'We can't compete': why universities are losing their best AI scientists

It was the case of the missing PhD student.

As another academic year got under way at Imperial College London, a senior professor was bemused at the absence of one of her students. He had worked in her lab for three years and had one more left to complete his studies. But he had stopped coming in.

Eventually, the professor called him. He had left for a six-figure salary at Apple.

“He was offered such a huge amount of money that he simply stopped everything and left,” said Maja Pantic, professor of affective and behavioural computing at Imperial. “It’s five times the salary I can offer. It’s unbelievable. We cannot compete.”

It is not an isolated event. Across the country, talented computer scientists are being lured from academia by private sector offers that are hard to turn down. According to a Guardian survey of Britain’s top ranking research universities, tech firms are hiring AI experts at a prodigious rate, fuelling a brain drain that has already hit research and teaching. One university executive warned of a “missing generation” of academics who would normally teach students and be the creative force behind research projects.

The impact of the brain drain may reach far beyond academia. Pantic said the majority of top AI researchers moved to a handful of companies, meaning their skills and experience were not shared through society. “That’s a problem because only a diffusion of innovation, rather than its concentration into just a few companies, can mitigate the dramatic disruptions and negative effects that AI may bring about.”

She is concerned that major tech firms are creating a huge pay gap between AI professionals and the rest of the workforce. Beyond getting the companies to pay their taxes, Pantic said the government might have to consider pay caps, a strategy that has reined in corporate salaries in Nordic countries.

Many of the best researchers move to Google, Amazon, Facebook and Apple. “The creme de la creme of academia has been bought and that is worrying,” Pantic said. “If the companies don’t pay tax it’s a problem for the government. The government doesn’t get enough money to educate people, or to invest in academia. It’s a vicious circle.”

When Murray Shanahan, another Imperial researcher, received a job offer from DeepMind, Google’s London-based artificial intelligence group, he thought hard about the decision. He saw plenty of positives to joining the company. It was a chance to pursue his work without the burden of other academic duties. He would have access to fabulous computing resources. And he would work alongside some of the best in the field. But despite the long list of pros, Shanahan paused.

View image in fullscreen

“The potential impact on academia of the current tech hiring frenzy was one of the issues that bothered me,” he said. Shanahan decided to negotiate a joint position. It allowed him to have a foot in both camps, keeping his chair at Imperial while becoming a senior scientist at DeepMind.

For those with the right skills, the hiring boom has obvious positives. Heavy investment from tech firms means there are many more jobs in artificial intelligence than there are qualified candidates. To recruit the best talent, companies offer high salaries, impressive computing facilities and technical challenges that have the potential to affect the lives of billions.

In the past, brilliant mathematicians, physicists and computer scientists headed to the City for serious money. Now they are as likely to train in AI and move to tech firms. “There are fantastic opportunities in industry, the sorts of opportunities that make working in the City seem really dull and not particularly well paid,” said Zoubin Ghahramani, professor of information engineering at Cambridge University and chief scientist at Uber, the ride-hailing firm. “It’s both intellectually interesting and, from a lifestyle point of view, very difficult to turn down.”

Ghahramani announced his move to Uber in March. For now, he commutes to the company’s San Francisco office one week every month. Next summer, he will move to the city full time. Beyond the difference in salaries, he lists a host of other reasons that academics are lured into industry. University roles come with administrative duties that some find onerous: teaching, marking, being on committees, and the endless chasing of grants. In industry, star recruits can focus purely on their research.

But there is more to it than that. The explosion of interest in artificial intelligence is driven by the success of machine learning, a field that uses algorithms to find meaningful patterns in data. To work well, many of today’s algorithms must be trained on huge amounts of data, a task that takes a lot of computer power. Without collaborative projects, universities can rarely compete with the big tech firms on data or computing. Instead, they focus on new ideas: building algorithms that learn from less information, for instance.

View image in fullscreen

Ghahramani began working at Uber part time when the company bought Geometric Intelligence, his AI startup, last year. As chief scientist, he will oversee the use of machine learning algorithms to understand how cities work and how people move around them. The end goal is to match the supply of rides to demand. “The interesting thing about this is we’re doing machine learning in the real, physical world of cities. We’re trying to optimise the movement of people and things around the world,” he said.

Ghahramani sees no sign that industry’s demand for talented AI researchers has peaked. “It’s very fierce right now and it’s yet to show signs of tapering off,” he said. “Universities will have to train enough people to meet the demand, and that’s a challenge if lecturers and postdocs are being lured into industry. It’s like killing the geese that lay the golden eggs. Companies are starting to realise that and some of the major tech companies are starting to give back to universities by sponsoring lectureships and donating funds.”

Steven Turner joined Amazon Web Services in Cambridge last year. He helps companies to build their own Amazon-style “recommendation engines”, and use image recognition, computer speech and chatbots in customer service. One financial institution he has worked with now uses technology to answer simple questions, such as on customer mortgage rates, freeing up humans for more complex queries.

In academia, he saw departments fighting for funds to continue their work and keep people from leaving. The main reason he left was to work on real problems rather than more theoretical concepts. But the culture at Amazon turned out to be more vibrant than in academia. At university, Turner found being a PhD student isolating at times, even though his supervisor was a brilliant mentor. “I personally think that having a greater focus on culture and social interaction to ensure researchers don’t feel as isolated as they can do would have a significant impact on retention,” Turner told the Guardian. He said universities should also focus on researchers’ career development, giving free access to external training and teaming up with business schools to broaden researchers’ knowledge.

Ghahramani believes UK universities will have to become more flexible about researchers holding joint positions. “They need to be flexible about intellectual property arrangements. They need to be flexible about PhD students who might want to spend time in a world-leading industry AI lab. That’s what we need to get around the problems. The universities that have been flexible have benefited,” he said.