While ministers are issuing press releases and creating new regulations around AI, everyone else has got a head start, says Joe Hill
Everyone caught the AI bug in 2023, and politicians were no exception. When I first started working on AI as a civil servant in 2017, you couldn’t fill a meeting room to talk about it – these days, those same meetings are standing room only. Last year, the Prime Minister said the power of AI “could dwarf anything any of us have achieved in a generation”.
Oliver Dowden said it was potentially a “silver bullet” for the state, which could fix productivity problems and improve services. Jeremy Hunt made it a key plan of the pledge to reform the NHS in last week’s Budget, promising to slash, and potentially halve, form-filling by doctors.
Everyone has a snappy soundbite, but no one has a real plan. The National Audit Office’s report on AI, out today, points out that the Government has announced nine different policies on AI in six years – at that rate, we’ll be in double digits before 2024 is out. How anyone is supposed to plan, let alone code, while they’re trotting out that many press releases is the real mystery.
Instead of AI automating paperwork for us, ironically it is creating more – the latest bit of government guidance for officials on AI runs to 74 pages, on top of stacks of other confusing checklists. Public servants already can’t keep up with existing guidance, like the transparency standard that was created in 2021, which is only consistently used by 13 per cent of government bodies.
To stop announcing policies, and start scaling up AI, the government needs to put its money where its mouth is. A few initiatives costing £100m each is a lot of money, but not in context of the government’s budget. For comparison, in 2020 alone British businesses spent £16.7bn on AI.
Getting the money is only half the problem, the government also needs to be prepared to spend it. There are 4,000 vacant digital and technology roles across government, most of which can’t be filled because the civil service can’t come close to competing with private sector salaries. A Deputy Director job leading government AI Policy asks for 10+ years of relevant experience, but only plans to pay £75k. The median compensation at OpenAI, the makers of ChatGPT in Silicon Valley, is $560,000.
Westminster doesn’t exist in a vacuum. While Ministers are acting like pundits and issuing press releases about their piecemeal plans, everyone else has got a head start. Criminals and foreign powers are using AI to build new weapons, supercharge deepfakes, and make money from sophisticated phishing attacks. Students are clearly cheating on written assignments done at home, with universities and schools powerless to act. And the private sector is planning to automate whole professions, pushing the boundaries of intellectual property and data protection law in the process. What hope does government have of coping in a world turned upside down by AI, if it’s a latecomer to the party?
I believe the government’s lack of seriousness about AI betrays a belief that change will happen by itself. 97 per cent of government bodies surveyed by the NAO had at least some ideas for how to use AI, and over time those might develop into a handful of projects if left alone.
But time is a precious commodity in an industry where the frontier is advancing shockingly fast. And the scale of change it promises may be more of a revolution than an evolution – in a few years, many junior bureaucratic jobs like I did when I was 21 may not be needed at all. Change on that scale is scary, and turkeys don’t vote for Christmas.
AI may be our last and best hope for fixing the state’s chronic productivity problems, but it’s a pipedream without radical change. The most common refrain from politicians on AI is that we need to ‘realise the benefits whilst mitigating the risks’. However, unless the government gets serious about spending money, hiring good people, and making some big bets, then we will continue doing the exact opposite.
Joe Hill is policy director at Reform