Would democracy survive an AGI-supercharged economy?

AGI could lead to growth of 30% — and double digit unemployment. It’s not clear that democratic institutions would be able to survive

Port Everglades dockworkers picket the entrance to the port October 2024. Credit: Joe Raedle/Getty Images

By Chris Dorrell

If you believe the prophets of Silicon Valley, the global economy is on the verge of a technological revolution thanks to artificial general intelligence (AGI).

In this vision, economic growth will accelerate to 20-30% a year on a sustained basis.

“The emergence of AGI will mark the end of the Industrial Age and usher in a new economic era, just like the Industrial Age ended the Malthusian era,” argues Anton Korinek, an economics professor at the University of Virginia, and one of the leading economists working on AI.

If the potential rewards are immense, then so is the scope for disruption. This new world might be a world of mass unemployment and rampant inequality, or a world of contented leisure and plenty.

Would democratic institutions be able to cope with this rate of growth? Or would they prevent it from ever happening?

What the economy might look like

The case for 20-30% annual growth is largely based on the idea that AI will at some point be able to automate any task that a human can do, whether that’s technical research or making a coffee. Moreover, it will perform those tasks more effectively and at a lower cost. The process of generating new ideas, the driving force of capitalism, may even itself be automated.

Productivity would explode. So might the labor market. Why would anyone employ a human when an AI could do the job better and cheaper?

Korineck and Donghyun Suh, an economist at the Bank of Korea, have attempted to model the impact of AGI on the labor market. Their conclusions are not comforting.

“If automation proceeds sufficiently slowly, then there is always enough work for humans, and wages may rise forever. By contrast, if the complexity of tasks that humans can perform is bounded and full automation is reached, then wages collapse,” they write in a paper for the National Bureau of Economic Research.

It may be that humans get lucky and there are some tasks too complex for AI. In that case, Korinek and Suh argue wages can keep rising — but only if investment in new productive assets (capital accumulation) keeps pace with automation. Investment could create demand for new roles, helping reabsorb displaced workers. But if automation instead advances faster than capital builds up, labor becomes too abundant relative to capital, and wages can collapse even before full automation.

This is just one forecast, and not everyone agrees with Korinek and Suh’s gloomy prognosis. Daniel Susskind, professor in economics at King’s College London, argues that we might voluntarily constrain the number of jobs offloaded to AI systems. There are some roles, he notes, that we might prefer humans for — such as artists, teachers or psychologists. In other cases, we may feel a moral obligation to keep a human involved — such as in medicine or politics. Susskind accepts that our sense of these preference and moral “limits” might change over time as AI improves, but they could be a block to rapid automation in the short-term.

It is also possible that AI will end up creating more jobs than it destroys, just as previous technologies have. Some argue AGI might be different, because in theory it will be able to do everything, but it is impossible to tell what roles might yet come into existence.

Still, it is likely that a ‘growth explosion’ — which is necessarily fast and far-reaching — will lead to high levels of unemployment, at least in the short-term.

Alongside the danger of mass unemployment, it is also very likely that the rich will get ever richer as the returns to capital increase. This will clearly be the case for the tech billionaires, who own the machines which will be responsible for driving the gearshift in economic growth, but it is not just the tech billionaires who will benefit.

Emma Rockall, Marina Tavares and Carlo Pizzinelli at the International Monetary Fund (IMF) modelled the potential impact of AI adoption on wealth inequality using data on UK households.

In the UK, workers in the highest income groups had the lowest share of their total income from wages, and held the largest wealth. They were also most likely to benefit from improving stock market performance.

“As such, high-income workers are not only better placed to insure themselves against potential adverse labor market impacts from AI, but also stand to benefit most from a possible AI-driven rise in capital returns,” they wrote.

They also found that high-income jobs were more likely to be complementary with AI, rather than be substituted, potentially compounding the process of widening inequality.

Impact on democracy

There’s a very real danger, then, that without intervention, AGI will create conditions which stand in conflict with the egalitarian ideals of our political systems. It is not certain that democracy would survive the transition.

Periods of mass unemployment pose major challenges for democratic politics. The rise of fascism in the 1930s took place against the backdrop of widespread joblessness. German unemployment in 1933, the year Hitler took power, was just over 30%, for example.

More recently, the fallout from the eurozone crisis in the early 2010s eroded the legitimacy of democratic institutions in many European states. Portugal, Greece, and Spain all saw declining rates of trust in their political institutions as unemployment rose.

It is easy to see why. Democracy is a system in which decision-makers are accountable to their citizens. Democratic institutions should respond to changes to the economy at an early point to promote the welfare of the population at large, which is difficult to square with mass unemployment. Popular resistance seems likely.

“In cases of significant economic and social disruption, there’s often popular resistance,” Miranda Bogen, founding director of the AI Governance Lab at the Centre for Technology and Democracy told Transformer. “This will likely evolve in different directions along different feedback loops, not progress in a straight line.”

She predicted that there would be a number of issues along the way, such as the quality of work that humans were doing and the question of exploitation, which could spark government intervention.

Speaking at an Institute of Progress event in January, the economist Tyler Cowen also predicted that there would be significant popular resistance to automation.

“Once it (AI) starts changing what the world looks like, there will be much more opposition. Not necessarily on doomster grounds, but from people who say ‘I grew up and trained my kids for a different kind of world, and I don’t want this’. That’s going to be a massive fight, and I have no prediction about how it is going to go.”

This fight is already happening. Last year striking dockworkers in the US demanded a total ban on the automation of gates, cranes and lorries to protect their livelihoods. Demands for a total ban on automation were rejected, but the dockworkers did secure protections against the introduction of any fully automated equipment.

This followed 2023 strikes in Hollywood, in which screenwriters won protections against the use of AI in scriptwriting. The eventual contract, which followed nearly five months of industrial action, prevented studios from treating AI-generated content as "source material."

Politicians have expressed interest in halting automation, too. Sen. Josh Hawley recently called for a ban on self-driving cars and trucks, arguing that they “would be terrible, terrible for working people.” If a large enough coalition of lawmakers feel similarly, regulation could potentially stop an automation wave in its tracks.

Who benefits?

If, on the other hand, governments were to act as the handmaiden for the new age, they would find themselves in a very different world.

The most important task of any government, particularly a democratic one, is to provide for the welfare of its citizens. But widespread automation could fundamentally break how governments do this.

The challenge is simple: if labor were to become redundant, then spending on social security would have to increase to support the newly unemployed. But widespread unemployment also means a significantly smaller tax base.

Taxes levied on employment in the US made up just over 67% of government revenue in 2024, and 46% in the UK. If employment collapses, governments will find themselves with a huge shortfall.

Capital taxes are an obvious way to plug the hole. Nobel-prize economist Daron Acemoglu notes that capital is taxed much more lightly than labor in the US, offering a big incentive to automate. Taxes increase the cost of labor to employers by around 25% compared to just 5% for things like software and equipment.

“The US tax system encourages companies to buy machines while discouraging them from adding workers,” he writes. Changing the tax treatment of capital investment could therefore slow the pace of automation and boost state coffers.

Similarly, Bill Gates proposed a robot tax back in 2017, suggesting that a company should pay a levy if it wants to automate a job previously performed by a human. “A human worker who does, say, $50,000 worth of work in a factory, that income is taxed and you get income tax, social security tax, all those things. If a robot comes in to do the same thing, you’d think that we’d tax the robot at a similar level,” he told Quartz.

Rather than taxing the robots themselves, Anthropic CEO Dario Amodei suggested in May that a special tax on AI companies might be necessary.

“If AI creates huge total wealth, a lot of that will, by default, go to the AI companies and less to ordinary people,” he told CNN. “It’s definitely not in my economic interest to say that, but I think this is something we should consider and I think [a tax] shouldn’t be a partisan thing.”

Capital gains, which are taxed well below income across most advanced economies, could also be increased to help shore up the tax base.

As ever, there is a trade-off. Capital is highly mobile: set the tax rate too high, and democratic governments might scare off the golden goose. And while capital taxes would help to control inequality and be an important source of revenue, they could also limit productivity growth distorting where investment is directed and limiting AI adoption.

In their study on UK inequality, researchers at the IMF found that a 15% tax on capital income would have a major impact on national income, even if it restricted inequality:

"Even though many households would have higher total incomes following the policy, the deadweight loss to society is large — the total income lost across all households exceeds the tax revenue raised by more than a factor of two."

But given the need to support the new swathes of unemployed, the tradeoff might be worth it. Many have argued that some form of universal basic income — a guaranteed income to all citizens, regardless of whether they are employed — will be necessary in the AGI age. Elon Musk, Sam Altman, and Geoffrey Hinton have all endorsed the idea.

Capital levies will almost certainly be required to fund such a program. But they raise another issue.

A world in which governments captured a healthy surplus from big tech companies to disperse among the wider population would not necessarily be undemocratic, but it would increase the risk that government decisions become responsive to a narrower group of corporate interests.

“Democratic governments might opt to tolerate dramatic shifts in the economy if they perceive it to be in some sort of long-term interest. The question is then whether they will be responsible to the public if popular pressures demand change, or whether other interests will sway their decision-making,” Bogen said.

The balance of power

Sam Altman neatly summarized the potential danger of the new era of AI in a blog post earlier this year: “The balance of power between capital and labor could easily get messed up.”

Rebalancing that relationship will be the fundamental challenge for democratic governments in the era of AI. This will involve challenging the interests of the tech companies, the dominant capitalists of the new world, either through heavier taxation or regulation.

Failure could be fatal for democracy. Looking back in history, there are plenty of signs that disruption can breed destruction.

Tyler Cowen noted that the most damaging dictators of the 20th century emerged in the context of “a highly disoriented civilisation.” Hitler, Stalin and Mao came to power “right after the period when the world was changing the most,” he said.

If managed poorly, a transition to the age of AGI could be every bit as disorienting, and damaging, as the changes of a century ago.

Chris Dorrell is a freelance journalist.