My Times column on artificial intelligence:
As a member of the House of Lords select committee on artificial intelligence, whose report is released today, I was struck by two things during the course of our inquiry: how well placed Britain could be to take advantage of the new technologies that go under the name of AI, should we choose to play our cards right; and how pervasive and invisible this technology will prove to be.
The first point was driven home during an evidence session with a more than usually brilliant German professor, Wolfgang Wahlster, chief executive of the German Research Centre for AI. He said: “We have a very special approach that is based on Germany’s industrial strengths . . . This is quite different from the US approach, which is based more on internet services. We are more in the physical domain; you know that Germany is well known for its engineering and manufacturing.” He added: “Historically, the UK was the leading country in Europe in AI. It started in Edinburgh.”
The point is, we can play to our strengths. The Germans will make clever washing machines and cars, the Americans and Canadians will concentrate on the internet, the Chinese will get worryingly good at smart espionage, but there is a niche for Britain, a country with a hotspot of world-beating expertise in biomedicine, finance, law and research as well as a reputation for getting ethical regulation right, for example in reproductive technologies. We will never compete with the United States or China in investment, but there will be an opportunity for whoever unleashes a torrent of well-channelled experiments to transform healthcare, the law and finance. The UK has more tech startups and more leading universities than the rest of Europe put together, and they rub shoulders with doctors and lawyers in Shoreditch, east London.
The most high-profile achievement of AI in recent years was the victory of AlphaGo, a program developed by the London-based DeepMind, over Lee Sedol, a world champion at the game of Go, in 2016. Watched live by up to 60 million people across Asia, the six-day tournament showed the underdog program, badged with a Union Jack, unexpectedly defeating the Korean equivalent of Garry Kasparov by four games to one, to gasps of excitement and groans of despair in a packed press room. AlphaGo’s 37th move in the second game was considered suicidal; it is now thought stunningly original. DeepMind is owned by Alphabet, Google’s parent company, having been bought in 2014 on the condition that it stayed in London and expanded under its founder, Demis Hassabis. British technology has never had such free PR.
True, games are only games. What counts is how far AI seeps into everyday life and transforms the prosperity of humankind. The main gains to society from electricity, say, or the internal combustion engine, are not the profits of the electricity companies or carmakers, but the efficiencies we all experience in our normal lives as a consequence of having those technologies. It will be the same with AI: consumers, not producers, are key. So while a thriving British AI development industry would be great, it’s not the main prize.
The peculiar fact is that is as soon as something becomes possible in the world of AI, we stop calling it intelligence and think of it as better software. The present excitement is over specialised AI applications, not general intelligence of the kind that could rival a rounded human being. But at their heart lies a new approach. Whereas the previous generation of AI was based on “expert systems” pre-loaded with human expertise, today’s algorithms know almost nothing at the start, but crunch lots of data to learn how to do something. It is the access to deep draughts of data, and the ability to learn from them, that make the difference. Which is, of course, the issue. The handling of personal data by all-too-human intelligence has turned into the biggest ethical challenge of this brave new world. Not job losses: automation has led to more employment, not less, in every technological revolution since the threshing machine, and will this time too, in my view, though it is now lawyers and medics who may have to readjust rather than farmhands and factory workers. Not inequality: nothing is proving more egalitarian than the smartphone, and AI should likewise be mostly available to rich and poor alike.
And not Hal, the computer that learns to be malevolent in defence of its off switch in 2001: A Space Odyssey, a film that this year celebrates its 50th anniversary. This week the hip speech-fest known as TED will take place in Vancouver, and it features the usual dystopian worries about AI. For instance, Max Tegmark, president of the Future of Life Institute at the Massachusetts Institute of Technology, will argue that “in creating AI, we’re birthing a new form of life with unlimited potential for good or ill. It’s truly urgent that we begin imagining different AI futures so that we can build the one we want. We only have one shot at this.” Instead, society must grapple with the dilemma of preserving people’s privacy and ownership of their data, while letting machine-learning algorithms harvest insights of value to everybody. The starry-eyed optimism that greeted, for example, the Obama campaign’s use of Facebook data in 2012 has given way to rage at similar practices used by the Trump campaign or Russia. Is there a sensible compromise to be found? Of course.
Technology has a habit of changing everything and nothing at the same time. The rise of AI will make life richer, better, more interesting, and then, without any fuss at all, be taken for granted, as electricity and cars are. Yet the same old, all-too-human ethical and political dilemmas about how much power to allow the state, or private enterprise, will persist.