Tech Frontiers™

A choice of the future

Daron Acemoglu argues that AI’s impact will be shaped by choices, not algorithms

By Eddie Milev, senior analyst, policy & insights, Economist Enterprise

5 Aug 2025

A thought-provoking image questioning who will own the future, featuring diverse individuals in discussion.Source: Green Transition Forum

“Who will own the future?” has the ring of marketing copy promoting a cryptocurrency scheme. But at the Green Transition Forum in Sofia this June, that was the title of an event with Daron Acemoglu, a Nobel prizewinner at the Massachusetts Institute of Technology. He took the question seriously—though not literally. For him, the pertinent issue is not who will own the future, but who should. And whether the rest of us are paying attention.

In a conversation co-moderated by Economist Enterprise, Mr Acemoglu sounds like a rational optimist, but his message is sobering. America is run not in the national interest, he notes with a serious face, but in the president’s own. The world is moving at a glacial pace to address climate change, with governments often misguided and corporations almost always half-hearted. Meanwhile, rich countries are ageing, their populations are shrinking and inequality is widening. Western democracies under-deliver, hobbled by mistrust and misinformation. 

Yet it is artificial intelligence (AI) that receives the bulk of Mr Acemoglu’s attention. Here too, his assessment is unflinching—but maintains an optimist’s trust that, as he puts it, “human nature is more adaptable than most people think”.

Central to his argument is the thesis of Power and Progress, a book from 2023 that Mr Acemoglu co-authored with Simon Johnson, a fellow professor at MIT and now joint winner of the 2024 Nobel Prize in Economics. It challenges conventional wisdom that technological innovation inevitably fosters shared prosperity, showing instead that over the past millennium new technologies have exacerbated inequality by consolidating wealth and influence unless they were deliberately steered toward more inclusive outcomes. “That will be true for AI if we do not make different choices today,” Mr Acemoglu says before a packed hall in Sofia.

A man in a suit and tie gestures while seated on a stage, engaging with the audience during a presentation.Source: Green Transition Forum

In question are two distinct paths for AI development. The first, embraced widely by tech evangelists and, increasingly, by western governments searching for productivity gains, seeks automation—to replace as many tasks performed by humans as possible. Under this model Mr Acemoglu likens AI to “so-so technologies”: they reduce labour’s share without meaningfully boosting job quality or overall productivity, raising corporate profits but not workers’ standard of living.

An alternative path, Mr Acemoglu argues, is that of augmentation: harnessing AI as a tool to enhance rather than supplant human capabilities. Recalling the “human-machine symbiosis” imagined by computing pioneers, such as Douglas Engelbart in the 1960s, Mr Acemoglu calls for AI that would enable “an inexperienced electrician to perform tasks at the level of a seasoned expert”. 

Yet steering AI development toward human augmentation rather than pure automation is not straightforward. One big challenge, Mr Acemoglu says, is that the agenda-setting power today rests almost exclusively with American tech giants, a few Chinese firms and their respective governments. “The decisions about AI are being made by a small elite, without input from roughly six billion people,” he remarks, more diagnostic than indignant.

Asked how societies might correct AI’s current course, Mr Acemoglu argues for a fundamental shift in incentives. Current economic and policy structures, especially in the West, heavily favour automation over human-centred innovation. Tax codes, he points out, typically impose lighter burdens on capital than labour, artificially cheapening investments in technology designed primarily to replace workers. Governments, he concludes, need to realign incentives by rewarding innovations that complement—not eliminate—jobs. But better regulation will not suffice; the culture embraced by AI builders needs to change, too. “If 60-70% of AI developers woke up tomorrow determined to prioritise technologies that benefit working people,” Mr Acemoglu notes, “the problem would disappear overnight. Unfortunately, I don't know how we could get there.”

These concerns build logically from Mr Acemoglu’s earlier work, notably, Why Nations Fail (2012), co-authored with James Robinson of the University of Chicago and the third joint Nobel laureate. In it, Messrs Acemoglu and Robinson argue that “inclusive institutions”—political and economic frameworks designed to distribute power broadly—drive lasting prosperity, whereas “extractive institutions” concentrate wealth and power, leading to stagnation and inequality. Current AI trajectories, Mr Acemoglu fears, risk establishing a similarly extractive technological order, concentrating economic gains and decision-making power in the hands of a privileged few. It would lead to “a new digital feudalism,” entrenched with inequality and with deeper political divides, he says with genuine concern.

Not set in silicon

But even if Mr Acemoglu is right to argue for human-augmenting AI that is accessible to many, he may overstate the grip of today’s tech titans. Open-source AI models—from Meta’s Llama line to those of France’s Mistral—are loosening the hold of heavyweights like OpenAI, letting students in Lagos or São Paulo to tinker with code that, a year and a half ago, sat behind corporate gates. Around this free-for-all core, a crop of ‘small-compute’ start-ups, such as Spain’s Multiverse Computing, is pruning AI models so they run on laptops or mobile phones rather than a football pitch of servers. If such firms succeed, they could slowly spread AI to places where capital, bandwidth and electricity are scarce.

A deeper question concerns Mr Acemoglu’s view of the economy’s knack for self-correction and expansion. When asked what is the one thing about the future which is assumed by most but, in his view, is not likely, Mr Acemoglu says that “in ten years the economy will have evolved and AI will be used more widely” but, he adds, “the AI revolution—and it is a revolution—will be slower and more gradual than many expect.” History, though, shows that even measured revolutions unleash powerful feedback loops. Each leap in technology, from steam engines to silicon processors, lowered costs, spurred new demand and, crucially, expanded the stock of ideas that lead to further growth. If AI can accelerate discovery—even modestly rather than at the break-neck pace dreamt of by Silicon Valley technologists—it will widen markets and spawn tasks that do not yet exist. Those gains will feed back into spending and hiring, blunting the labour-saving shock that first arrives with AI.

And still, Mr Acemoglu has a point. No technology is destiny, and how AI impacts societies and the world economy will depend on the institutions, incentives and everyday choices that steer it. Ever the realist, he considers premature fears around artificial general intelligence, a hypothetical AI capable of all human intellectual tasks, and asks us to focus on the current: “Science-fiction scenarios are intriguing,” he notes, “but they should not divert us from the practical challenges in front of us.”

In an academic fashion, then, Mr Acemoglu describes a bleak reality but also offers a way out: the world does not face predetermined technological inevitability, but it will take conscious, collective choices to prioritise human-centred AI development. Citizens, policymakers and innovators will each play a role. In short, who will own the future—whether a privileged minority or humanity at large—remains up to us.