Dr. Stuart J. Russell is a British computer scientist, Professor of Computer Science and Smith-Zadeh Professor in Engineering at the University of California, Berkeley.
France is known for its high-fly thinkers whereas the UK has championed in pragmatic and technical development (industrial revolution, ruling the seas…). What is then the most favoured nation for AI? And how could the others catch up?
The US has a significant lead. Interestingly, however, most advances that have occurred have come from non-US-born people. So it’s the influx of thinkers and academics into the US from other countries that has played a huge role in US supremacy. The UK as well has done a pretty creditable job of making academia more attractive to people outside its borders. One obvious reason is that English is the lingua franca of research; it’s more difficult for France to achieve the same level of openness because that means people having to learn three languages. Most senior scientists already have decades of experience speaking English no matter what country they come from. So that’s an issue, but I think that most universities will probably run and are already running degree programmes and research labs with English as the primary language.
Art conveys culture and mirrors civilisations. What will be its place in an AI based world?
People are already using AI as an adjunct to artistic expression, particularly in music: composers work with computers to create and perform music. It’s like having a jam session where some musicians are actually AI programmes and they produce responses and so on. You can imagine that the AI systems have become more capable linguistically–so they can understand what people are saying and can respond in ways that are relevant or sensible–you can imagine that you might have ways of producing plays which, rather than being completely scripted, actually involve AI programmes that participate in real time. A new art form if you’d like. Virtual reality, it is a medium that could allow for far greater artistic expression than any medium we’ve seen before. It remains, however, extremely hard to author because of the vastness of the choice and the volume of authoring decisions that have to be made in order to produce an interaction with the viewer. I think that we will inevitably need AI systems to help in the authoring process and they’ll be part of what constitues the VR experience because AI systems will be driving other characters in the environment. But I think there’s another important part to this which has to do with how humans manage their interaction with AI. How our society organises itself to deal with the presence of AI systems. There we need the help of the cultural and artistic community to understand what’s happening and to figure out how to avoid the pitfalls. For example, if you’re worried about cultural enfeeblement where people essentially hand over their civilisation to machines – which would be the end of human culture–there is a need for the development of a cultural resistance. We need to install a sort of spartan way of resisting, otherwise we’re finished.
What is the role of AI in security and defence? After the Cold War’s nuclear arms race, could we witness a new race take place? One to AI weaponry?
We already are. It’s very difficult to see how we’re going to avoid a full-blown arms race. If there’s no treaty banning autonomous weapons, it is undoubtedly going to get out of hand very quickly. Governments are not listening to the reasons for which you would not want autonomous weapons for they already are in the grips of an arms race. We’ve been trying to convince governments that the end point of an arms race would be weapons of mass destruction which would be very likely to be used against human populations, a truly disastrous outcome.
The many things AI is going to learn are they not always biased? How can we expect them to emit fair and just propositions when what it will learn is from a long history of biased news and historical inaccuracy.
When you read a book, what you have evidence of is that the author of the book put those words onto the page. You don’t have evidence of the truth of those words. One advantage the machines have is that they can detect inconsistencies, and the nice thing about truths is that they are always consistent. I suspect that, in the long run, machines should be better at detecting inconsistencies and false information than humans, partly because they have access to so much more information than we do.
What is the role of AI in economic development?
Economists have been using mathematical systems in economic modelling for ages. The Gosplan is a great example of this. However, although econometrics could substantially help eliminate world poverty, I don’t it think that is the main issue. If we wanted to end world poverty, we could. It is simply not in the interests of the governing classes who have a great deal of power in their respective countries.
Finally, AI, a force for good in the world?
A force in the world. I think we can obviously use it for both good and evil. It can eliminate ongoing conflicts over resources and promote economic development but can also be used to create even more inequality. However, even if you see AI as this tremendous force for good that will solve economic problems world wide, one very important question remains, and that’s what people are going to do with it. This links back to the conversation about enfeeblement which is another way of describing the unemployment problem: we become primarily consumers and not producers. In the process, we lose the incentive to learn, to create, to manage; we become enfeebled in the process. Whichever way you want to think about it, that’s a very serious issue.
Interviewed by Alexander Lake
Franco-British Portrait Gallery