It’s hard to recall a more controversial technology than AI. Businesses are caught between FOMO (fear of missing out) and FOGIW (fear of getting it wrong). Both have potentially terrifying consequences.
It’s not just business. Governments are also trying to find a safe path between AI innovation and regulation. On one side the interests of the private citizen, on the other the interests of the economy.
In the past month, the UK government has put AI at the heart of its economic strategy with claims that the technology could unlock public sector productivity, saving £45 billion. That’s about a quarter of the annual NHS budget.
In February, the UK government announced new legislation to limit the use of AI in the creation and distribution of child sexual abuse images.
The same go-faster, slow-down ambivalence can be seen in AI policy everywhere in the world. On the one hand, a desire to unleash innovation. On the other, a frantic drive for regulation to make the consequences of innovation as benign as possible.
AI is ambivalent about its own potential. Asked if AI is safe, ChatGPT opines that “AI can be safe, but like any powerful technology, its safety depends on how it’s developed, deployed and used…”
The trash can of history
Governments love the limitless possibilities of AI, but only when they can control them. DeepSeek was widely praised for its innovation, but the China-developed AI is also self-censoring. Don’t expect DeepSeek to tell you about historical events that show China in a less than favourable light, such as Tiananmen Square. The AI has “forgotten” them.
The DeepSeek furore illustrated another truth about AI, which is that control will be hard to exert. The late January run on technology stocks was fuelled by the fear that export controls on foundational AI technology, Nvidia GPUs, were not working. DeepSeek had apparently been created using an earlier generation of Nvidia device which wasn’t subject to the rules.
The other remarkable aspect of DeepSeek were claims that it had been developed at a fraction of the cost of other LLMs. Great news for innovation but terrible news for tech companies whose own multibillion dollar investments now looked massively overpriced.
The unlevel playing field
As geopolitics and industry politics play out, it’s clear that for all the talk of transformative benefits to humanity, the IT industry’s quest for an unfair competitive advantage continues unabated.
Government’s AI policies also risk clashing with others, notably environmental policy. The race is on to build hyperscale data centres with many times the energy consumption of today’s facilities. Data centres already account for 2-4% of global energy use and 20% in the Republic of Ireland. (The country has a high concentration of data centres thanks to a combination of its location between the US and mainland Europe, low corporate tax rates and a cool climate.)
This month, the former chief executive of Google, Eric Schmidt, told the BBC that he feared what “rogue states” could do with AI. North Korea, Iran or Russia could use it to develop biological weapons. Unsurprisingly, Mr Schmidt is a fan of export controls.
Like other AI advocates, the ex-Google boss thinks regulation is a good thing – up to a point. He issued the familiar warning that while AI should be regulated, over-regulation would stifle innovation.
How much is enough?
What just enough regulation looks like is the problem governments will wrestle with for the foreseeable future. The problem, as ever, will be enforcing it. History teaches that regulation has little effect on lawless states and cybercriminals and is a mixed blessing for the law-abiding.
Less than a year ago the EU passed sweeping legislation in the form of the AI Act. But the first national regulations were introduced three years earlier, by China. Enthusiasm for regulation is not a reliable indicator of intent.
Which is the stronger impulse, FOMO or fear of getting it wrong? The headlong rush to an AI-enabled future, fuelled by the media, driven by investors and whipped up by government rhetoric about economic transformation is full-on FOMO. The prevailing sentiment is that we need to innovate the hell out of AI. The much less exciting business of building guard rails will continue to bump along behind.
