Iain Clacher & Con Keating counter yesterday’s blog focussing on the gains in growth likely to happen if we adopt artificial intelligence as we could. Are they right to be sceptical – Pension Plowman.

Iain Clacher and Con Keating

In his foreword to the Government Response to Matt Clifford’s AI Opportunities Action Plan, the Prime Minister asserts:

“Artificial Intelligence is the defining opportunity of our generation. It is not a technology that is coming; a future revolution on the horizon. It is already here, … Harnessing AI and using it to deliver our Plan for Change requires ambition, purpose and focus. This is a unique chance to boost growth, raise living standards, transform public services, create the companies of the future in Britain and deliver our Plan for Change.”

The Response adopts all 50 recommendations of the AI Opportunities Action Plan. Other than references to existing private sector investment plans of around £25 billion, there are no costings of any of the multiple elements of the proposed government action. The International Data Corporation forecasts that worldwide AI spending will reach $628 billion in 2028.

The problem unaddressed by this response is that there is a very high failure rate of AI projects in the private sector. In his Harvard Business Review article, “Keep Your AI Projects on Track”, Iavor Bojinov observes

“Most AI projects fail. Some estimates place the failure rate as high as 80%—almost double the rate of corporate IT project failures a decade ago.”

The Rand Corporation, in their analysis of failures,

“The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed.”

offer the following five causes:

“First, industry stakeholders often misunderstand—or miscommunicate—what problem needs to be solved using AI. Too often, trained AI models are deployed that have been optimized for the wrong metrics or do not fit into the overall business workflow and context.

Second, many AI projects fail because the organization lacks the necessary data to adequately train an effective AI model.

Third, in some cases, AI projects fail because the organization focuses more on using the latest and greatest technology than on solving real problems for its intended users.

Fourth, organizations might not have adequate infrastructure to manage their data and deploy completed AI models, which increases the likelihood of project failure.

Finally, in some cases, AI projects fail because the technology is applied to problems that are too difficult for AI to solve. AI is not a magic wand that can make any challenging problem disappear; in some cases, even the most advanced AI models cannot automate away a difficult task.”

It is clear the Prime Minister sees AI as a panacea:

“AI-powered scans can help doctors detect disease earlier. AI can cut NHS waiting lists by scheduling better appointments. It allows teachers to personalise their lessons to their children’s needs. It can support small businesses with their record-keeping, spot potholes more quickly, and help speed up planning applications. Indeed, right across our public services, it offers frontline staff the precious gift of time. A chance to reconnect with the human, face-to face aspects of their job, which I know is something that attracts so many people to public service in the first place.”

The Prime Minister has instructed Cabinet ministers to make driving AI adoption and growth in their departments a top priority.

An important element of the adopted plan is the building of more data centres within the UK, which Government had already declared a “critical national infrastructure” in September. The two critical inputs of a data centre are electricity for operation and water for cooling. With the cost of electricity in the UK being twice that of our European neighbours and our water utilities in disarray, the UK’s competitive position is hardly attractive, The Government response, the formation of an AI Energy Council charged with “understanding the demands and challenges AI presents for energy companies” seems unlikely to resolve these issues.

While the presence of Demis Hassabis of Deepmind and AlphaFold being appointed as an advisor is reassuring, how much time will he have for this crucial role remains an open question. More fundamental to driving AI innovation, and industrial policy more widely, is consistency in policy from government. To have seen the Exascale Supercomputer at the University of Edinburgh cancelled in August last year , only to be making moves to reinstate it 6-months later does not create the necessary conditions for innovation and cutting-edge science, nor does it create the conditions for crowding in private sector investment.