ARIAn Heresy
The UK's new research organization is a step in the right direction, but does it grasp the reason why it is needed?
One recent success for the British boffin has been the setup of the Advanced Research and Invention Agency (ARIA). While attributed to Dominic Cummings, ARIA is part of a more general rush for ARPA-inspired organizations across the low-growth world. This stems from lots of civil servants reading Mariana Mazzucato’s work on state-led innovation from 2017 onwards. Germany and Japan have set up their own organizations. They are less inspired by the actual institutional peculiarities that made the original ARPA and DARPA unique, and more by the thought more government research is generally a good thing. The point of difference for ARIA was that it was being driven by and would be made to the specifications of Dominic Cummings. The hope was he ‘gets’ what really made the initial ARPA successful. But with its most notable institutional champion left out of the agency’s final formation, the revealing of its board raises some scepticism as to whether it can distinguish itself from preexisting conventions and priorities in British science & technology policy.
Ilan Gur, the CEO of ARIA, was formerly a program manager at ARPA-E, an ARPA clone set up in the U.S. in 2007 to specifically tackle advances in the energy sector. Gur’s public pronouncements have described ARIA as a have-a-go organization that is about making the world a better place. While reasonable, it cuts very differently from what drove the original ARPA set up in 1958. There, the goal was very clearly military-technical supremacy for the U.S. over all others. It was ‘winner-takes-all’ in its worldview. If DARPA improved the world, that was at most a byproduct of America’s ability to zap the Red Menace from space.
ARIA’s website offers a series of focus areas for inspiration. These include global AI governance, developing a chatbot for detecting and countering bias, and eliminating livestock methane emissions (a byproduct of bovine belching). These are marketed as suggestions, but they imply ARIA’s internal culture is burdened with the well-worn nostrums of the ‘science and tech’ mainstream. This includes climate change mitigation and getting wrapped up in Yudskowskeite eschatology about killer AI. The emphasis already appears skewed toward technology mitigating negative externalities, rather than the original mission statement of ARPA, which was to prevent strategic surprise to its government and impose strategic surprise on its enemies. If we are going to ban bad AI, there are other organizations to undertake that task. ARIA was not created in a vacuum. Contrary to the thoughts of the free-trade wing of the Tory party, optimistic countries do not vote for Brexit. There is a palpable sense of decline regarding Britain’s economic position, backed up by an evergrowing trove of data. ARIA was not set up to solve climate change or negotiate with AGI. It was established on the pretext of returning Britain specifically to the first cohort of scientific powers.
Profiling the newly announced board of ARIA, we have Matt Clifford, Kate Bingham, and Patrick Vallance, all well thought of. We have the erstwhile David MacMillan, a renowned Nobel-prize-winning chemist. Antonia Jenkinson, the former chief financial officer of the UK Atomic Energy Authority (AEA), is also on the board. The AEA primarily heads research in nuclear fusion but is spectacularly ill-financed to engage in such AOE-style wonder construction. The more active government nuclear body is currently the Nuclear Decommissioning Authority (NDA). Jenkinson has previously been responsible for budgets between £200 - 300 million annually at UK AEA, so ARIA’s £800 million over the rest of the parliamentary term is nothing new. She is experienced in shuffling BEIS money between different stakeholders, so she is a conventional choice for ARIA.
More strange is the inclusion of Stephen Cohen, a civil servant who sits on the British gambling commission. Like it or not, gambling is an industry in which the UK excels, and the inclusion of one of its regulators onto a board tasked with producing “0 to 1” innovation is perplexing. Why is someone primarily known for scrutinizing negative externalities overseeing ‘have a go’ funding for hitherto unthought-of technological wonders? Cohen’s other role is on the civil service commission, where he scrutinizes Whitehall appointments.
Another board member is Sarah Hunter, the director of public policy at X, Google’s ‘moonshot factory’. With Demis Hassabis of Deepmind as an advisor, Alphabet has more of a presence on this board than all of British Industry. Another advisor is Stripe’s Patrick Collison, whose business is primarily in San Francisco and Dublin. The advisory board includes Hayaatun Sillem, the CEO of the Royal academy of engineering. Sillem is not an engineer with commercial experience but rather a biochemist turned PhD in cancer research. Upon completing her 2002 PhD, she entered the royal academy as a policy advisor. Sillem herself has argued improving Britain’s technological standing is indexed to reducing the proportion of white men in STEM.
One might have hoped that the Royal Academy of Engineering had some independent capacity to prioritize things other than modern workplace boilerplate, even with civil servants like Dame Sue Gray sitting as one of the twelve board trustees. The shift from the 1976 mandate of pursuing engineering excellence has been heavily caveated to include sustainability and inclusivity. This is likely to influence ARIA’s hiring and research practices going forward.
ARIA has an overbearing American influence. Its inaugural CEO previously worked at ARPA-E, and he was brought in as a late replacement for the would-of-been CEO Dr. Peter Hingham, who had been the deputy director of the U.S. DARPA. Hingham was the initial choice to lead ARIA but quietly backtracked for unknown personal reasons. In addition, the presence of Alphabet affiliates on both the board and the advisory committee stresses just how America-brained this endeavour is.
ARIA’s remit is expansive, in its mind harking back to the original organization before it became constrained by scrutiny from the Department of Defense in 1972. But given its apparent fixation on tackling climate change and cognitive bias, overseen by a board that lacks much in the way of institutional outsiders, it may not achieve either the anarchic inventiveness of the original ARPA or the military sorcery of the post-1972 DARPA. It risks being an adjunct to the established research base that turns additional funding into diminishing returns.
I wish the organization all the best. Perhaps its budgetary ringfencing and safety from freedom of information requests can allow it to pursue truly risky programs. But at least at the outset, this does not appear to be a unique organization that can think beyond the imagination of Google or the conservatism of the civil service. The original ARPA was set up because scientists, the U.S. secretary of defence, and President Eisenhower in particular, had had enough of wasteful and parochial generals being constantly behind the curve. ARIA should remember this as it considers its relationship with prevailing orthodoxies.