David Willetts was Minister for Universities and Science 2010-2014.
Two Brits won Nobel prizes last year for their work on AI. Sir Demis Hassabis created Deep Mind which uses large datasets not just to learn to win at Go but also to identify protein structures; Geoffrey Hinton won his for the original insights into how a large language model could work.
Their research careers overlapped at one key place: UCL’s Gatsby computational neuroscience unit, backed by the great David Sainsbury.
We don’t just have great research strengths we also have great data. This became clear when Boris asked me to join him on a science and innovation trip to the US East Coast. We were going to see how we were doing on life science and genomics research compared with leading research institutes in America.
Our life science researchers’ aim was to link up genetic data with health data collected by GPs and hospitals throughout someone’s life; connecting up that data in the US is virtually impossible, as they have a far more fragmented system.
So we have brilliant innovative thinkers and precious public service data. But as so often we have not done so well at converting these extraordinary advances and resources into commercial companies or useful British applications. If you think not of the Nobel Prizes but of the Oscars, we seem to end up with the best supporting actor awards.
When Sir Demis sold DeepMind to Google, the increase in Google’s share price was far more than they paid for it. I asked him why he had sold. His answer was that DeepMind needed access to compute power costing about $1b a year and there were no public or private funders with that funding capacity in the UK.
Faced with such American financial power, my aim as Science Minister was at least to keep supporting our great technologies and start-ups for long enough to ensure they put down deep roots here – and even if the Americans bought them they could not move them lock stock and barrel to the West Coast. The dynamism of Kings Cross today is evidence for that.
In those days there were still sceptical Tories, such as Sajid Javid, who thought the Government had no role backing key technologies. But that view of the world has now passed, not least thanks to the unlikely partnership of Tony Blair and William Hague, who are eloquent on what technology can do with Government backing.
Rishi Sunak got all this and took a personal lead on AI. But he decided to go big on the safety and security angle. That is a genuine concern, but the intense focus on it squeezed out the more positive case for AI.
There was also the classic problem of a British inferiority complex with America. This meant he did not go for a British company to lead in co-ordinating access to health data. That contract went instead to Palantir – probably a missed opportunity.
But Sunak did at last get on with boosting our computing capacity. We had fallen woefully behind. It is a case study in Britain’s problem at getting anything done. There were endless reviews, and ministers and advisers changed frequently (they usually started with the question why can’t we just use Amazon Web Services).
By the time they had been taken through the answer to this and begun to see the need, they were moved on and the briefing began over again.
As well as exascale computers we were also falling behind on the workhorses of AI: graphics processing units (GPUs), first developed for computer games which have become key to this new technology. At one point I estimated we had fewer GPUs across all universities in the UK than they had at Stanford alone.
But the strategy of starving universities of funds meant that they simply did not have the resources to provide this sort of kit. Expecting them to train the next generation of AI professionals whilst they run at a loss both on teaching and research is not a viable strategy.
George Freeman and his successor Andrew Griffiths got all this, and the last government announced a new exascale computer – we have not had such investment for a decade. Then Labour cancelled the plan. Thankfully, it has been revived as part of a wider package based on Matt Clifford’s excellent report.
Above all there is a growing recognition that if we are to run better public services at lower cost we have to be able to harness innovation and AI is key. So much clunky data collection and manual reporting can be replaced by AI.
Peter Kyle and Wes Streeting are genuine believers in all this. A key question is exactly who can access health data and on what terms it can all be linked up. There are still GPs still refusing to link their patient data with the Biobank’s records. A lot of the data is still manually recorded or has been digitised so poorly it is hard to use. There are still gruelling battles both to get it to happen and to maintain public confidence.
The Government can also look beyond health at other applications such as handling welfare claims – and not just to identify possible fraud. Police officers still waste a shocking amount of time filling in reports rather than investigating crimes. AI is our best hope of transforming the efficiency of public services.
One very interesting area is education data. Ministers in the previous government were opposed to any shift to online learning and joined media campaigns against it. But it can be a useful aid for teaching. Teachers can track how someone is learning key stroke by keystroke and see where they get stuck. But the debate on ethical constraints on education data way behind where we are on health data.
Yesterday’s announcement makes a lot of sense. It is just possible this is an area where, with cross-party support, things might actually happen.