National AI Strategy - can the UK be the superpower it imagines itself to be?
Does Whitehall have the human intelligence, expertise, and skills to make the UK’s new National AI Strategy work?
The UK government launched its long-promised National AI Strategy on 22 September. The announcement was made during London Tech Week, with the capital still the epicentre of the artificial intelligence sector - not just in the UK, but also (whisper it) in Europe.
The 35-page document presents a 10-year plan to turn the UK into an "AI superpower", though most of the detail only covers the next 12 months. This suggests real tactical urgency.
In her introduction to the Strategy, new Secretary of State for Digital, Culture, Media, and Sport (DCMS), Nadine Dorries, wrote:
This National AI Strategy will signal to the world our intention to build the most pro-innovation regulatory environment in the world; to drive prosperity across the UK and ensure everyone can benefit from AI; and to apply AI to help solve global challenges like climate change.
That statement about a "pro innovation regulatory environment" should be put in the context of growing Whitehall chatter about divergence from Europe, including from GDPR, and the government wanting the incoming Information Commissioner to have a more commercial and enabling role, rather than a punitive one.
One possible interpretation of this is that the government wants to lower protections on sharing citizens' data to achieve its AI superpower ambitions, thus getting rid of all that EU red tape we hear about. So, let's pretend businesses aren't wading through reams and reams more of the stuff, with costs attached, since Brexit.
Moves are already afoot, according to the AI Strategy. Before the end of this year, the government plans to: publish a framework for government's role in enabling better data availability in the economy; determine the role that data protection plays in AI governance (presumably a reduced one); and develop an all-of-government approach to the technology. It also plans to draw up a draft national strategy for the use of AI in health and social care, via the NHS AI Lab.
With all this happening on a three-month timescale - there are longer-term plans too - the impression is that the government wants to sweep aside the status quo before Christmas.
The adequacy problem
The problem with all this is easily stated: the UK's fragile Data Adequacy agreement with the EU, where most UK data is stored, transferred, and/or processed, relies on the UK not diverging from European regulations. Especially at a time when the global direction of travel is towards GDPR-style rules, not away from them. The elephant in the room has ‘THE CLOUD IS BASED ON LAND - MAINLY IN THE EU FOR UK ORGANIZATIONS' written on it.
Incidentally, there's another elephant, which has the words ‘Together, the 27 EU nations have data on nearly 450 million citizens. That means much better-trained AI systems than the UK can hope for' written on it. But let's pretend the elephants aren't there. After all, that's what everyone in the Cabinet is doing. (Does a maximum data set of less than 60 million people equal a superpower? That's only twice the number of people in Texas.)
So, it stands to reason that, to stand any chance of becoming an AI superpower outside the EU, the UK needs access to all the big data sets it can get its hands on. Inevitably, that means data on you and me, almost certainly via the NHS and other government sources. In short, it looks like the ‘pingdemic' is set to become the kerching-demic. (I apologize for coining the word.)
The government's determination to commercialize public data is surely the subtext of the new AI Strategy. It must be, because a lot else has happened over the past five years.
For example, the Industrial Strategy put AI at the core of economic renewal (alongside robotics, new transport solutions, and green energy). Whitehall's Office for AI was set up, and the AI Sector Deal successfully combined public and private investment. Meanwhile, the Hall-Pesenti Review set out a clear vision for the technology, and we saw the creation of the independent AI Council, the Alan Turing Institute, and the Centre for Data Ethics and Innovation, along with new investment in NHS AI, Innovate UK's Brexit-inspired Expert Missions on AI and robotics, and £100 million in funding for 1,000 AI PhDs.
In other words, you could be forgiven for thinking that the government already had an AI strategy, one that has been working well for years. But lest we forget, the Industrial Strategy - the coherent, forward-looking document that did a good job of galvanizing public and private investment - was torn up in the Spring. In its place came Boris Johnson's ‘Strategy for Growth' (last measured at 0.1%).
The conclusion is inescapable: the current Prime Minister has no patience with anything his two predecessors dreamt up. Even if it delivered real progress and linked innovators, investors, and academics around shared goals, via challenge funds, Hubs, catapults, and more. This is Global Britain in 2021, folks! Who needs things that work anymore?
In 21st Century Whitehall, it's all change - and change and change again; a constant process of disruptive, witless upheaval that puts all previous gains at risk. Even the Minister launching the AI Strategy was new: DCMS Minister for Technology Chris Philp was only appointed last week, and reports to Dorries, who has been in her post for less than a week.
At the London launch of the Strategy, Philp said:
The Strategy is constructed around three pillars: Investing in and planning for the long-term needs of the AI ecosystem to continue our leadership as a science and AI superpower; supporting the transition to an AI-enabled economy, capturing the benefits of innovation in the UK, and ensuring AI benefits all sectors and regions; and ensuring the UK gets the national and international governance of AI technologies right to encourage innovation, investment, and protect the public and our fundamental values.
It wouldn't be a Johnson era presentation without talk of being a superpower, but Philp added:
The UK is in an admirable position, with a rich legacy of spearheading many of the greatest leaps in AI over the decades; we have advanced scholarships at universities and research centres across the country; and, in London, the most vibrant start-up scene outside of San Francisco, with companies like DeepMind, Benevolent AI, and Improbable pushing the envelope of what's possible in AI in their respective fields.
The UK saw 20 tech firms reach Unicorn [billion-dollar valuation) status in the first half of this year, including Tractable and Zego. We have ten privately owned tech firms valued at over $10 billion. And in your [the AI] field, Exscientia, which uses AI to discover new drugs, raised nearly a quarter of a billion dollars this year. From Alan Turing to [DeepMind founder] Demis Hassabis, the UK has always led.
This is true. But DeepMind is evidence that while the UK has always had countless brilliant innovators and can hothouse new ideas (not just in AI, but also in Fintech, robotics, and quantum technologies), scaling them to global status without a US behemoth buying them first is difficult. It's even harder now, outside of the EU and with few trade deals signed.
The government is completely committed to maintaining and building the UK's leading tech position, including in AI. The UK is the clear European leader in AI and third globally behind only the USA and China - and I know we can catch them up. So let me say today: we want AI innovators to locate and scale up in the UK. We want you to succeed in the UK. We want the UK to lead the world in this field. It is a critical national priority.
It is critical because AI is a profound technology. It is the future. Your field has the potential to - in fact, it will - infuze every aspect of our personal and business lives in ways we cannot currently imagine. As [computer scientist] Andrew Ng has argued, AI will fulfil a similar role in the coming century to the one that electricity and then regular computing played in the last century: a meta-enabler which underpins activity in a huge range of fields, including those without initially obvious applicability.
How to play catch-up
Can the UK really catch up with the US and China? In research, ideas, and innovative applications, no doubt. But the troubling facts are these: first, the UK has never backed its rhetorical superlatives with much more than piecemeal central investments. Granted, the investments in AI could be said to add up to £2.3 billion, but only if you throw in - as Philp did - investments in numerous related technologies, such as autonomous vehicles, robotics, and more. Even so, there are cities in China with more government AI funding than that.
And second, short of opening an office in San Francisco, the traditional route for scaling UK tech start-ups has always been through Europe. Today, that is much harder - unless the new enterprise sets up shop in the EU itself, as some are doing. Brexit is the gift that keeps on taking.
Casting around for non-European allies was the real driver for Innovate UK's Expert Missions programme in 2018 and 2019: to find new partners for the UK's entrepreneurs and academics, such as the US, South Korea, Japan, and Israel. I know, as I was on two of them: the desperation to open new doors as Europe's avenues closed was palpable.
The government now plans a new series of Innovation Missions and will launch a new national AI research and innovation programme that will align funding across UKRI (what was wrong with the old one?!). It will also launch a joint Office for AI/UKRI programme to stimulate the development and adoption of AI in high-potential sectors.
Good news. So, what else is in the AI Strategy? In the short term, the government will consult on options for a ‘Cyber-Physical Infrastructure Framework' (whatever that is); host AI and data science skills bootcamps; and publish an AI strategy for the Ministry of Defence.
In the medium term - which the document defines as six to 12 months (I'd argue that is short term) - the government says it will publish research into what skills are needed for employees to use AI in business and identify how national skills provision can meet those needs.
The context for this is a recent government report which found that thousands of AI and data science jobs are being left unfilled, because of a lack of people with relevant AI, data science, and business skills. As I noted in a separate report this week, thousands of jobs in cybersecurity (10,000, in fact) are also unfilled. In field after field the UK can't find enough workers, even to pick strawberries.
The Strategy will also support the National Centre for Computing Education to ensure AI programmes for schools are accessible, and support bringing a broader range of people into AI jobs by ensuring career pathways highlight new opportunities.
Speaking at the July eForum on AI, the UK's computing skills champion Professor Dame Wendy Hall, co-author of the independent AI review, said that diversity is the real key to fulfilling the UK's AI ambitions. Not just diversity in gender and ethnicity (vital in an industry that is overwhelmingly white and male), but also in terms of age and background. "If it's not diverse, it's not ethical," she said.
The government will also evaluate the private funding needs and challenges of AI scale-ups - I can help here: they need the pathways they once had in Europe - and implement the US/UK Declaration on Cooperation in AI R&D. The latter suggests the UK is drifting across the Atlantic - but not towards a US trade deal, as yet. The government also plans to roll out new visa regimes to attract the world's best AI talent to the UK.
Intriguingly, the Strategy says the UK will extend aid to support local innovation in developing countries - at a time when the government has slashed foreign aid as a sop to bank benchers and supportive newspapers. The implication is that UK aid now comes with strings attached: aka ‘Work for us, we need new friends - and can't fill our own job vacancies'. Whatever you do, don't mention the Empire.
The government also plans to build an open repository of AI challenges with real-world applications and plans a new white paper on governing and regulating AI. Meanwhile, it will work with the Alan Turing Institute to update guidance on AI ethics and safety in the public sector. More good news.
Encouragingly, it also plans a complete, in-depth analysis on algorithmic transparency, with a view to developing a cross-government standard. Not before time: Whitehall's record on algorithms is poor, as the outgoing Education Secretary could tell you. A new AI Standards Hub will coordinate UK engagement in AI standardization globally, and the government plans to identify ways in which AI could provide "a catalytic contribution to strategic challenges".
Some final bullet points: the government plans to undertake a review of its international semiconductor supply chains (don't mention tensions with China!) and will monitor and use the National Security and Investment Act to "protect national security while keeping the UK open for business".
A bold, comprehensive document, whose aims are undermined by the government's actions in other areas of domestic and international policy.
But the conclusion is clear: as the UK's single biggest IT user, government plans to turn AI on itself and its citizens, because - outside of the telescopes that are mapping the universe - that's the biggest data we have.
So, the question then becomes: Does Whitehall have the human intelligence, expertise, and skills to make the strategy work? Answers on a postcard, please.