The innovation game
Can innovation strategy keep up with AI?
This essay is a submission for the TxP Progress Prize. If you enjoy it, I suggest seeking out the fellow submissions.
Britain needs some new ideas: innovations, like miracles, can grow economies and solve huge problems. Artificial intelligence, which scientists and economists claim marks “the invention of a method of invention”, will accelerate or even automate innovation. Now is the time for politicians to build an innovation strategy that’s ready for the reinvention of invention itself.
Through an explosion of innovation came the Industrial Revolution, and so we have innovation to thank for how global living standards have skyrocketed since 1800. The government anticipates AI could herald something just as transformative. With AI “creating machines that exceed human intelligence”, the 2017 Innovation Strategy declared, we’re sitting “on the cusp of transformative industrial change unlike any the world has ever seen”.
Transformative change is needed. Today, Britain feels weighed down by too many years of economic stagnation and daunted by huge challenges like the climate crisis. Some economists even venture that a slowdown in the rate of innovation is to blame for waning global economic growth.
To begin with, Britain has no choice now but to innovate our way through climate change. A green transition will necessitate embracing green technology like clean energy, including wind and nuclear, and alternative proteins, from labs not slaughterhouses.
AI will be key. Already being deployed to find new pharmaceutical drugs, AIs will become increasingly important to innovation in the coming years. Google DeepMind claimed the first “genuine, new scientific discovery [...] by a large language model” last month, when their AI cracked a previously unsolved mathematical problem. DeepMind co-founder Demis Hassabis believes that AI will achieve a “renaissance of scientific discovery”. More than that: Hassabis thinks that artificial general intelligence, a system clever enough to outthink us, to have new ideas before us, “could be the ultimate general purpose tool to help us understand the whole universe”.
As the UK crafts its AI regulation, playing catchup to the EU and the Biden administration, the Conservatives say they don’t want to be “stifling innovation” and Labour say they “don’t want to smother these innovative companies”. Both parties will have to find a place for AI within a broader innovation strategy.
Around 2020, the Conservatives were having an innovation moment. Gross domestic expenditure on R&D (GERD) looked like it had been languishing at 1.7% of GDP for years. (OECD average was 2.5%). But now, scientific research was set to save us from the pandemic, and Dominic Cummings – yet to fall out of favour with Boris Johnson, who was yet to fall out of favour with MPs – was making innovation and technology central to the Prime Minister’s agenda. And so Johnson pledged to spend £22bn on R&D by 2024/25 (soon delayed to 2026/27), emphasised “innovation-led growth” in 2021’s ‘Build Back Better’ industrial strategy and revised Innovation Strategy, and committed to establishing the Advanced Research and Invention Agency: a new ‘high risk, high reward’ research group, modelled after America’s DARPA.
Something held over from the 2017 Innovation Strategy was a target to get GERD, which includes both public and private R&D spending, to 2.4% of GDP by 2026/27. A £22bn investment wouldn’t be enough to hit that target, but it could get the ball rolling: economists like Mariana Mazzucato argue that if the government bears the risk of initial investment for innovation, private investors will subsequently ‘crowd in’ with finance of their own.
While the £22bn hasn’t yet fully materialised (there’s no promise like a Boris promise), we’ve hit the 2.4% target thanks to non-publicly funded R&D. In fact, it turns out we’ve been at 2.4% – not 1.7% – for years. Since at least 2018, the Office of National Statistics had been undercounting R&D worth billions from small businesses and higher education.
Mission accomplished? Eh. As a proportion of GDP, our public R&D spending is still only half the OECD average. We’re lagging the world on publicly funded innovation, even as new research finds public R&D leads to more innovative patents and bigger benefits for firms and the public. And Rishi Sunak’s new VISA rules are set to drive away academic researchers in innovative fields like AI even though our universities are British innovation’s saving grace, contributing twice the R&D (as a percentage of GDP) of higher education in the US, Germany and France.
Most public R&D expenditure goes to UKRI, an assembly of nine research organisations. But for four years from 2021, £800m has been earmarked for ARIA’s eight workstreams. CEO Ilan Gur and Chair Matt Clifford are overseeing moonshots like controlling the weather, or building brain interface technology. Clifford played an important role in the AI Safety Summit last November, and ARIA has similarities with the AI Safety Institute. Both cultivate the nimble culture of a startup, and both are precarious, their futures beyond this year’s election uncertain. But the similarities run deeper. ARIA was inspired by DARPA’s mission “to avoid and impose technological surprise” and AISI is on a mission “to minimise surprise” from AI. Most importantly, they’re doing similar research. A former UKRI CEO suggested that ARIA could focus on “developing artificial general intelligence”. Instead, both organisations are researching ways to ensure an AGI that can outsmart humanity doesn’t outmanoeuvre us, with over a dozen employees at AISI working on ‘foundational safety research’ and one ARIA workstream exploring whether mathematical proofs are the secret to building safe AGI.
An ‘advanced research projects agency’ like ARIA was Cummings’s condition for joining Johnson in government. His WhatsApp bio was his creed: “Get Brexit Done, then Arpa”. The plan faced resistance from ex-UKRI leadership, who felt the still-new organisation could take on the ‘high risk, high reward’ agenda, and some reluctance from the Commons Science and Technology Committee, who advised picking one or two research directions lest ARIA spread itself too thin. Labour voted for an amendment to focus ARIA exclusively on net zero technology. Although Cummings is no longer involved and hasn’t so much as spoken to ARIA’s CEO, if Keir Starmer assumes the premiership, he may want to be rid of this strange agency, the brainchild of a deeply unpopular man.
But if AI does, as Joe Biden has predicted, blaze forth more technological progress in the next five to ten years than we’ve seen in the last fifty, Britain will need new institutions to create and capture innovative value. AI is a new method of invention, so we’ll need new mechanisms of innovation governance, like ARIA, to keep up.
We know that innovation, the white heat of technology, matters to Labour. The leadership have met with Mazzucato about her vision of an entrepreneurial state, and their Green Prosperity Plan means investing billions into climate infrastructure. If we invest and innovate, then we don’t have to choose between saving our planet and growing our economy, Starmer believes, because through green growth they will save each other. But we don’t know much about Labour’s innovation strategy. We can assume the Green Prosperity Plan will fund some green technology R&D, but they’ve yet to make commitments on R&D spending — and they’re yet to embrace AI’s transformative role for innovation.
This election year, we’re facing new challenges. Let’s vote for new ideas.