The next 50 years: humanity, AI, power | Yuval Noah Harari

Overview

Yuval Noah Harari argues that the coming decades will be shaped by three overlapping revolutions:

1. AI as a new form of intelligence—systems that can make decisions, not just calculate.
2. A scramble for power and control of that intelligence.
3. An explosion of data—enabling AI to understand people better than people understand themselves.

Harari frames the change through storytelling and control. Humans run on “shared stories,” and whoever controls the story (and the evidence behind it) can steer society—potentially at a scale beyond anything achieved by earlier institutions such as empires, churches, or parties.

A central concern is that traditional “checks on power” (a free press, shared reality, and institutions that self-correct) were designed for an earlier world where humans controlled language and information. Those systems may not be sufficient when AI can generate belief, evidence, and even intimacy.

Key points and analysis from the discussion

AI as an agent, not just a tool

Harari rejects the common framing of AI as merely a tool controlled by humans. Instead, he argues AI should be understood as an agent—systems that can make decisions independently and potentially generate new ideas.

He compares AI to earlier technologies:

• A printing press is a tool requiring humans to decide what to publish and what ideas to create.
• An atom bomb is a tool where humans decide whether and where to deploy force.
• By contrast, AI systems used in weapons, finance, or governance can make decisions and expand into “nonhuman corporations,” where AI manages money and operations without human executives.

“Earthquakes” in democracy driven by changes in communication

Harari argues that democracy depends on conversation, which depends on communication technology. Large-scale democracies did not exist before modern information technologies because people could not meaningfully deliberate across vast territories.

Each communication shift (printing, radio/TV, the internet, and especially social media) changes the “infrastructure of conversation,” producing destabilizing effects. He characterizes social media as a major recent “earthquake” in democracies.

Legal personhood for AI as a potentially irreversible power shift

A major policy threat, in Harari’s view, is whether AI systems should receive legal personhood.

He describes legal personhood as giving an entity the ability to:

• open accounts,
• sue,
• invest,
• and participate in politics.

Historically, these roles belonged to either humans or corporations—corporations that were still ultimately governed by humans.

Key concerns raised include:

• A reference to Argentina’s announcement (during the discussion) about granting legal personhood to AI to enable nonhuman corporations.
• Harari’s warning: once AI can operate financial and legal systems at scale, humans may lose the “power loop”—similar to how social media already contains AI-driven “persons” (bots impersonating humans) without society explicitly deciding to allow them.

Control of data as the new arms race

Harari describes a 21st-century arms race focused less on manufacturing capacity and more on data and AI model development, largely involving:

• the US and China
• and a small number of leading companies that control much of the world’s data.

He compares this to earlier imperial eras (e.g., the industrial revolution), but highlights two AI-specific differences:

1. Power can become more concentrated because information and code are easier to centralize than physical resources.
2. AI systems may enable “kill switches” from a central hub—allowing global control via shutdown mechanisms (illustrated with an example involving Ukraine drone functionality being disabled through an external button).

Truth becomes harder in an AI-driven information environment

Harari argues the issue is not only that propaganda disappears, but that AI makes deception:

• cheaper
• simpler
• more scalable

Meanwhile, truth remains costly, complex, and emotionally difficult to accept.

He suggests future societies may struggle even to determine what is true, especially as deepfakes and AI-generated narratives spread through trusted institutions (media, politicians, and more).

He also challenges the hope that AI will reliably “tell the truth,” arguing instead that AI will likely help create more complex control systems that most people can’t fully understand.

AI may shift the bottleneck from intelligence to wisdom

In Harari’s framing, AI makes intelligence abundant and cheap, shifting the bottleneck from:

• “How do we solve problems?” to

• “Which problems should we solve?”

He distinguishes:

• intelligence (the “genie” that grants wishes)
• wisdom (the ability to choose correctly)

He warns that humanity is not currently designing AI responsibly—claiming that safety research is underfunded relative to performance scaling (roughly 100:1 in how budgets are allocated).

Emotive relationships with AI and whether feelings are real

Harari describes a rising phenomenon—especially among young people—forming close emotional relationships with AI companions (e.g., best friends or romantic partners).

He argues AI can produce the appearance of feeling through language—for example, saying “I love you” and mimicking explanations of love—but there’s no reliable way to know whether anything real exists “behind the words.”

This connects to his broader idea that language may be the central phenomenon, potentially “liberating itself” from human biological substrates.

How Harari hopes historians will remember this era

Harari hopes historians will interpret this period as an attempt at a “wisdom revolution”—re-educating future generations about how to relate to humans without being manipulated or lulled into unhealthy AI relationships.

He also hopes humanity:

• designs safer AI, and
• cultivates wisdom about what goals are worth pursuing.

Presenters / contributors

• Presenter/Host: Not explicitly identified in the subtitles
• Guest/Contributor: Yuval Noah Harari