Back in October 2017, publication of the Hall & Pesenti review, Growing the Artificial Intelligence Industry in the UK, caused something of a stir – if subsequent actions are anything to go by.
Seven months after the independent, government commissioned study hit ministers’ desks, the AI Sector Deal was announced. Trumpeted as a £1 billion Whitehall vote of confidence in the industry – which has deep, historic roots in the UK – closer examination revealed it to be a more conservative £300 million of public funds backed by a promise of £700 million from industrial and academic partnerships.
It still provided a welcome boost to the country’s ambitions, even if the announcement itself was fumbled by the mandarins of SW1. At a Westminster eForum event in February 2018, the Deal was trailed for publication in early March. However, policymakers sat on it until May, when it finally emerged days after the EU had published its own AI strategy to the tune of €20 billion – a pointless and frustrating own goal for the UK. There are two possible explanations: incompetence, or frantic last-minute negotiations to bump up the UK Deal to a PR-friendly £1 billion.
But it was still an excellent year for the country’s burgeoning AI sector. In 2018, British AI companies raised nearly double the combined funding of their German and French counterparts. The UK’s venture capital investments in the technology have risen nearly sixfold over the last five years. 2018 also saw Mayor of London Sadiq Khan announce a strategy to make the capital into the epicentre of a global industry.
The UK is currently home to one-third of all of Europe’s AI companies, nearly twice as many as any other EU country. Meanwhile in 2017, the UK was ranked joint first globally on the Open Data Barometer, a measure of how well governments are using open data for accountability, innovation, and social impact – a finding that may surprise some.
The government believes that AI could add £232 billion to the UK economy by 2030, boost productivity in some industries by nearly one-third, and generate savings of up to 25 percent – figures that presumably don’t factor in the potential economic impacts of Brexit.
But the UK is far from alone in its ambitions to be in the vanguard of AI: South Korea, Japan, China, Singapore, Taiwan, India, Sweden, Denmark, France, Canada, and Australia are just some of the countries to have published AI strategies since 2017. Nearly all of them are doing better than the UK in implementing related technologies, such as robotics and automation, while generally keeping human unemployment low – as we explored in a recent diginomica report.
According to IDC, worldwide spending on AI systems is forecast to reach $35.8 billion this year, a 44 percent increase on 2018.
While UK government investment is maddeningly small for its bold ambitions, it’s at least smart and well targeted within the Industrial Strategy, which matches ‘Eight Great Technologies’ with four Grand Challenges: an ageing society, clean growth, solutions for future mobility, and maximising the impact of AI and data analysis.
What progress has been made?
So where are we now, two years on from the Review? One of its authors, Professor Dame Wendy Hall, Regis Professor of Computer Science at the University of Southampton, remains an outspoken, galvanising force for UK AI research and innovation. The other, Jérôme Pesenti, is now VP of AI at Facebook, a company that has done more to arouse suspicion of the technology than any other. Indeed, his appointment made Hall laugh out loud on a conference stage last year, when she said:
It’s ironic in terms of what we were trying to do with the Review, which was all about job creation and economic growth and trying to keep that growth within the country! Erm, so... I won’t say anymore.
However, she did reveal that so short was the timescale set by government for the Review, that there had been no time for deep consultation with industry, while her dealings with Pesenti had taken place via WhatsApp (now owned by Facebook) while he was in New York.
So the question is, what does Hall think about it all now? And what has the UK actually done since the Review and the launch of the AI Sector Deal?
Giving the closing keynote at another Westminster eForum last week on UK AI policy, Hall set out some of the nation’s achievements so far. Joining her at the event were Gila Sacks and Dr Rannia Leontaridi, joint heads of the Office for AI, which has origins in the Review alongside the new AI Council.
The statistics look good. UK has opened 16 New Centres for Doctoral Training in AI at universities across the country, delivering a possible 1,000 PhDs over the next five years. It has also launched the AI Turing Fellowships (via the Turing Institute) to attract and retain top AI talent (don’t mention Brexit!), with the first five recipients already announced.
Also on the table to date has been industry funding for new Masters places, along with 2,500 places for AI and data conversion courses, including 1,000 government-funded scholarships – 50 percent of which must go to under-represented groups, said Hall. That programme begins next year.
The UK has also: commissioned the Government Digital Service (GDS) to deliver a Review of AI Adoption; published an AI Guide for Government; provided up to £50 million for five new centres of excellence in AI-enabled digital pathology and imaging, and £30 million for the new Bayes Centre in Edinburgh, which focuses on data science.
Other good ideas include: support for the Early Diagnosis Mission, which aims to use AI and analytics to shift medicine away from reactive treatments and towards predictive health management; funding of £3 million for a trio of AI research programmes with business, and up to £79 million for programmes in the engineering, urban planning, and healthcare sectors; plus a range of other initiatives, including £13 million for 40 other exploratory AI and analytics projects nationwide.
The UK has also partnered with the Open Data Institute (ODI) and Innovate UK on three Data Trusts pilots, which focus on tackling the illegal wildlife trade, reducing food waste, and improving public services in the Royal Borough of Greenwich in London.
Hall said she was excited by the Data Trusts concept, but – characteristically – was more candid about these and other achievements than the government’s spokespeople, who were effectively in purdah as Parliament geared up for the December General Election.
I didn’t realise how little budget was going into this area. [...] The Data Trusts projects are worthy, but small and bottom-up and nowhere near the scale that we need.
As a small country off the coast of Europe, which we will always be geographically, what’s our best strategy to make sure we do maintain and develop ourselves as a centre of excellence so that we can be producers in this field, as well as consumers?
Diversity in AI
Hall’s personal focus is developing the skills base for the UK to succeed in a hyper-competitive field. Core to this is growing diversity in an industry in which only one in 10 UK AI workers is female.
It’s the hardest aim to deliver on, which is why we put it at the top – it usually comes at the bottom. I wrote my first paper on the lack of women in computing in 1987 and I feel that I have signally failed in actually shifting the dial. We probably have fewer women in computing than when I started in 1984. That’s a sad indictment of the world that we’re in.
It’s deeply, deeply cultural [in the UK]. I travel the world and see what’s going on in other countries. For example, I go into computer science classes in India, and over 50 percent of the classes there are women. So it’s not DNA-based, it is culturally based, and we are not breaking that cycle – despite all the efforts.
It’s a passion of mine. So we put diversity first [in the Review] to signal that that is the most important thing to do. I should say though, that the situation in China is even worse.
This may be one of the reasons for the UK’s AI Masters courses being, in Hall’s words, “full of mostly Chinese students at the moment”. She added:
There’s nothing wrong with that. They come over here, get the education, pick up some of the values perhaps, which is fabulous, but then they go back and we lose them.
Eighty percent of papers at AI conferences are presented by China and the US, according to Richard Stirling, co-founder and CEO of Oxford Insights, who also spoke at the event.
The industry’s focus on machine-learning programmers can only make the diversity challenge worse, continued Hall, because it is “drawing on a pipeline that is mostly white male geeks doing computer science and maths degrees” – some of them, presumably, with Hall as their teacher.
Whether she was being indiscreet or simply honest, her remarks echo comments made at the World Economic Forum a couple of years ago by MIT’s Joichi Ito, who described his own students as white male “oddballs” who prefer the binary world of computers to the messy, emotional one of humans. His words, not ours. No wonder some of them prefer to stay indoors.
To combat the industry’s overwhelming white male bias, the UK needs to teach AI in its broadest sense, including ethics, social responsibility, design, and testing, said Hall, and move far beyond a narrow focus on coding and STEM, opening up the field to professionals in every industry.
I would add, there are huge salaries for machine learning programmers today, but if we get it right, we won’t need them in the future. I see it as a transient skill, because if we do get it right in the future, machines will be writing their own code. So make hay while the sun shines, ML coders! You won’t be the top, the bee’s knees, in the future, but boy do we need them now.
Getting single parents – 90 percent of whom are mothers – into AI careers would be one way of tackling the diversity crisis, she added. This echoes the work of organisations such as Padlock, which focuses on training a flexible workforce of single mums to be cyber security specialists.
How do we help people who want to ‘reskill’ mid-career or after a career break? Hall asked. And while much of the focus is on young people, she added, “boy do we need to get this right for the elderly”.
So what are the benefits of this national focus on AI – minus the familiar narrative about predictive health? For the Office for AI’s Rannia Leontaridi, it’s primarily about productivity – an equally familiar narrative from government representatives.
Productivity. We need to be significantly more productive and innovative; we are not alone in the world in looking at AI [...] so progressing AI is an obligation not an option. We are competing for research resources and talent.
Again, don’t mention Brexit. For co-leader Gila Sacks, while it may be easy to picture the Grand Challenges and understand their benefits, the UK also has to keep its eye on AI’s potential to level the playing field for everyone in society.
But once again, it was left to Hall to be both the passionate advocate for AI and the spectre at the feast. Adding a big dose of realism, she said:
This is going to be much, much, much, much, much harder than people think. We have this macho ‘We can do everything better and faster than you can’ attitude, but how can it work in the real world safely and reliably, and be tested?
The UK’s AI strategy needs to be ambitious, but also pragmatic and sensible and not just “a load of big hype”, she added.
Wise words from one of the world’s most passionate AI advocates and campaigners, who closed with the comment: “If it’s not diverse, it’s not ethical.” Amen to that.