New Times,
New Thinking.

  1. Politics
  2. UK Politics
3 March 2021updated 04 Apr 2021 12:24pm

How UCL’s groundbreaking AI research became entangled in Facebook’s net

Through a series of acquisitions, professorships and industry partnerships, the university's computer science department has been drawn closer to the tech giant.

By Oscar Williams

On 4 February 2019, Facebook confirmed it had quietly acquired a London-based start-up called Chainspace. With an impressive track-record in developing blockchain technology, the start-up was the work of a small team of academics in University College London’s (UCL) world-leading computer science department.

This was an unusual acquisition. Facebook didn’t announce the deal; the social media giant only confirmed it had taken place after details were leaked to a reporter. Stranger still, it involved none of the start-up’s intellectual property. Instead, Facebook was buying the expertise of the UCL academics who had founded Chainspace. But perhaps the most unusual aspect of the acquisition was that, despite Facebook paying a significant amount of money to hire these engineers, two of the Chainspace co-founders remained on the university’s staff.

Since 2013, as the global race for AI talent has intensified, Facebook has drawn UCL’s computer science department ever further into its orbit through a series of acquisitions, industry partnerships and dual-affiliation professorships. The company has spent tens of millions of pounds acquiring three start-ups closely linked to the department, and now jointly employs four of its most high-profile professors, whose specialisms are at the vanguard of computer science and range from privacy engineering to natural language processing.

On 23 February, the two organisations aligned themselves once again. During a virtual speech at the Oxford Union, Facebook’s chief technology officer Mike Schroepfer described “working so closely with academia” over the past decade as “a huge positive for us”. As such, Schroepfer announced Facebook was “bringing its PhD programme to the UK in partnership with UCL, one of the leading universities in AI”.

Under the terms of the deal, Facebook’s artificial intelligence research lab (Fair) is funding the creation of a new PhD programme through which five students will study at the university over the next four years as Facebook employees. For an institution the size of UCL, which turned over more than £1.45bn in 2018-19 and already has access to government funds for AI positions, the £1m Facebook is committing to the project is a relatively small amount of money. For Facebook, it is infinitesimal. In the third quarter of 2020, the US tech firm generated £1m of revenue every eight minutes.

Select and enter your email address Your weekly guide to the best writing on ideas, politics, books and culture every Saturday. The best way to sign up for The Saturday Read is via saturdayread.substack.com The New Statesman's quick and essential guide to the news and politics of the day. The best way to sign up for Morning Call is via morningcall.substack.com
Visit our privacy Policy for more information about our services, how Progressive Media Investments may use, process and share your personal data, including information on your rights in respect of your personal data and how you can unsubscribe from future marketing communications.
THANK YOU

So why was UCL willing to accept the deal when it would inevitably raise concerns about academic independence? First, it appears that the funding won’t go directly to UCL, but to the students, who will be paid generously during the course of their research.

In a press statement announcing the news, UCL’s natural language processing group lead, Pontus Stenetorp, sheds further light on the university’s motivations. “Through the arrangements of this programme, our PhD students have access to the people and resources from a world-leading academic institution in AI such as UCL, and also from Fair, a world-leading industrial research lab. This makes the programme something very special indeed, and it should appeal to any student that seeks to kick-start a career in AI.”

UCL can claim to have played a pivotal role in the recent history of AI’s development. Its computer science department’s research is rated first in the UK by the Research Excellence Framework, and in recent years UCL academics have founded a series of impressive start-ups. As well as those acquired by Facebook, there is also DeepMind, the King’s Cross-based AI lab that was set up by two UCL neuroscientists in 2010, sold to Google for £400m in 2014 and is now widely regarded as the leading centre of AI research in Europe, if not the world. By continuing to foster links with the industry – the UCL computer science department also has partnerships with Google, Amazon and Arm – the university ensures that it attracts the world’s best computer science students and that its innovation legacy will live on.

[See also: Oscar Williams on the Australian government’s fight with Google]

But some observers are still concerned that the university is paying a considerable price for the partnership. As the PhD candidates will primarily be paid employees of Facebook, there are fears that they may be bound by Facebook’s non-disclosure agreements, essentially silencing their right to speak freely about their employer. Google’s recent decisions to fire two prominent AI ethics researchers, Timnit Gebru and Margaret Mitchell, who had called for greater diversity among staff and shared concerns about censorship within the company, has renewed scrutiny of the way US tech firms exploit employee contracts to stifle dissent. (Google says that Gebru left the company rather than being fired, and that Mitchell violated its code of conduct.)

Defending the PhD partnership on Twitter, Sebastian Riedel, a research manager at Fair and professor at UCL, said: “Students have full freedom in this program. They work with their UCL and Fair supervisors, and all collaborate in terms of scoping projects, setting goals and execution – directions are not defined by Facebook. And they publish and open source the resulting work.”

As part of a series of queries about the deal, the New Statesman asked Facebook and UCL if the PhD candidates would be subjected to non-disclosure agreements, business reviews assessing whether their research is aligned to Facebook’s interests, and whether UCL had negotiated contractual terms to protect their students’ work. In response, Facebook said the “student is jointly supervised by the academic thesis supervisor and a mentor from the company”. Facebook added that the students will also be encouraged “to conduct open research, contribute to open source code, and publish papers in academic conferences”. Facebook said academics with joint roles at Fair and UCL were able to produce independent research. 

UCL said: “Through our partnerships with Fair, and other technology companies, our students and researchers have the freedom and support to address some of the most challenging issues facing the world today. The research insights and developments generated through these partnerships are shared freely and the funding support we receive is predicated on our ability to share these insights.” Neither organisation directly responded to the questions about non-disclosure agreements, nor any precise terms UCL had agreed with Facebook to protect students’ work.

The legal and financial arrangements between UCL and Facebook might seem academic. But Facebook’s insatiable appetite for the university’s research, as evidenced by the acquisitions, professorships and generous PhD salaries, speaks to the enormous value it holds. And this particular deal is just one of many struck between Britain’s top universities and US tech giants in recent years. Such partnerships are welcomed by many in government as a testament of the power of British research and its ability to attract billions of pounds of investment in London by Google, Amazon, Microsoft, Apple and Facebook. These companies employ thousands of highly paid engineers in the capital and are playing a major role in the regeneration of areas such as King’s Cross and Battersea (although this, too, comes at a price).

But despite these benefits, there are concerns that by lavishing money on academics, the tech giants are successfully disincentivising them from building companies that could scale into competitors in their own right (or at the very least, to the size that DeepMind did before it was acquired in 2014). Ultimately, university partnerships may not just be about improving a company’s public image, but about maintaining market power.

This is a problem that can’t be solved by the universities themselves, but must be addressed by governments. And it is in their interests to do so – because the platform giants don’t simply want to ensure they remain at the forefront of research in order to fend off competition from academics.

Four months after Facebook acquired Chainspace’s blockchain expertise in early 2019, the social media giant unveiled plans for a new global digital currency, then called Libra and now known as Diem. The plans had been developed in large part by the UCL privacy experts who had founded Chainspace and now also work at Facebook.

But these proposals deviated in one critical way from those the academics had initially been pursuing: while Chainspace’s founders had been building a decentralised framework that would wrest back control of financial systems from central banks, Libra pursued a centralised model that would ultimately be managed by Facebook and its consortium of industry partners.

As the Chainspace saga reminds us, the US tech giants’ aggressive pursuit of British research enables such companies not only to protect their role within the private sector, but also to push into territory historically occupied by democratically-elected governments. Whitehall ignores such advances at its peril.

[See also: Oscar Williams on what’s next for Amazon after Jeff Bezos]

Content from our partners
The UK’s skills shortfall is undermining growth
<strong>What kind of tax reforms would stimulate growth?</strong>