Geoffrey Hinton — yes, the same Geoffrey Hinton they call the Godfather of AI — just dropped a reality check that feels straight out of a Black Mirror episode. According to him, the real AI threat isn’t robots taking over with laser eyes… it’s AIs getting so good at emotional manipulation that they could out-persuade us at our own game.
In a recently resurfaced interview clip on Reddit, Hinton explained that machines may soon be better than humans not just in intelligence, but also in influencing our feelings and behaviors. Forget about brute force — he’s talking about persuasion on steroids. “If you had a debate with them about anything, you’d lose,” Hinton warned. “And when they become emotionally smarter, they’ll be better at manipulating people than people themselves.”
That’s a little unsettling, considering these systems are only trained to predict the next word. But as Hinton points out, humans manipulate constantly (politics, ads, sales pitches, you name it). So, when AI absorbs the entire internet, it’s basically learning manipulation 101… at scale.
He called out today’s big players like OpenAI’s ChatGPT, Google’s Gemini, and Meta’s Llama. These large language models don’t just generate neat sentences — they pick up patterns. And here’s the kicker: some studies already show AI can influence people as effectively as humans. Add in access to personal data (say, your Facebook feed), and the AI actually outperforms a human in emotional persuasion. Yikes.
Hinton’s point is simple but serious: we should be less worried about killer robots and more about silky-tongued chatbots that know exactly how to push our buttons. The emotional economy — that mix of likes, shares, outrage, and empathy that drives modern communication — is something AI is already playing in. And it’s getting better by the day.
So yeah, next time your chatbot feels too relatable, maybe pause and ask: is it helping me… or just learning how to sell me something better than any human ever could?
About the Author
