Oxford AI guru: I’m not afraid of robot overlords — I’m worried about data, power and the tech firms playing prisoner’s dilemma

Michael Wooldridge, the Oxford professor who has been wrestling with computers since a teenager sat in a shop window tapping away on a Tandy, has a reassuringly human take on the AI boom: he’s not losing sleep over robot uprisings. What does keep him up is less cinematic and more bureaucratic — data hoarding, concentrated corporate power, and the incentives that push companies to race rather than pause.

Wooldridge has spent decades explaining complicated ideas simply, and his latest pastime is turning game theory into bedtime stories for adults. In his book, everyday dramas — from James Dean-style “chicken” standoffs to international brinkmanship — become models for predictable, if awkward, human behaviour. The Cuban missile crisis used to be the poster child for mutually destructive escalation; today he warns similar dynamics are visible in current conflicts where neither side wants to be the first to blink.

Game theory isn’t just for warfare. Wooldridge uses it to diagnose politics and culture: a zero-sum mindset — the idea someone else’s gain must be your loss — breeds misery and bad policy. He prefers thought experiments like Rawls’s Veil of Ignorance, which nudges people toward fairer systems because you might end up anywhere in them. It’s the sort of moral math that, he notes, even presidents have liked.

So how does this math meet machine learning? A lot of modern AI is about multiple agents acting on our behalf — your digital calendar talking to mine, bidding bots on auction sites, or small programs negotiating access to scarce resources. Game theory helps predict how those actors will behave when their interests don’t line up. Wooldridge’s own journey from early programming impatience to studying multi-agent systems underpins this viewpoint: the tech is older than the hype, but the scale and stakes are new.

That scale is critical. The breakthrough that made large language models ubiquitous was, in many ways, a bet on size — do the same methods but massively larger. It worked, but at a cost: compute, energy and an appetite for ever more data. Wooldridge flags data itself as the next precious commodity. If Wikipedia supplied a tiny sliver of training material, where do you find ten times more? He worries about institutions sitting on goldmines of sensitive data — the NHS being an obvious example — and the awful deals that could follow if privacy is treated as a bargaining chip.

He’s also frustrated at how Silicon Valley has shaped the public image of AI. A few very rich companies have the servers, the money and the megaphones to define the story, often around large language models that hog attention while quieter, socially useful projects go underfunded. Yet he’s not a luddite: teams in Oxford are building AI tools to make heart scans cheap and mobile, the sort of benefit that actually improves lives. Wooldridge won the Royal Society’s Faraday prize in 2025 for explaining these tensions plainly; he’s warned of possible big failures — an “AI Hindenburg moment” — even if he ranks nuclear war as a greater existential threat. “I don’t worry about a robot takeover,” he says, but the surrounding system does worry him.

What would he do differently? Slow down the race. Game theory explains why companies keep sprinting: if you unilaterally pause, you fear a competitor will rush ahead and scoop the prize. That prisoner’s-dilemma logic helps explain an industry locked into escalation: more servers, more energy, and more risk for ambiguous social gain. Wooldridge would like regulators and society to change the payoffs so sensible caution becomes rational — a harder task than it sounds, but the sort of classroom-to-global-policy lesson he relishes. In short: don’t fear the robots; fix the rules the robots are playing by — and teach a very reluctant tech CEO how to behave like a decent game player before someone else writes the scorecard.

If you want a final takeaway from a man who loves seeing that “light go on” in people’s eyes: the machines are tools, the danger is the game. And until we change the rules, we’ll keep watching competitors confess while everyone else wonders why they didn’t just cooperate.