UK's plan to become AI world leader is flawed

Professor George Buchanan, Deputy Dean of Research, School of Computing Technologies

“Kier Starmer’s AI Opportunities Plan aims to reduce government costs, boost business productivity and resolve mounting long-term political and economic problems. But it seems more likely to create another black hole for the UK government’s budget than save it through scientific ‘white heat’.

“The new needs to be understood within the nation’s sixty-year political tradition. In 1963 the then leader of the United Kingdom’s Labour Party, Harold Wilson, spoke about Britain’s need to benefit from ‘the white heat’ of a scientific revolution. Subsequent prime ministers, including Margaret Thatcher, Tony Blair, and now Kier Starmer have placed great emphasis on new technology changing Britain’s government and economy.

“The plan has numerous problems. From a computing point-of-view it repeats failures of the past: the UK technology sector has been weakened by decades of selling off key businesses to overseas buyers; its university sector is in a financial crisis caused by declining funding, particularly in computing; and the body politic has a track record of poorly understanding technology and backing politically convenient, but technically or organisationally naive, partners.

“The current plan ignores significant privacy issues, and trusts in deanonymisation as a guarantee of privacy. It appears to rely on easy access to large-scale government data by third parties without considering its potential misuses. This has already failed in the past with much less powerful computers almost immediately de-anonymising the clinical records of Welsh government ministers. The high-risk, low-privacy approach puts the UK government at odds with an EU marketplace to which they desperately need improved access post-Brexit.

“More concerning though is the long-term capacity of the UK government, of whichever party, to execute technological grand-strategy. Tony Blair’s government saw the largest civilian computing project in the world, the National Health Services’ National Framework for IT, fail with huge costs. Subsequent conservative governments fared no better, with failing projects on ankle tags for criminals and others delivered many years late and over budget.

Professor George Buchanan is a world-leading researcher on large-scale information systems and has contributed to a book on the usability of government systems. He has worked for the United Nations and UK Government on strategic IT projects.

Dr Dana Mckay, Senior Lecturer in Innovative Interactive Technologies, School of Computing Technologies

"AI is seen by many – including governments, economists, and tech companies – as a panacea for economic woes. The problem with this approach in the UK is threefold.

“The first problem is that we are already experiencing an energy crisis, a problem that has already been prominent in UK news, and AI takes enormous amounts of energy to run. That small (and often not very useful) ‘AI Overview’ Google serves you with your search results increases the energy consumption of your search by more than a factor of 10. The UK is not well placed to generate the significant new energy Starmer's plan would require.

“The second problem is technical knowledge. The UK has been underinvesting in education and research and development (R&D) for decades, and is unlikely to be able to create a new 'silicon valley' readily without these skills. The Google DeepMind project mentioned in Starmer's address was created well over 20 years ago, as part of a previous wave of AI hype.

“The third problem is that AI is often inaccurate, and the further you are from being a straight, white, wealthy man, the more likely you are to be harmed by its inaccuracies. Public sector workers using AI, for example, has been trialled here in Australia with disastrous results in the Robodebt scheme. Recent research right here in Melbourne has surfaced many problems with the use of AI in education, one of Starmer's proposed target areas. Starmer's plan also aims to hand over sensitive data, such as health records, to AI, without clear legal requirements for privacy or human rights, likely reinforcing or worsening existing inequalities."

Dana McKay is Associate Dean, Interaction, Technology and Information in the School of Computing Technologies at RMIT University. Dana's research focuses on ensuring advances in digital information technologies make the world a fairer and more equitable place.

***

General media enquiries: RMIT External Affairs and Media, 0439 704 077 or news@rmit.edu.au