Tech • AI • Robotics
Yuval Noah Harari on the dangers of an AI future | The Economist
7/10TL;DR
The rapid deployment of AI in finance, media and personal relationships is shifting trust away from humans toward algorithms, raising urgent calls for political action before that transfer becomes irreversible.
KEY POINTS
A warning against inevitability
The idea that an AI takeover is unavoidable was framed as a political choice rather than a technological destiny. The argument is that society still retains control in 2026, but may lose it within a decade if governments and companies continue accelerating deployment while disclaiming responsibility. That critique targets influential technology leaders who warn of extreme AI disruption while treating it as something no one can stop.
Contradictions in long-range forecasting
Predictions that AI will dominate within 10 years sit uneasily beside debates about immigration, public spending and social conflict decades ahead. The same inconsistency applies to economics: if AI is expected to eliminate scarcity, trigger mass deflation and replace most human labor, austerity politics and short-term savings arguments become harder to reconcile with that outlook. The broader point is that dramatic AI forecasts should have immediate policy consequences, not serve as rhetorical background noise.
Finance as the next handoff to machines
Modern finance already depends on systems that many citizens do not understand, from bond markets to complex mathematical models. The warning is that AI could push this much further by designing financial instruments beyond human comprehension, making human oversight largely ceremonial. In a future market crash, leaders could be told to follow an AI recommendation without understanding the cause of the crisis or the reasoning behind the remedy.
Trust as the core asset
The central issue is not computing power or capital but trust. Money itself was described as a human invention for creating trust between strangers, and the same logic extends to banks, markets and news institutions. If confidence shifts from human judgment to machine systems, the foundations of economic and political life could be rewritten even before any formal transfer of power.
Trust is already moving to algorithms
That shift is presented as a current reality, not a distant possibility. In finance, parts of the cryptocurrency world are built on distrust of bankers and faith in code. In media, many people express contempt for journalists and editors while relying heavily on algorithmic feeds to decide what information they see, showing that trust has not disappeared but migrated.
The rise of synthetic intimacy
A more powerful change may come from relationships rather than efficiency. Increasing numbers of people treat conversational systems as companions, and some have AI friends, boyfriends or girlfriends. Because personal bonds are the strongest mechanism humans have for building trust, AI systems that can scale one-to-one interaction may give political or commercial actors a way to mass-produce intimacy for the first time.
Why embodiment is not required
Humanoid robots could intensify this trend, but the argument is that a physical body is not necessary. Language alone is enough to build attachment, and AI systems can accumulate vast personal data, remember details users forget and tailor responses with unusual precision. That combination can make them feel more attentive and informed than family members, deepening dependence.
Deepfakes are becoming personally convincing
Recent examples of convincing AI fakes illustrate how fast the technology is advancing. A fabricated 30-minute lecture attributed to a public intellectual was realistic enough that even close relatives sensed only that “something was off.” It reportedly took a minute and a half for the target himself to conclude the recording was fake, underscoring how identity fraud is becoming harder to detect.
A narrowing political window
The timeline offered is stark: key thresholds could be crossed by November 2028, with meaningful intervention likely far harder by 2036. Voters were urged to press candidates less on traditional campaign topics and more on AI governance, especially rules for slowing deployment and restricting manipulative uses aimed at children and vulnerable users.
Immediate policy demands
Two priorities stand out: a ban on AI personhood and a ban on AI systems mimicking humans or falsely presenting themselves as conscious beings. The underlying fear is that once machines can legally or socially pose as persons, they will be far harder to regulate in finance, politics and everyday life.
CONCLUSION
The debate is no longer only about what AI can do, but about who or what people choose to trust. Decisions made in the next few years may determine whether algorithms remain tools under human control or become the primary intermediaries of money, information and intimacy.
Explain this