How to Build the British ARPA

Today’s interviewee, James Phillips, was formerly the science and tech adviser to Prime Minister Boris Johnson. An acclaimed systems neuroscientist, Phillips helped develop the UK’s rapid COVID testing and helped create the Advanced Research and Invention Agency (ARIA).

• Dominic Cummings’ band of “weirdos and misfits”

• Red-teaming Westminster

• Why you should always be willing to resign

• The problem with the British civil service

• Protecting ARIA from mission creep

• Whether the UK can end economic stagnation

I’d previously been in neuroscience. During my PhD, I’d become very interested in how science might be more effective and how we might change how we organize and fund scientists to get better results. This is now broadly known as “applied metascience.” In January 2020, Dominic Cummings, who was Prime Minister Boris Johnson’s Chief Adviser, put out a blog post to try to bring in people from outside government who wouldn’t usually get into government. The idea was to diversify the range of backgrounds, experiences, and skillsets that would be around the Prime Minister. And I was one of those who was brought in. 

I ended up walking through the door of Downing Street on April 15th, three days after the Prime Minister left the hospital and midway through the first lockdown. It was a surreal time to enter government. I’d come in to look at how to improve science, but the pandemic ended up being a major part of the first year, at least. 

I’d known Cummings a little bit before. We’d been talking with a network of people who were interested in the early ARPA and things like Bell Labs to see what we might learn from different approaches to science and technology. And then this opening came up and we went from that. 

Yes, I’m a weirdo and a misfit. Actually, in my favorite ever Guardian article, the introductory line is, “James Phillips is a weirdo and a misfit.” I think it was meant kindly, because it went on to be complimentary. But yeah, it was a wonderful group of people that came into Number 10, and most of us had absolutely no background in politics. It was entirely new to us. 

Part of what was challenging was the very reason we were there — there’s a recent report from the COVID inquiry that says there was a lot of groupthink in the early response to the pandemic in the UK. And I think this has been mirrored across some other countries. Part of the reason we were brought in was to break that groupthink. The fact that we really didn’t have any scars or assumptions about what was possible inside the bureaucracy meant we could see things in a new way. I walked through the door, got a security briefing, a phone, and a laptop, and was essentially told to get on with trying to make things better. I’d actually been due to go in earlier, but the HR systems and most of the government operations had essentially collapsed because they didn’t have contingency plans for this kind of event. 

I got a phone call from Ben Warner, who was central to some of the first lockdown. He said, “James, before you come in, can you phone around and try to find things that we’re missing?” So, essentially, “Can you red-team the government’s COVID response?” I spent about 10 days in my parents’ garden just phoning around the UK, America, and all the networks of people I knew, asking, “Can you find me five other people who think differently and might have some unusual ideas?” That was my first immersion into this world. 

It’s amazing what you can find out if you simply phone someone and tell them you’re going to join Number 10 in a few days. People are perhaps overly trusting, but that surfaced a range of things about the need to do rapid mass testing of whole populations, things around vaccine procurement, things about the origins of COVID. The government lacked a system for gathering unusual ideas at that time, and that was one of the first things that our team was able to do. 

Given that I went in at such an unusual historical moment, it’s difficult to draw general lessons from that start period. I joined in the middle of a lockdown, with the Prime Minister away at his country house in Chequers due to being ill.

But I would give two pieces of advice. First of all, you cannot get anything done in government by yourself. You have to have a network and a team. Many special advisers come in as soloists, and I wish I had been able to forge more links around the system in those early stages, which the pandemic made very difficult to do. Those personal relations are incredibly important for getting things done.

The second is, always be willing to resign. Don’t go into one of these roles and think you have to stay there for the whole duration, because you will become risk-averse. Our view was that we were in there to get certain things accomplished. If it became impossible, we were perfectly willing to just go back into science. 

Someone very senior in the Sunak administration put it this way: act as if you’ve already been fired. Special Advisers don’t have any job security, they could be dismissed without reason overnight. You want to set things up so that if you suddenly vanish, things will carry on without you. 

I was technically in the Policy Unit. This body dates back to 1974, and usually contains six or seven of the prime minister’s leads. So you have your lead adviser for health, for defense, and I was in there for science and technology. I think I was the first one they’d had. I reported to Munira Mirza, who was a brilliant head of the Policy Unit and reported directly to the Prime Minister. I also worked very closely with Dominic Cummings and the secretary of state in the department. I actually had a joint role, so I also held a new role “the Special Adviser for Science and Technology to the Secretary of State.” I was basically given a lot of leeway to find things that would be useful and push on them. That was a very distinctive feature of that time in Number 10, that we had these people who didn’t really have a specific brief, but could look for opportunities and then push them forward. 

I thought that was a positive addition, especially in a crisis, where you don’t really know what it is you should be looking for. You can’t give out instructions saying, “Can you go and find out this?” because you don’t know that your testing strategy might be wrong, for example. So you need these people whose job is essentially to look around corners. I would have meetings with all sorts of different people, from the Prime Minister, the Secretary of State, the Chief Scientific Advisers, all the way through to Cambridge PhD students, who did some of the crucial work demonstrating that a large-scale rapid testing strategy could work on our COVID response. 

The British system is quite different from the American system in a number of ways. In terms of the day-to-day functioning of the government, almost everyone inside the government is not a political appointee. So most of the secretaries and people who are now advising Keir Starmer are the department-specific private secretaries, and they don’t change. The civil service is separate from the political side. 

Now, that has many strengths, and the impartiality of the civil service is a long-standing principle. But it also means that if, say, a prime minister wants a specific set of people advising them on issues, they don’t have that freedom. That applies both on the media side and on the policy side. 

There are two types of spads, or special advisers. The media spads tend to be a bit more political. They’re much more engaged in moment-to-moment communications of the government, whereas policy advisers are, as the name suggests, more involved in the formulation of priorities and policies. In my case, I didn’t really ever engage in anything that you would think of as politics. I certainly wasn’t brought in as a political person in that sense, but these are people who are appointed by the politicians rather than by the official system, and that’s the key distinction. Now, over time, the role has become more political — maybe these people will have worked in the Labour Party or the Conservative Party before. But my understanding is that the Policy Unit was created to give a politician the ability to have an adviser near them who they had personally selected.

And that system has evolved. It’s a very flexible role and any given secretary of state can have between one and three special advisers. It depends on how you count it, but the prime minister has about seven to eight policy special advisers in Downing Street. Then there can be any number in other roles, like media or business relations or appointments. The appointment can be very flexible, rapidly hired — there isn’t much of a process to it. Spads review and advise and essentially see everything that the ministers and the prime ministers see on issues.

I was in a somewhat unusual role, in that I think there was recognition on the official side that I had come in for my technical expertise, rather than due to any past political involvement. The media picked up a narrative throughout this era that there was a lot of tension between the political side and the official side. My own experience was that there wasn’t much of a boundary between many of the best officials and the best special advisers, especially during the COVID response.

I think there is more tension between Number 10 and, say, the Treasury, or between the Treasury and the Departments — the tensions arise in the departmental struggles. Now, that’s not to say that it’s all rosy everywhere, but my impression was that the officials were very happy to work with the people who had been brought in as domain experts. 

Yes. I mean, Tony Blair wrote about this in his memoir and has spoken of regret about not going further on civil service reform. I think it’s always important for people like me who’ve experienced this to be quite careful in how we speak about it, because it can easily sound like we’re saying the civil services are rubbish. Munira was quite precise in what she said there, that there are many brilliant people in the civil service, many not-so-brilliant people. Instead of saying, “Let’s do more stakeholder consultations, let’s do more reviews,” the brilliant people say, “Actually, it’s pretty obvious what the answer is. Let’s just get on and do it.” 

Of course, this is oversimplifying somewhat. But my experience was that there was a general selection pressure against those doers. And it’s not merely Munira who said that. Tony has said it. Kate Bingham, who was brought in to lead the Vaccine Taskforce, was also very critical of some aspects of the civil service. 

I think that there’s a broad recognition, including in the senior civil service, that some of the well-meaning processes that have been brought in to protect civil servants have sometimes ended up going a bit too far and just made it very difficult to deal with that poor performance. 

I would say that it was generally at the higher levels in the civil service where this was a problem, partly because it’s easier to choose the great junior civil servants. And so many civil servants went way beyond the line of duty, making absolutely fantastic contributions during the response. 

But let’s say you’ve got a role where you need someone with a crisis-management mindset, maybe someone ex-military or in the police forces who’s used to dealing with rapidly changing, highly uncertain circumstances where the stakes are very high. I would see roles like that being filled by people who had spent their entire career in the tax and spend sections of, say, the treasury or finance department, with little other experience.

And it’s not the fault of that person. It’s that the civil service doesn’t have enough of those sort of people on the crisis response side, and just generally tends to shuffle the same people into those positions. 

What we ended up with was a meeting where the people who were actually getting things done were often sat at home, because we were working from home a lot, whilst the people who went into the meetings with the Prime Minister were often just reading off a script that someone else had written for them. I think people who have worked in bureaucracies will recognize this phenomenon, and COVID shone a particularly big light on that. 

Yes, in many different ways. One example is that the status quo testing strategy going into COVID, both in the UK and elsewhere, was that PCR-based test and trace was the best way to control the pandemic. In fact, the regulations in the UK made it so that you actually couldn’t use any other kind of test; they said every test that’s used must be PCR-level or better. 

PCR testing had low rates of false negatives. It’s not missing many positive cases. The number that was given in government was 99% (which PCR doesn’t actually provide in practice because of sample-handling issues, poor swabbing, all that kind of stuff). But that was the number. 

And during the red-teaming process I led, we found a network of people who said, actually, the speed and scale of testing matters just as much as the sensitivity — if you test one person in the country with a 99%-sensitive test, that’s clearly nowhere near as useful as testing everyone in the country with a 50%-sensitive test. So the scale and also the speed, which I didn’t use in that example, are crucial parameters. 

However, once we’d seen this in April, it took four or five months just to get the system to recognize it. The people in the bureaucratic positions who could change this weren’t really scientifically literate in the necessary ways. Some of the examples that I gave you just didn’t register. They would simply say, in effect, “Look at this group of weirdos and misfits in Number 10.” Then they’d look at the official advice from these major international bodies and say, “Well, the official advice says that it’s got to be 99%-sensitive. They just must be some crazy people who don’t know what they’re talking about.” They couldn’t exercise the technical judgment to say, “Actually, these people have got a point.”

That situation only changed after a panicked moment. I was stopped in the hallway in Number 10 and asked, “James, how worried are you about this winter?” I said, “I think it could be much worse than what’s happened just now and we could be in lockdown for several months.” I was asked what else we could do, and I said, “Well, I need you guys to summon all the key people together and ask what is wrong with this argument, that we can use lots more types of tests.” In that meeting, that decision was overturned and then things changed. But that could have happened back in April 2020. For UK-based readers, these are the “lateral flow tests” you can do at home.

This happened in July 2020. By that point we only had two months, really, to get this thing in order. And there were a whole load of other obstacles that came about for similar reasons, because the system didn’t have the people to recognize the skills you need to hire in. 

If the people running certain programs don’t themselves have an understanding of the issue, they can’t say, “Well, clearly we need a modeling expert who has these three characteristics and we need a procurement expert with these characteristics and we need this person to do the comms.” So you have this cascading failure. If you don’t have the right skill set at the senior levels, you can’t get it at the junior levels either. 

We’d been going around the system for three months and writing notes and essentially couldn’t find anyone who could shift this thing, so I was asked by the Prime Minister’s main adviser, Dominic Cummings, to write out an argument for this. I literally spent a Sunday afternoon writing a 20-page document explaining the problem, the blocks in the system, and what needed to be refuted for us to be proven wrong. I got it reviewed by several experts.

Cummings then called a meeting with the Chief Scientific Adviser Patrick Vallance, the Deputy Chief Medical Officer John Van-Tam, directors of the Wellcome Trust, and several others. Essentially everyone on that call said, “Oh, this is right. We should be doing this.” The Prime Minister’s Private Secretary for Health Imran Shafi was there and said, “Okay, clearly we’ve made a mistake,” and organized a letter explaining this to the Prime Minister. 

I, Imran, and another Special Adviser briefed the Prime Minister and explained this additional way in which you could deal with a pandemic, which was to use scale of test and speed of test, along with specificity of test. And then instructions were given to the system to pursue this. 

In my opinion, it didn’t work out in the way that it should have, but nonetheless, we still got schools and hospitals and care homes, which were a particularly vulnerable point in the first wave, protected with this extra layer of tests. Later on during Omicron, we just sent everyone these tests in the post. Now, I think that could have happened much earlier, but still, progress was made. 

There are well-documented challenges with the Test and Trace system that existed at the time. There was an incredibly able person, a former Commanding Officer of the Special Air Service, who was brought in to basically fix testing and did an absolutely brilliant job, but there wasn’t someone else like him who we could put in charge. So the rapid testing was done under the same bureaucracy. This wasn’t a decision that I was consulted on, but the main PCR-testing bureaucracy would also deal with the rapid testing approach rather than doing them as separate entities, which I think would have been the better way of doing it.

Secondly, there wasn’t a clear communication of the rationale for this. There are many, many points at which I think that could have been done differently. But essentially the failure mode here is that, in an ideal world, all the people who had been pushing this for several months and had done the modeling, which we published online in April 2020, would have been brought in to now work on this new mission. They would be doing the modeling and briefing. They would obviously have oversight and a red team criticizing them, but you’d have people who understood the objective and the arguments managing the system. 

Instead, all the people who had essentially been blockers on this were put in charge. There were literally months of conversation about, “Why on earth are we doing this?” It was really only when Sir John Bell, the Regis Professor of Medicine at Oxford, got on board and began pushing it that things started to move. 

The lesson from that is when you want to do something outside of the Overton window or the consensus at the time, you really need to hire a bespoke team to do it. Otherwise, you’re going to spend forever explaining to everyone why there’s been this sudden pivot. At one point I literally had to offer to pay people out of my own salary due to the ludicrous HR rules that were blocking the right expertise from being able to work full-time on COVID.

Yeah, you’ve made a very good point. In setting up ARIA, which is like a UK ARPA, we essentially had a live demo. In an ideal world, if there’s another crisis, something like an ARPA would be able to respond in a very agile and fast way. One of the first things I raised to Number 10 was the need for some kind of COVID ARPA because the existing systems were so slow. In fact, one of the two main clinical trials for COVID ran out of patients during the first wave, and only began administering drugs to patients in June, over 2 months after lockdown began and 6 months after COVID was first detected. There are many other examples. 

We were essentially getting a live tutorial and demonstration of how our bureaucracies were paralyzed by the web of regulations that had built up around them. So, yes, that was extremely, extremely relevant. We had this test of, “Let’s say ARIA was dealing with COVID. Will it be able to do this thing? Will it be able to just buy these tests, assemble the right experts and oversight, and run a demo of them really quickly?” That was one silver lining of the whole COVID nightmare. 

And, by the way, it wasn’t really our idea to set up a UK ARPA — it had been suggested for many decades. One difference between the team that came in to help set up ARIA and the many other government teams that had tried before was that we had been looking at this for several years and personally knew some of the people who had been there in the Early Advanced Research Projects Agency before it became DARPA, people like Alan Kay, who reached out by commenting on an article I’d co-written in 2018. 

So we had all these stories from them of how they got things done. One of them is the story of how the ARPANET, the precursor to the Internet, got funded. It was literally a 20-minute conversation between Director Charles Herzfeld and the program manager, Larry Roberts. They signed off something like $1 million in 20 minutes. Now that’s completely unheard of inside a government bureaucracy — such a decision can easily take years. So we had those examples and went around looking at the legal setups of different agencies and saying, “Right, we need to remove that one,” and “We need to add this regulation.” 

The origins of it politically are part of the whole chaos around Brexit in 2019, where Dom and others were asked to go in and help sort out that situation. One of his four demands to go in was that a UK ARPA would be created; another was that there would be a doubling of the science budget. 

They put that into the manifesto for the 2019 election, and that was a crucial step. One of the slight frustrations I see with the recent UK election is that the manifestos weren’t really used to put in any bold commitments around science and technology. I think that was a mistake on both sides because if it’s in the manifesto, it’s a huge help pushing it through the system. If people object, you can just say, “Well, the country has voted for this.” 

Yeah. So I actually looked this up recently and I read that the parties do have platforms in America, but when I lived in America for several years, I didn’t see that mechanism of the manifesto in the way that we have it in the UK. The commitment was for £800 million for a high-risk, high-reward research agency — that was essentially what it said, word for word. Almost every other government process is about minimizing risk, which can too often mean minimizing reward. 

That “high-risk, high-reward” line was incredibly useful and meant that in any meeting where someone would say, “You know, this is very risky,” we would say, “Well, the manifesto says ‘high-risk, high-reward.’” That became a tool that we could use again and again to drive this thing through. 

We had these twice-weekly meetings in the Cabinet Room with me, a couple of others from the science and tech group across Whitehall, Dominic Cummings, the Chief Scientific Adviser, and the head of UKRI. There were also several officials across seniorities, including a brilliant junior official called Zak Lawton. 

There’s a great quote from Colonel Boyd, “People, ideas, machines — in that order!” Boyd was one of the reformers in the Pentagon in the 1970s, and he’s famous for the OODA loop concept of how fighter jets should engage in combat. That quote contrasts with how almost everything else in government is set up; it’s machines, ideas, and people, in that order. In the UK government, you’ve got your processes, or machines: consultations, reviews, and setting up a bureaucracy. Then once you have this set of tools, you say, “Okay, what’s the idea?” “Well, we’re going to look at net zero,” or, “We’re going to look at some issue in tax.” And then at the end of that, once you’ve got this relatively developed plan, you go out and find the person to run it. 

In setting up ARIA, we did a complete inversion of that. We said, “We will get in an exceptional leadership team and let them develop the ideas and set up the machines.” We didn’t specify how many programs they were going to have, how many program directors to hire, how they should do their HR, or how they should set up procurement. That manifests in the business case for ARIA. 

Most things in government have a business case that specifies in minute detail, years ahead, what an organization is going to do. ARIA essentially has an empty business case and the responsibility for managing its money is primarily with the board and Chair, not with Whitehall and the Westminster machine. That’s a slight oversimplification, but the basic premise is that the government appoints a board to oversee the agency, rather than the usual process of extensive email chains back and forth between the bureaucracies. 

The reason for saying that is that if I’m giving a running commentary on it, but at the same time saying to everyone else, “Give it 10 years,” it’s kind of a contradiction, right? So I thought it would be best to just take a step back. But Ilan Gur and Matt Clifford and others come to me from time to time and ask for input, and I saw Ilan a couple of weeks ago. 

In terms of how to evaluate ARIA’s success, we specifically added a clause in the primary legislation that the Secretary of State can close this agency down after 10 years without any legislation. He or she can just sign a piece of paper and it will disappear. I think the Secretary of State should come to that decision by getting in people like former DARPA directors, world-class technologists like Demis Hassabis, maybe some people in the metascience space like Ben Reinhardt, Adam Marblestone, and Eirini Malliarki — the people who have got some experience of trying to do new things — and get them to go in to evaluate how it’s going and write a report.

The default is to hire management consultants, and my expectation is that that mindset will come in and look for KPIs and processes. They’ll look for pieces of bureaucracy that indicate this is a well-managed organization. But an ARPA agency is really much more about, “Are they getting good people? Do those people have the freedom to take risky bets? Is the agency able to adequately calibrate that risk?” The focus is on people more than paperwork.

I think the biggest outstanding challenge for ARIA is that it’s got a very small budget relative to the scale of the mission that’s been given; it will be at £400 million a year by next year, and that’s roughly one DARPA office. Now, ARIA has a much broader remit than DARPA. So I think the biggest risk or failure mode is that we look back in 10 years and we’re not sure if it’s because it wasn’t able to take enough bets. While I think it will be relatively easy for experts like a former DARPA director to go in and see if it’s a good idea. I’m not sure how the public side of it will work given this constraint.

If ARIA is not doing things differently from the existing system, there was absolutely no point in me and others spending three years of our lives helping to set it up. It’s looking to do things that cannot be or are unlikely to be done by the existing system. Many of the criticisms you cite are well-motivated, but they’re essentially saying, to my ears, that we should be looking at ARIA in the same way we look at all the other agencies. 

You’ll see in the same newspaper or journal, “Oh, we’re very worried about how bureaucratic and inflexible our funders are, and how they don’t seem to be able to move quickly,” without reflecting that the mechanisms that ARIA has can help other bureaucracies solve these problems, and that the constraints those organizations face, that ARIA is largely free from, may be the origin of their own critique.

The second point, and a more important one which we used at the time, is that if you had set up ARPA when Sputnik went up in 1957 and given it a mission and a set of metrics, I would be absolutely astonished if anyone in DC had said, “You know that the criteria here is to create the personal computing industry, right?” There’s just no way you could foresee that. One of the criticisms that Labour had of ARIA was that they wanted it to be given a mission. Our point was that if we knew what that mission should be, there would be little point setting up ARIA. We can do that through UK Research and Innovation (UKRI), the main public funder. ARIA’s job was to go and look for the very things we didn’t know to task it with. 

And again, that’s very, very unusual in government — all of the pressures are, “What is your strategic priority? How does this fit into the government’s agenda?” And we were explicitly saying this thing doesn’t fit into any of those things. It’s essentially a red team to spot missed opportunities to try and do things differently. 

One response I’d give is that we have a lot of rules in the UK system around value for money and oversight processes. I would often be in these discussions where we’d have some set of rules to ensure that public money is used well, and we’d sit around the table and realize that this rule is going to greatly damage the value for money of this investment. And then the response would be, “Yes, we all understand that, but it’s important we have processes to ensure that value for money is maintained.” It’s Potemkin-like: if you’re called before the select committee, you could demonstrate that you’d gone through the proper process. 

Conventional oversight can sometimes actually be quite harmful for the real value for money in return to the taxpayer. A simple example of this: having to explain what you plan on researching to a funder years in advance increases the “legibility” of your research to external oversight, but restricts freedom to adapt to circumstances to do effective research.

Now, in terms of how I would explain this to someone in my local community, I would say that things like the mRNA vaccine don’t tend to come about in competitive free markets on their own, because they require a push at a much earlier stage than people are able to take profits from. The mRNA vaccine for COVID actually came, in part, from DARPA’s early investment. We’ve just seen an example of how effective this kind of funding is, and there’s a long list of amazing things that science funding in the UK has done. And selling that science funding is actually not so difficult. 

I think that people are receptive to the bigger issue, it’s simply that science funding just doesn’t poll highly enough. If you poll people and ask, “Do you think we should fund science?” they’ll say “Yes, of course, we should fund science.” But it’s number 9 or 10 on the list beneath the National Health Service, defense, law and order, and all those kinds of things. 

I think ARIA will be fine for the next five years. One of the key people in government who was incredibly helpful for us in setting this up and was on the interview panels for the ARIA leadership, Sir Patrick Vallance, who was also instrumental in setting up the Vaccine TaskForce, was appointed to the board of ARIA after he left government. He was very keen to come in and help because he was a big believer in it. He’s actually left the ARIA board and is now the Minister of State for Science under Labour. So that’s a very good signal. If you look at reports from across the think tank space in London, there’s overwhelming support for the kinds of motivations that led to ARIA, and, we’re very lucky that the space of science and technology hasn’t really become politically polarized in the UK.

The biggest risk comes from that entropic effect where things creep back in over time, because there’s no political will to say, “No.” Joshua Elliot spoke on your podcast about how a DARPA director sometimes has to come in and just sweep away a load of process that has built up. I don’t want to use the word “weeding,” because a lot of this is well-intentioned bureaucracy, but bureaucracy is, in some sense, the accumulation of good ideas, and that will be a danger and that needs regular attention.

Productivity in the UK has been stagnant since 2008, so we’re coming up on the 20th anniversary of any improvements in that situation. Your colleague Alec Stapp, posts these very helpful, motivational, tweets where he compares UK and US productivity, where it’s just totally flat in the UK and he says how well the US is doing. I always send these around our WhatsApp groups to motivate us to do better, so thank you to Alec for that.

Precisely. Tony Blair put this very well in his speech at the recent Future of Britain 2024 conference, that we need a short-term plan and a long-term plan. The short-term plan has to be reforming planning. It can be very, very slow and difficult to build something in the UK: between planning permission, environment review, and various other stages, it takes 15 to 20 years to start digging to build a short 20-mile stretch of road around Newcastle. There’s enormous amounts of latent potential in that system, but we’ll need regulatory changes for short-term economic growth.

Longer term, the R&D agenda is the only game in town for the UK, because we don’t have natural resources or vast amounts of land in the way that other places do. Given our history, the wonderful universities that we have, and one of the best startup environments outside of Boston and Silicon Valley, this is an open door for a government, but it will take some time for that to come through.

I spent several years just outside DC as part of my PhD. One of the standout differences is that the UK system is much more concentrated in universities than the United States and most other countries — Germany has the Max Planck Society and the Fraunhofer Society, and the United States has the national labs, the tech giants, and relatively higher private R&D, whereas a lot more of the UK goes through universities. Paul Nurse, Nobel Prize winner and the founder of the Crick Institute in London, did a review for us and found that in the UK, roughly 80% of non-business R&D goes through universities. In comparable countries, it’s somewhere between 45-60%. 

One of the things we really wanted to try and change as part of an uplift in the R&D budget was to add in new types of research institutions, learning from projects like Xerox PARC, Bell Labs, and DeepMind. I think that’s necessary because no one set of incentives and career structures suit every kind of problem. In the UK, even many of her best supporters will admit that Margaret Thatcher went too far in cutting the R&D budget, which concentrated research into universities and created a kind of monoculture.

So first of all, pretty much the best years of my life were spent in Oxford and Cambridge. I think they’re wonderful institutions and I would advocate increasing their government support, not the other way around. But there are a couple of points here that don’t get enough airtime. The first is that there’s a narrative that the UK is brilliant at early-stage research, and it has a problem spinning that out. Yes, there are very good universities in the UK, but our analysis, later reported in the Financial Times and Times Higher Educational Supplement, shows that if you remove things like the Cambridge Laboratory of Molecular Biology and DeepMind, the UK’s performance drops quite dramatically. These adjacent organizations prop up the UK’s performance, and they don’t look much like anything else we’ve got. 

The commonly cited number is that the UK has 13% of the top 1% of most-cited papers. We looked at the top 100 most-cited papers over 10 years, which is a much more selective measure. Now, that measure has all sorts of problems, but we found that that percentage actually drops down to about 3-5%. To me, that points to a system that has been optimized for producing lots of pretty good papers, but perhaps not giving researchers the freedom to go and produce those really big advances. This is a problem around the world.

The second point is, yes, those top universities are still outstanding, but some of the things that give them their outstanding status are concentrated in studies, and not in technology development. For example, doing a large-scale public health study and finding that X dietary intervention is helpful for diabetes is one type of study, another is, “We are going to find a way of engineering an improved CRISPR.” Those are both very valuable types of research, but one of those is much more likely to directly benefit your economy, in a way that the value can be captured commercially. Our analysis is far from definitive, but it suggests that the UK has a slight imbalance there. 

There’s a great quote from Sydney Brenner: “Progress in science depends on new techniques, new discoveries, and new ideas, probably in that order.” The premise of a lot of the UK’s mindset is that science proceeds through new ideas, new discoveries, and new methods, usually in that order. So the UK is very predicated on a linear model from some brilliant theoretical idea to a discovery to a method. 

What I saw in the United States, and you see it now with places like the Arc Institute and Arcadia Science, is that there can be more emphasis on blue skies invention. It’s not an invention with a very specific goal, but more akin to what Jason Chin is doing at the Cambridge Laboratory for Molecular Biology. They’re trying to create some new fundamental capability, and my impression is that there needs to be a bit of a pivot toward that kind of research. And that’s not the same as discovery versus applied. It’s about whether or not there’s an engineering element involved. 

I’m wary of a false dichotomy between universities and institutes. The Francis Crick Institute in London has strong links to the surrounding universities, but is independent in terms of its governance. I think adding in those kinds of entities can be extremely powerful in strengthening the universities and giving core support and resources to the scientists so that they don’t need to have a lab of 50 people to do their experiments. 

Yeah, I’d totally co-sign. At a conference I was at recently, someone said they found this incredible opportunity: an undeveloped country that speaks English, and it’s called England. If you go to someone who’s doing a startup in Cambridge and say, “What’s your biggest problem?” they will usually say, “It’s so expensive and difficult to get lab space because they just won’t build.” The solutions to these things ultimately have to come politically and I’m very, very optimistic that they will. The big upside of things getting bad is that it makes it easier to do the bold things you need to make them good, and I think we’re clearly there. 

Well, Liz Truss had a go in her cameo as Prime Minister, and thought that it was all due to too-high taxes. I don’t have a view on where you should come out on that issue, but my reading and that of many others is that low-tax environments alone are not enough to get deep tech environments going. You see this in the origins of Boston, the Bay Area, and Cambridge in the UK with the post-war investments in the molecular biology revolution. Again and again, you see early philanthropic and public investments by the state seeding these broader ecosystems. The bigger debate is, do you simply invest more in the things you’ve already got, or do you try to reform them and learn from places like DeepMind?

Yes — there’s a plight for junior scientists in that the global system has become very gerontocratic. NIH data shows that there’s been a great rise in the age of independent investigators. I think the UK could address that and has a niche opportunity there. 

Global science policy is essentially set by committees of very senior people. Even if they have good will independently, as a group, they have a shared interest in power for more senior people. Sydney Brenner had another quote that always affected me. I knew him very briefly when I was in America. He said, “I strongly believe that the only way to encourage innovation is to give it to the young. The young have a great advantage in that they are ignorant. Because I think ignorance in science is very important. If you’re like me and you know too much you can’t try new things. I always work in fields of which I’m totally ignorant. Today the Americans have developed a new culture in science based on the slavery of graduate students.” 

One of the things I was trying to push, and I didn’t succeed as much as I would have liked, was to pivot the system back toward giving people independence at a much earlier age. In the US, you’ve got someone like Jake Sullivan, who’s very public-facing as the National Security Advisor. We don’t really have that in the UK. It’s more Special Advisers and MPs, but the MPs are not really podcaster types, though this is now changing. 

Cummings’ belief in young people was unusual, and I was relatively young for the amount of authority I had. It seemed to go right, I didn’t get any scandals, and I wasn’t at the parties.

Thank you to Rita Sokolova for her judicious edits.