Steven Roberts is an academic. In a leafy corner of Oxford, in a building built on the site of a now defunct iron-smelting works, he is busy turning out the kinds of minds both banks and hedge funds are want to recruit. It’s a relentless kind of work: “I’ve only had a few days off this summer,” he says. “The students here are all postgraduates, so they don’t take big undergraduate holidays.”
Roberts is a Professor of machine learning at Oxford University. He’s also director of the Oxford-Man Institute, which does, 'academically outstanding research that addresses the key problems facing the financial industry.' He was immersed in the world of machine learning long before it became synonymous with entryism. He has seen it evolve.
“I’ve had graduate students for 25 years,” says Roberts. “Decades ago, students used to go into the City or into academia. Now they also have the choice of Amazon, Google, Microsoft or Facebook. The students here don’t need to find jobs, they usually get poached before they’ve finished.”
Scarcity is an aid to employability. Roberts has between 40 and 50 students at any one time and only half specialize in finance. This is a lot compared to elsewhere. “We’re a big group by university standards – a lot of academic machine learning teams are just a dozen people.”
Predictably, applications have increased dramatically in recent years. “Very heavily over-subscribed” is the phrase Roberts uses to describe the rush of interest from students looking to burnish a first degree with an Oxford machine learning PhD or post-doctoral fellowship. “Every university that works in this domain has the same problem,” he says. “A lot of young people see that this is not just an exciting area to work in, but a gateway to opportunities in their career downstream.”
There are plans afoot to increase capacity to 70 students in future, but for the moment the course is oversubscribed 25 to 30 times. The selection process is, “very time consuming.” “The first pass is always academic merit, but we try to look at everyone as an individual, to interview as many people as we can.” The students who are chosen are “humblingly good.” “…We end up with staggeringly brilliant people.”
Staggeringly brilliant students don’t find jobs through the milk-round process. There’s no line of hedge funds waiting to seduce Roberts' postgraduates with wine and canapes on campus. Instead, students attend big machine learning conferences like NIPS and ICML. Recruiters also attend and pick them off. “They’re like big recruitment fairs which have maintained their academic rigour,” says Roberts. “It’s become very much a seller’s market”
If you want to move into machine learning, an online or short course won’t have the same effect. Roberts says online machine learning courses such as that run by Andrew Ng are impressive for their “quality and depth of understanding,” but that most people who get jobs with big banks and big technology companies have PhDs. “It doesn’t mean you’re smarter. It means you’ve been exposed to machine learning techniques and have an experience base.”
Machine learning has applications far beyond finance. Financial services isn’t even its ideal realm of use (“In finance the rules change from move to move. You never have full 'observability' and you never know what was going on, except with hindsight.”) At Oxford, students are also engaging machines to explore everything from deep space to animal welfare. “A cheap camera in a chicken shed can monitor the motion of the birds and generate data that turns their motions into a series of patterns which correlate to stress and welfare outcomes,” says Roberts.
This variability is one reason Roberts has never jettisoned his job for the higher pay in hedge fund land. “I could get paid 10 times in finance," he says (stressing that he by no means intends this as a negative observation), "but only one third of my academic life is spent working on problems directly related to finance.” There are crossovers anyhow: “In astronomy, you’re often looking for a weak signal in a hubbub of noise and we have been able to transport techniques from astronomy to applications relating to markets.”
Roberts regrets that universities don’t have resources to develop solutions for problems like global warming, food security, disease, ocean plastic, or solving nuclear fusion. Nor, though, does he judge his students for choosing technology companies or hedge funds ahead of academia or government. “Money talks. Students want to solve interesting problems,” he says. It’s just that solving commercially interesting problems usually pays more than solving worthy ones.
It’s fortunate, then, that hedge funds and tech companies are often no more than a temporary stop-off for many of Roberts brilliant people. “I never think it’s a shame that people go into finance,” he reflects. “They go into finance, and then they come back when they don’t need the money and can afford to live on an academic salary.” How long does this take? “It’s typically ten years or so,” says Roberts. “It depends how lucky they are.”
Have a confidential story, tip, or comment you’d like to share? Contact: firstname.lastname@example.org in the first instance. Whatsapp/Signal/Telegram also available.
Bear with us if you leave a comment at the bottom of this article: all our comments are moderated by human beings. Sometimes these humans might be asleep, or away from their desks, so it may take a while for your comment to appear. Eventually it will – unless it’s offensive or libelous (in which case it won’t.)