“It seems to me that this plan could easily be mad,” Marc Warner, CEO of the AI consultancy Faculty, told Dominic Cummings, according to the former Downing Street adviser’s parliamentary testimony. The plan, to pursue herd immunity and avoid locking down, “could be incredibly destructive. Has this really been tested? Have you really thought it all through? Should I and some others start thinking about a plan B?”
Less than a fortnight later, after consulting Demis Hassabis, CEO of the AI start-up DeepMind, and the mathematician Timothy Gowers, Cummings invited Warner, Warner’s brother Ben (a Downing Street data scientist), and the Prime Minister’s private secretary for health into Boris Johnson’s study to discuss an alternative pandemic approach. Together, the foursome sketched out plan B on a whiteboard.
Ben Warner and Cummings met Johnson the next morning. “We showed him this graph. It has here ‘no mitigation’, ‘our plan’. This is all the NHS broken,” said Cummings, gesturing to the original plan. “And here is an actual plan, which is lockdown.”
Nine days later, on 23 March, the government announced the country’s first lockdown. “It was a totally, totally crazy time,” Marc Warner tells me when we speak via video call in July. “It’s seven days a week, incredible hours, but everyone really working together trying to get this stuff sorted. And so we could look at the data and see where things were going.”
In his select committee appearance in May, Cummings described Warner as “one of the smartest and most ethical people I have ever met in my life”. He said that if he, Cummings, had been prime minister, he would have placed Warner “in charge of this whole thing”. In Cummings’s world, Warner would have had “as close to kingly authority as the state has legally to do stuff, and pushing the barriers of legality”. He would have been “in charge of everyone, and he can fire anybody, he can move anybody, and he can jiggle the whole thing around”.
Warner rolls his eyes as I remind him of this. “Again, we just try and do our best. There’s not really much to it. The customer has a problem. We try and help them fix it to the best of our ability. I don’t think there’s much more to it than that.”
Nevertheless, Cummings was unambiguous about Warner’s role in averting an even larger disaster than the UK ultimately endured. “Without him, thousands of people would be dead,” he told MPs, “and I know that his involvement in it has come at a lot of personal cost for himself and for his company.”
Raised in Bedford, Warner, 36, had what he calls a “very gentle childhood”. He went on holiday to seaside resorts such as Poole and Southwold and attended three schools each less than three miles from his family home. Warner’s mother was a PE teacher and his dad an accountant, but it was his grandparents who inspired his passion for science: his grandmother taught chemistry and his grandfather taught physics.
“They definitely encouraged my interest in science in general and I remember doing some night experiments with pH as a kid. I don’t know how much I understood it, but I certainly liked that different liquids could change the colour of the paper.”
[See also: Join New Statesman and Tech Monitor to hear leading public sector IT professionals discuss digital transformation themes across central government, local administration and healthcare bodies in the UK]
Warner says it wasn’t that he “knew young that I wanted to be a physicist”, but that he had “a physics teacher who can help you out with a bunch of things at home so you do better in it, and so I suppose it’s a somewhat self-fulfilling prophecy”.
Warner moved to London at 18 to study physics at Imperial College London. After graduating, he specialised in quantum computing, completing a PhD at the London Centre for Nanotechnology before taking up a research fellowship at Harvard.
It was there, in the early 2010s, that Warner began to research artificial intelligence. “I started to get interested in AI through the kind of AI safety angle, and people like Nick Bostrom, who were writing very, very early – before he wrote his book Super Intelligence.
“I thought that was interesting and important, and understudied, but I thought AI was magical. I had a PhD in physics but my assumption was AI was going to be too hard, and the maths would be too complicated,” he says.
“It was only when I actually sat down with a textbook, when I was doing my research fellowship at Harvard, and started looking through it and I thought, although it sounds magical, actually it’s similar to the maths we do in physics. In some sense it’s actually easier to master.”
In 2014 Warner faced a decision: continue pursuing quantum computing research or move into the AI sector. “I stumbled across these programmes in the US that helped academic PhDs become commercial data scientists. I thought that that has got to be a good thing for the world. We should just set that up while I’m thinking about these bigger picture things.
“And then basically, the truth is, I just enjoyed working with my friends,” says Warner. He met Andrew Brookes, now Faculty’s chief technology officer, at secondary school and Angie Ma, its chief people officer, while studying for his PhD. “When I was originally thinking, well, we’ll set this up and then I’ll make a big decision, it basically came down to loyalty to friends.”
In August 2015, Warner, Ma and Brookes launched ASI Data Science, Faculty’s original moniker. The company developed a fellowship programme for recent PhD graduates to carry out data science secondments with Faculty’s clients, using advanced modelling to solve complex business problems. The company has built a terrorism propaganda detection tool for the Home Office and analysis to help the UK coastguard distribute resources more effectively.
With 5 to 10 per cent of UK-based maths, physics and engineering PhD students applying to take part each year, places on the programme are now fiercely contested; just one in ten applicants are accepted. The programme has thereby given Faculty the opportunity to select the very brightest data scientists and then employ them in-house.
“The Faculty engineers and data scientists were really amongst the best we’ve ever worked with,” said one source who worked with Warner and his team at the start of the pandemic. “I’d say the most impressive thing… is the quality of the talent that they’ve managed to recruit.”
Not only are they “very, very brilliant engineers and data scientists, but they have a kind of grit that a lot of people who are very talented in that way [lack]. Faculty has somehow kind of creamed off the people who combine those two things.”
Warner possesses these qualities too, the source said. “He’s incredibly bright, he’s got a fantastic temperament, he’s a very level-headed, laid-back person, who’s got that ability to talk to CEOs and someone who’s totally operational as if they were the same ranking individual. He’s the same person with everyone.”
It was Warner’s temperament that appears to have converted Cummings into one of his most committed acolytes. “One is definitely a fan of the other; there is some form of friendship,” said the source. “But I don’t know how close they really are, or where that dates from. With Dom Cummings you never really know because he admires people like Marc in just such a sort of demented way.”
Reflecting on Cummings’s claim that Warner was one of the most ethical people he’d ever met, the source added: “Cummings really is big on his superlatives, I mean people are the most ethical people in the world or they’re the most fucking hopeless or they’re the most corrupt. There are very few shades of grey.”
Last March, Faculty was enlisted by Downing Street to build a Covid-19 data dashboard for the NHS and government officials. Developed in partnership with Peter Thiel’s data engineering firm Palantir, the dashboard analysed how much stress Covid-19 was placing on hospitals across the country. Over time, it also predicted demand on NHS services, allowing managers to distribute resources such as PPE, ventilators and beds accordingly.
A number of the world’s largest consultancy firms, including McKinsey and Accenture, have also built models for the government during the pandemic. But the source said Faculty’s were the most useful: “The so-called early warning system that [Faculty] developed for NHS England was a tremendous, phenomenal bit of work.”
However, despite the utility of Faculty’s work, the company’s links to Cummings have attracted much scrutiny from campaigners. Cummings’s private consultancy company paid Faculty £260,000 over the course of two years in exchange for work that has not been explained. Faculty’s data scientists had previously worked to build political models for Vote Leave and were praised in one of Cummings’s blogs.
“Physicists and mathematicians regularly invade other fields,” the adviser wrote, “but other fields do not invade theirs so we can see which fields are hardest for very talented people.”
Since 2018, Faculty has won more than a dozen government contracts, but Warner denies that his links to Cummings have led to preferential treatment. “We’ve worked with government long, long before Johnson was PM. I was appointed to the AI council on AI under Theresa May. We think it’s critically important that companies like ours work with government to make services better.
“From our perspective, the work of the government is just totally mission driven. In many ways, the private sector is more lucrative, and there’s less scrutiny. But as citizens of our country, we just don’t feel like we can step away from some of the important problems that the country faces. Ultimately, the government has to get better.
“The NHS has to be better and cheaper, especially with all the spending that’s gone out over the last 18 months. Realistically, one of the important ways of doing that is going to be technology companies working to help governments get better and cheaper.”
Warner has a gentle manner and is softly spoken, but becomes frustrated as he reflects on the criticism his company’s work has received. “Ultimately fair scrutiny is just totally critical to the good functioning of democracy and a government,” he says. “But it’s important to make sure that the people doing that scrutiny are honest with the public and themselves about the separation between facts and opinions.”
Many working in healthcare and government feel disappointed that AI – a field of technology that Silicon Valley executives describe as more important than fire or electricity – has failed to help combat the pandemic. While hundreds of AI tools have been built, few have played a significant role in managing the pandemic. Warner’s work is an exception to this rule, but even Faculty relies on large numbers of highly educated engineers to deliver its products. And the company’s most important contribution to Britain’s pandemic response was arguably Warner’s criticism of the herd immunity strategy, which stemmed from his own analysis, rather than any modelling Faculty had produced.
The extent to which AI reshapes businesses, governments and economies in the coming years will determine Faculty’s growth potential. “Faculty, I think, has a chance of being a very important company, because we can actually ensure that organisations get the value from AI,” Warner says. “Everyone knows that AI is a big deal. The fact is, quite a lot of people know it’s going to be a huge deal and a massive trend, but a lot of them experimented with AI, and found that although in principle it should have been really useful, in practice it didn’t really create the value.
“That’s because there’s a whole bunch of things you need to get right, and you need to apply [AI] to the right problem. You need to build good models. You need to scale those models effectively. You need to embed it in the organisation. You need to connect things across the organisation. Unless you get all of that right, you won’t actually see the value.”