brain activity. And actually, there’s a pattern recognition. And so, mathematics is actually quite a right brain activity, interestingly. So not a language side, but more about new environments where often we see right brain activity. And interestingly, right brain controls left hand. I think there’s evidence there are more left-handed mathematicians than you would have in a normal population. But I think one has to be very careful because we’ve seen that the brain is not as cleanly divided up as we thought it was in a left right brain.
How can you get students who don’t love maths in a later stage of life to actually fall in love? Let’s say quadratic equations, you know all those complicated math, how do you get them to love math?
I think coming back to mathematics again that you didn’t get the first time, for example, quadratic equations, you didn’t get it the first time, but we’re a bit like rabbits in a headlight. You know, we freeze. But then when you come to it again you say, I don’t understand why I found that so frightening. So going back to something, very often you can relax and you can find a new way in. In fact, I’m a big supporter of actually redoing years. There’s no point going on if you’ve failed the year. Much better to redo it rather than crashing on and getting more and more anxious. But I think, you know, our brains maturate. They get more sophisticated. And somebody who’s 18, for example, will begin to understand why the things that they were learning might be important for what they’re doing. So, for example, in London, the University for the Arts London, where a lot of artists go to train, there’s a course which is Maths for Artists. It is the fastest course to be signed up for. Because by then the artists who are 18, 19, 20 realise, oh my gosh, to make this thing or to plan this thing or I want to make something with this particular structure. And they realise, oh my gosh, but I actually need maths to this. Then they are suddenly trying to make things which they need mathematical ideas. And so they go, if only they told me that when I was 11, I might have concentrated but I thought it was nothing to do with what I love, which is painting or architecture. My books are mostly written for adults who I want them to value the maths such that changes the culture and it feeds down to their kids or their employees and they value maths and they realise, no, no, having somebody who thinks mathematically is a very powerful team player.
Someone once said that you should always let your children take a business course to learn about life.
I don’t know. I’ve heard less, I’ve heard a movement away from that. That actually, it’s much better. I’ve heard more and more people saying don’t do MBAs and things like this. That actually, it’s teaching a very set way of thinking. Actually, better to do something very, just expanding the mind. Go, do philosophy instead and expand the mind. I would prefer you not to go and do a business degree. Go and do a maths degree. It makes you think about solving problems. Then go and work in a business. Don’t go and learn a business degree. I think that’s a terrible waste of an opportunity.
…Concluded
This article by Shailaza Singh appeared in Rashtradoot Newspaper’s Arbit Section on July 11th, 2024
Leave a Reply