Machines may soon think in a language we don’t understand, leaving humanity in the dark: Godfather of AI sounds alarm

Geoffrey Hinton, often dubbed the “Godfather of AI,” has once again sounded a sobering alarm about the direction in which artificial intelligence is evolving. In a recent appearance on the One Decision podcast, Hinton warned that AI may soon develop a language of its own — one that even its human creators won’t understand.

“Right now, AI systems do what’s called ‘chain of thought’ reasoning in English, so we can follow what it’s doing,” Hinton explained. “But it gets more scary if they develop their own internal languages for talking to each other.” He went on to add that AI has already demonstrated it can think “terrible” thoughts, and it’s not unthinkable that machines could eventually think in ways humans can't track or interpret.

A Visionary’s Regret and Revelation

Hinton’s warnings carry weight. The 2024 Nobel Prize laureate in Physics, awarded for his pioneering work on neural networks, has helped lay the foundation for today’s most advanced AI systems, including deep learning and large language models.


But today, Hinton is wrestling with what he calls a delayed realization. “I should have realised much sooner what the eventual dangers were going to be,” he said. “I always thought the future was far off and I wish I had thought about safety sooner.”

That hindsight is now driving his advocacy. Hinton believes that as digital systems become more advanced, the gap between machine intelligence and human understanding will widen at a staggering pace.

Machines That Learn and Learn Together

One of Hinton’s most compelling concerns is how digital systems differ fundamentally from the human brain. AI models, he says, can share what they learn instantly across thousands of copies. “Imagine if 10,000 people learned something and all of them knew it instantly — that’s what happens in these systems,” he explained on BBC News.

It’s this kind of distributed, collective intelligence that could soon allow machines to outpace even our most ambitious understanding. AI models like GPT-4 already surpass humans in general knowledge, and though they lag in complex reasoning for now, Hinton says that gap is closing fast.

A Whisper Network in Tech?

While Hinton has made waves by speaking openly about AI risks, he says others in the tech world are staying quiet — at least in public. “Many people in big companies are downplaying the risk,” he noted, despite their private concerns. One exception, he says, is Google DeepMind CEO Demis Hassabis, who has shown serious commitment to addressing those risks.

Hinton's own exit from Google in 2023 was widely misinterpreted as a protest. He now clarifies, “I left Google because I was 75 and couldn’t program effectively anymore. But when I left, maybe I could talk about all these risks more freely.”

With AI’s capabilities expanding and governments scrambling to catch up, the global conversation around regulation is intensifying. The White House recently unveiled an “AI Action Plan” aimed at accelerating innovation while limiting funding to overly regulated states.

But for Hinton, technical advancements must go hand in hand with ethical guardrails. He says the only real hope lies in finding a way to make AI “guaranteed benevolent” — a lofty goal, given that the very systems we build may soon be operating beyond our comprehension.