The UK is a global hub for artificial intelligence (AI) innovation. Behind the US and China, it has a strong claim to be the third best place in the world to be an AI researcher or company founder. While it was sold to Google, Deepmind, which recently “changed everything” by solving the protein folding problem, is testament to the UK’s AI talent. But in the words of a House of Lords Liaison Committee, there is “no room of complacency”.
With Brexit and the pandemic, the UK Government has been a little distracted, but it’s trying to get back on track, announcing a plan to make the UK a global center for the development, commercialization and adoption of responsible AI. This plan will be published in the Autumn and to that end, we have put out a report as a starter for ten (well, eight recommendations) on what should be considered.
The US leads the world in AI with the only realistic threat to its crown coming from China. But that doesn’t mean the UK can’t be a significant player. And, as the report suggests, the UK has a few things to learn from its neighbours across the pond.
Séb Krier, author of the report and former Head of Regulation at the Office for AI, calls for the creation of a pool of cloud compute credits for the UK R&D ecosystem. AI research is computationally intensive and expensive, so is dominated by the likes of Google and Amazon. UK-based academics need access to that sort of power.
As the report details, following a recommendation by the National Security Commission on AI, last year the US introduced the National AI Research Resource Task Force Act. This “aims to spur and democratise AI-centred studies and applications by developing a national cloud for scientists and students to use – meaning that typically expensive experiments could be available to a wider range of institutions and researchers than is currently the case.”
While the report doesn’t recommend the UK goes the whole hog in trying to recreate a public AWS, it does make the case for subsidized cloud credits to level the playing field for AI academics.
The report also calls for the UK to upgrade public data infrastructure and open up datasets. The UK is often said to lead the world when it comes to open data, with the success of Transport for London opening up its data to create the likes of CityMapper (with an estimated £130 million a year to London’s economy) and the Bus Open Data Service initiative proving the value of of public data.
However, as Krier writes: “The government holds an enormous trove of data, yet a lot of it remains inaccessible, mislabelled, out of date, or not correctly formatted.” He points to an Open Government Network letter urging the government to improve its open government agenda, after the country was placed under review by the Open Government Partnership (OGP) for failing to meet the required standards.
The report calls for the UK to replicate the US Data Coalition’s proposed OPEN Government Data Act, which sets an official presumption that “Government data assets made available by an agency shall be published as machine-readable data…in an open format, and…under open Licenses.”
Krier also suggests that the UK should lower barriers to immigration to attract foreign AI talent, citing a Microsoft report that finds the UK is facing an AI skills shortage that isn’t being met by training. The UK needs immigrants to make up the shortfall, and while the UK is becoming increasingly open to international talent after leaving the European Union, it’s not clear to what extent this will offset the barriers to European talent since the end free movement.
Given the academic requirements for AI, scholarships might hold the key. Krier points to America’s success in attracting top talent has been largely been down to the targeted strategies of a few universities, such as MIT, of actively recruiting top foreign talent with funded admission offers. As such, the report recommends the Government provides additional funding to leading research universities to fund postgraduate scholarships in AI-related fields.
It’s not just about pinching good ideas – and talent. There are legitimate geopolitical – and even existential – concerns about the risks of AI. It’s something that concerns Dominic Cummings, the Prime Minister’s former chief advisor. There are fears AI could bolster authoritarian regimes, with the report suggesting that the UK works closely with the US to build affirmative alternatives to digital authoritarianism, especially for countries currently forced to decide between no growth and stability, or growth and stability but with authoritarian strings.”
In the words of Krier: “Amidst a global retreat towards digital nationalism and protectionism, the UK should push for an open and global technology ecosystem without compromising on safety and security.”