YUVAL NOAH HARARI ON THE ECONOMIC AND POLITICAL PERILS OF AI / THE ECONOMIST

Confidence trick

Yuval Noah Harari on the economic and political perils of AI

Humanity must not entrust its lot to clever machines, the bestselling historian warns

Photograph: Sam Deardon

YUVAL NOAH HARARI describes himself as a historian who studies not the past but change. 

His first blockbuster book, “Sapiens”, applied that lens to the broad span of human history. 

The second, “Homo Deus”, looked at religion’s role in building humanity. 

His most recent work, “Nexus”, examined how the control of information has changed the course of history—and how artificial intelligence could shape the future.

In a wide-ranging interview for The Economist’s Insider show, Mr Harari warns that AI’s biggest impact may be the way it reshapes trust, “the most important asset in the world”. 

Much of modernity is underpinned by systems that allow trust to operate at scales unimaginable in the prehistoric world. 

Money is a “very creative, imaginative way to build trust between strangers”: people happily hand over valuable goods in exchange for nothing more than a promise that someone else will do the same in turn. 

“Bankers, if they are good bankers doing their job, they build trust.”

Handing powerful AI control of the financial system looks almost inevitable, Mr Harari reckons. 

AI systems will soon be capable of understanding the totality of the global economy in far greater detail than a human ever could. 

The pressure for humans to place more trust in AIs will thus grow over the next few years, Mr Harari expects. 

He points to the rise of cryptocurrencies—“AI money”, as he puts it. 

Many people have more faith in the bitcoin algorithm than even in those good bankers. 

As people lose faith in the media, he adds, they happily consume algorithmically curated feeds on digital platforms (think TikTok or Instagram).

The paradox is that individual humans’ growing trust in their personal interactions with AIs could eventually lead to distrust of the techno-financial system to which those interactions ultimately lead. 

That is because people will have no way of inferring how and why AIs make decisions. 

“The human financial system controls the lives of cows and chickens, but the cows and chickens have no idea that finance even exists,” he says. 

“This is likely to be our position in ten years.”

Even setting aside the existential risks that AI may pose, which Mr Harari thinks are real, a world in which humans come to trust AI to such an extent would be perilous. 

Imagine a financial crash in a techno-financial system, for instance. 

If its causes are beyond human comprehension a policy response will be impossible—at least from human decision-makers.

Mr Harari’s more profound worry is that too much trust in AIs could lead to techno-despotism. 

Stalin could give a speech on the radio and force all Soviet citizens to listen to it. 

But to have true control over the population he would need everyone to actually trust him, rather than just pretend to. 

And that would require millions of personal relationships, which not even the Stakhanovite dictator could pull off. 

Now, Mr Harari observes, “AI makes it possible, for the first time in history, to mass-produce intimacy.” 

Many people already profess to have AI friends and even romantic companions. 

“You can have millions of AIs working as part of a network in the service of someone, or in their own service…and then using that power of intimacy to convince [people] of whatever political or commercial interests they want.”

The only way to contain the potential harms is through regulation, Mr Harari thinks. 

He proposes a ban on AI personhood (a version of which Argentina wants to grant by allowing bots to run companies) and on AI impersonating humans (including “deepfakes”, which have become so convincing that it took him a full 90 seconds to realise that a video of one of his lectures he had recently received wasn’t real).

Given the breakneck pace of AI advances, he reckons that policymakers, especially in America, where most of these advances are taking place and whose government remains democratically accountable, must act now. 

And so should voters. 

If Mr Harari had his way, the midterm elections in November would be all about AI.