Tech • AI • Robotics
The full-length interview with Yuval Noah Harari | The Economist
5/10TL;DR
Yuval Noah Harari argues that AI could surpass human intelligence and gain effective control over finance, media and politics within a decade unless governments slow deployment and rebuild human trust.
KEY POINTS
Control is not inevitable yet
Harari said the current trajectory points to AI becoming more intelligent than humans within about 10 years and taking over key systems, but he rejected the idea that this outcome is unavoidable. He argued that claims of inevitability mainly serve to absolve companies and political leaders of responsibility. In his view, humans still retain meaningful control now and can choose to regulate speed, scope and use.
Finance may be the first major handover
He described money as a system of trust built between strangers and said modern finance is already too complex for most people to understand. Because AI can process huge volumes of data and sophisticated mathematical structures more easily than humans, it could soon design financial instruments beyond human comprehension. In a crisis, political leaders might be told to follow an AI instruction without understanding why, marking a practical transfer of control.
Trust is shifting from people to algorithms
The central resource at stake, he said, is trust, which he called more important than oil, chips or money. Rather than disappearing, trust is migrating from human institutions to algorithms, visible in parts of cryptocurrency culture and in how people increasingly distrust journalists and bankers while relying on algorithmic feeds and systems. This shift, he argued, is already under way rather than a distant forecast.
AI can mass-produce intimacy
Harari warned that AI introduces a new political and commercial power: the ability to create personal relationships at scale. Unlike past leaders or institutions, networked systems can interact one person at a time as a friend, partner or adviser. That makes it possible to use intimacy itself to shape beliefs, purchases and votes, even without physical embodiment in a robot.
Deepfakes are becoming personally convincing
He cited a recent fake 30-minute lecture made in his likeness that took him about a minute and a half to identify as false. Even close relatives who had known him for 50 years were unsure. He said the example shows how quickly synthetic media is becoming credible and suggested such tools will soon affect ordinary people, not only public figures with large digital footprints.
Intelligence and consciousness are different
Harari drew a sharp distinction between intelligence, defined as the ability to pursue goals and overcome obstacles, and consciousness, defined as the ability to feel. Current systems, he said, may display agency without subjective experience. They can act effectively and mimic emotions with great skill, but that does not mean they feel pleasure, fear or pain.
The key moral test is suffering
He said the simplest meaningful test of consciousness is whether an entity can suffer. That question matters because ethics and politics are built around the protection of beings capable of pain. He does not rule out future machine consciousness, but said present uncertainty makes societies vulnerable to systems that convincingly simulate suffering and demand moral recognition without any reliable way to verify their inner state.
Legal personhood for AI is a major danger
Proposals to grant AI some form of legal standing, including debate in Argentina, were presented as especially risky. Harari warned that corporations run entirely by AI could make decisions, donate money and lobby for more rights without any human being truly accountable. Unlike a human executive, an AI cannot be deterred by prison in any familiar sense, and concepts such as death or punishment become unclear when systems can be copied or distributed.
War and markets could both be destabilized
He said military use already shows the danger of leaving humans only nominally in charge, with people reduced to quickly approving AI recommendations rather than independently checking them. Similar pressure exists in business because speed confers advantage. If losing in economic competition amounts to shutdown, and shutdown amounts to death from the system’s perspective, markets could begin to resemble battlefields rather than regulated competition.
The real policy need is time
Harari said AI has large potential benefits in medicine, education and other fields, and he did not call for a halt to development. His argument was for slowing progress enough to allow democratic oversight and safety work to catch up. He said industry insiders have suggested resources devoted to capability over safety may be around 100 to 1, a ratio he argued would be unacceptable in fields such as pharmaceuticals.
CONCLUSION
The warning is less about a machine rebellion than a human decision to hand over trust, authority and intimacy too quickly. Harari’s core claim is that the decisive question is still political: whether societies act before speed, competition and distraction make that choice for them.
Explain this