It is widely known that investment in artificial intelligence (AI) is concentrated in two Pacific Rim countries, China and the USA. But the strength of this duopoly is not generally appreciated, so here are two recent data points which throw it into sharp relief.
First, we have learned that Amazon’s R&D spend has reached 50% of the total spent by the UK on R&D. This means all the spend on any kind of R&D by the UK government and all the UK’s companies and universities. This astonishing fact becomes even more significant when you consider that a great deal of Amazon’s R&D spend is on AI, whereas not much of the UK’s R&D spend is on AI.
The second data point is a piece of research revealed by Jeffrey Ding, a Sinologist at the Future of Humanity Institute in Oxford, at the CogX conference in June. He started with the Chinese government’s widely-reported plan to invest $150bn in AI by 2020, and wondered if this included investments by China’s municipal and regional authorities. He discovered that it didn’t, and he managed to obtain data for much of that spend as well. It brings China’s total planned government investment in AI by 2020 to a breath-taking $429bn.
Earlier at the same conference, Matt Hancock, who, as minister for digital in the UK’s Department for Culture Media and Sport (DCMS), is the nearest thing the UK has to a minister for AI, claimed that with a projected investment of £1bn, the UK was “in the front rank of AI countries.” This frankly preposterous claim was echoed by the Mayor of London, on the publication of a report which trumpeted London as the centre of AI within Europe, which is not unlike claiming to be the best champagne producer in Maidstone, Kent.
It is misleading to describe the development of AI as a race, partly because there is no fixed point when the process stops and one team is declared the winner, but mostly because the enormous benefits that will flow from the development of better and better AI will accrue to people all over the world. Just like the development of the smartphone did. However, there are at least three powerful arguments why Europe really should make more of a contribution to the global project of developing AI.
The first argument is that AI will deliver enormous benefits to the world, and the faster we reap those benefits, the better. To cite just two of many examples, AI will improve healthcare so that people who would otherwise suffer or die will remain healthy, and self-driving cars will stop the appalling holocaust of 1.2 million deaths each year (and 50 million maimings) which human drivers cause. Europe has great wealth, great universities, and millions of smart and energetic people; it can and should be contributing more to realising these benefits.
The second argument is gloomier, but perhaps also more compelling. Europeans should not feel relaxed about the development of AI, humanitys’s most powerful technology, being so heavily concentrated elsewhere.
Jeffrey Ding argues that there is a far more lively debate in China about the government infringing on individual privacy than we in the West usually think. If so, this is great news, but it is hard to believe that China’s current approach to the development of AI would be acceptable in Europe. Most people here would be uncomfortable with schools using face recognition and other AI techniques to check whether the children are paying attention in class, and the way AI is being used to control the Uygur population in the western Chinese province of Xinjiang would also raise serious objections.
Many Europeans are also feeling slightly nervous about the great AI power to their west. So far, the development of AI in the USA has been a project for the private sector, but the government is showing signs of waking up to its importance, particularly with regard to military applications. The USA is currently a vibrant democracy, and has long been an invaluable ally. But things can change. President Trump’s ruminations about NATO and his willingness to initiate a trade war against the EU have shown that Europe cannot be certain that America will always share the benefits of its AI prowess.
The third reason is that AI might well be the source of much of the value in the economy in two or three decades’ time. Countries and regions which play only a minor role in developing the technology are likely to find themselves enfeebled.
To be clear, these are not arguments for autarky or self-reliant isolationism. We will all do better if the countries of the world collaborate to develop AI together, and share its benefits openly. That is the approach which Europe should champion. But sometimes, while planning for the best, it is wise to have a backup plan for the worst.
Jurgen Schmidhuber, one of the foundation figures of modern AI, argues that AI is currently dominated by the Pacific Rim countries for two main reasons: they both have huge single markets, and they both pursued muscular industrial strategies to promote the development of their technology industries. (Silicon Valley got started as a tech hub because of the sinking of the Titanic. To prevent a recurrence of that tragedy, the authorities decided that all ships must have powerful ship-to-shore radios, and it so happened that Silicon Valley was the home of a nascent radio industry. Later, the military research organisation DARPA funded cutting-edge tech research there, especially after America’s Sputnik moment – which was, of course, Sputnik.)
Schmidhuber urges that Europe should strengthen its single market (woops, Brexit – yet another reason for the People’s Vote), and that it should enact similarly forward-thinking industrial policies. He also observes that while the Pacific Rim countries clearly dominate the internet of humans, the internet of things (IoT) is still up for grabs. He argues that Europe is home to the leading companies in the development and manufacture of many of the component parts of the IoT, so the game is still wide open.
It is time for Europe to step up.