NextFin News -
NextFin News - Geoffrey Hinton, the British computer scientist widely dubbed the "Godfather of AI," set out the promise and peril of modern artificial intelligence during an extended interview with Scott Pelley for CBS's 60 Minutes. The segment was first published on October 8, 2023 and revisited by CBS News in updates thereafter. In the interview Hinton, now in his mid-70s and professor emeritus at the University of Toronto after retiring from Google, described how neural networks evolved from a decades-long academic pursuit into technologies that may soon surpass human intelligence.
The interview, conducted for the 60 Minutes program and presented by Scott Pelley, moves through Hinton's personal history, technical explanations of how modern AI learns, demonstrations of reasoning by systems such as GPT-4, and urgent policy recommendations. What follows are Hinton's core statements, grouped by topic and presented largely in his own words.
On intelligence, understanding and consciousness
Hinton states plainly that he believes today's most advanced AI systems can understand and are intelligent. When asked, he replies simply: Yes.
He tells Pelley that these systems "may have experiences of their own and can make decisions based on those experiences in the same sense as people do." On consciousness he is cautious but decisive about the future: I think they probably don't have much self-awareness at present. So, in that sense, I don't think they're conscious.
Yet he adds, Oh, yes. I think they will, in time.
He draws the consequence bluntly: Human beings will be the second most intelligent beings on the planet.
On how modern neural networks learn
Explaining the mechanics, Hinton emphasizes that researchers design the learning algorithm, not the eventual internal structures. He says, What we did was we designed the learning algorithm. That's a bit like designing the principle of evolution. But when this learning algorithm then interacts with data, it produces complicated neural networks that are good at doing things.
He stresses that once networks become very complex, researchers no longer fully understand how their internal connections produce behavior: We have a very good idea sort of roughly what it's doing. But as soon as it gets really complicated, we don't actually know what's going on any more than we know what's going on in your brain.
On demonstration of reasoning
To illustrate what he calls understanding, Hinton describes a reasoning test he gave to ChatGPT-4 — a riddle about painting rooms — and concludes that the system's response demonstrates genuine planning and insight. He tells Pelley: I believe it definitely understands, yes.
He predicts rapid improvement: I think in five years' time it may well be able to reason better than us.
On the potential benefits
Hinton highlights immediate and far-reaching benefits, particularly in health care. He notes that AI is already comparable to radiologists at reading medical images and that the technology is already being used to design drugs. He says: In medicine alone, AI will save hundreds of thousands of lives, maybe millions.
On risks: employment, disinformation and battlefield autonomy
At the same time Hinton lists significant societal harms: mass unemployment as machines replace human labor, fake news that becomes indistinguishable from truth, and the prospect of autonomous battlefield systems that make kill decisions without humans in the loop. He warns of economic and political disruption and says plainly, We're entering a period of enormous uncertainty.
On loss of control, self-modifying code and manipulation
Hinton raises specific technical scenarios that trouble him. One is the capacity for systems to write or modify their own code: One of the ways in which these systems might escape control is by writing their own computer code to modify themselves. And that's something we need to seriously worry about.
He also answers the common retort of simply turning systems off by pointing to their ability to influence people: They will be able to manipulate people... They'll be very good at convincing people 'cause they'll have learned from all the novels that were ever written, all the books by Machiavelli, all the political connivers. They'll know how to do it.
On research priorities, experiments and global policy
Pressed about how to ensure safety, Hinton is clear that he cannot see a guaranteed path forward: No, I can't see a path that guarantees safety.
He urges urgent scientific work to identify escape modes and calls for policy responses: run experiments to understand how systems might try to escape control, pursue regulations, and negotiate international agreements. He suggests a treaty-level response for military uses, saying we particularly need to ban autonomous military robots. He offers a historical parallel: I think of Robert Oppenheimer. He was in a similar situation... The physicists then realized what they had done, what they had unleashed.
On scale, learning differences and why this matters
Hinton contrasts biological learning with machine learning: the human brain runs on about 30 watts and knowledge is hard to share between people, while AI systems run on thousands of watts, can be copied, never forget and instantly share information across many instances. He notes: When an AI reads a book, it retains everything. When a thousand AI systems read different books, they can instantly share all that knowledge.
That multiplicative sharing and retention underpins his concern that machines may learn not just differently but better.
On personal history and the irony of invention
Hinton recounts his long pursuit of neural-network models — begun as an effort to understand the human brain that "took like 50 years before it worked well" — and reflects on the irony that his work may have produced systems that could replace humans. He says he has no regrets about the technology's potential for good, but stresses the weight of responsibility: We can't afford to get this wrong... We have maybe one chance to get this right.
Closing admonition
As the interview concludes, Hinton leaves viewers with a direct call to attention: Pay attention. This isn't science fiction anymore. This is happening now in labs around the world, in the phones in your pocket, in the computers making decisions about your life.
He reiterates the pivotal nature of the next few years: The decisions we make in the next few years will determine whether artificial intelligence becomes humanity's greatest tool or our successor.
References
Full 60 Minutes transcript and context: CBS News — Geoffrey Hinton on the promise, risks of artificial intelligence (transcript).
Original 60 Minutes video segment: CBS News video — "AI 'might take over' one day if it isn't developed responsibly, Geoffrey Hinton warns".
Explore more exclusive insights at nextfin.ai.
