“If you define that as AGI (artificial general intelligence), we’re probably going to get there in less than 20 years,” said Geoffrey Hinton, often hailed as the Godfather of AI. Hinton was talking at the FT Future of AI Summit in London where he was joined by NVIDIA CEO Jensen Huang, AI researcher Yoshua Bengio, Stanford Professor Fei-Fei Li, Meta’s chief AI scientist Yann LeCun, and NVIDIA chief scientist Bill Dally.At the summit, the world’s AI pioneers were asked how long until the world saw machine intelligence achieved human-level intelligence. Interestingly, their answers showed not just the timeline debate, but some fundamental disagreements about what artificial intelligence can and should become.
The speakers, all of whom are the 2025 Queen Elizabeth Prize for Engineering laureates, are the architects of modern AI. However, their perspectives on reaching human-level intelligence diverge dramatically, from “we’re already there” to “it’s the wrong question entirely.”
Story continues below this ad
A question without consensus
At the summit, Hinton offered perhaps the most concrete prediction. “If you refine the question to say, ‘How long before if you have a debate with this machine it’ll always win?’ I think that’s definitely coming within 20 years,” he said.
Hinton’s timeline reflects his pioneering work on backpropagation, which he traced back to 1984 when he created what he describes as “a tiny language model” with just 100 training examples. “It took 40 years to get us here,” Hinton said, “and the reason was we didn’t have the compute and we didn’t have the data.”
On the other hand, Bengio took a more nuanced approach, acknowledging both rapid progress and some fundamental limitations. He highlighted the rapid growth in AI’s planning capabilities. “The capability of AI to plan over different horizons has grown exponentially fast in the last six years. If we continue that trend, it would place AI at roughly the level an employee has in their job within about five years.”
However, he asserted that uncertainty is inherent in such predictions, adding, “I’m not saying it will happen, but we should be agnostic and not make big claims because there’s a lot of possible futures there.”Story continues below this ad
Meanwhile, Yann LeCun’s response was the most subdued. “It’s not going to be an event,” he cautioned. “The capabilities are going to expand progressively in various domains over maybe the next five to ten years to come up with a new paradigm. Then progress will come, but it’ll take longer than we think.”
LeCun’s skepticism comes from a fundamental gap he sees in current AI. “We don’t have robots that are nearly as smart as a cat,” he said, highlighting that “AI progress is not just a question of more infrastructure, more data, more investment; it’s actually a scientific question of how we make progress towards the next generation of AI.”
The wrong question
The premise was not accepted unanimously. NVIDIA’s Jensen Huang dismissed the question’s relevance entirely. “At this point, it’s a bit of an academic question,” he argued.
“We’re going to apply the technology, which will keep getting better, to solve a lot of very important things from this point forward. The answer is it doesn’t matter.” Huang’s stand reflects NVIDIA’s central role in powering AI infrastructure, where practical application trumps philosophical notions.Story continues below this ad
Bill Dally, NVIDIA’s chief scientist, echoed this sentiment. “Our goal is not to build AI to replace humans or to be better than humans. Our goal is to build AI to augment humans, to complement what humans are good at.”
He asserted that the objective is enabling humans “to do what is uniquely human such as being creative, empathetic, and understanding how to interact with others in our world.”
Human intelligence persists
Fei-Fei Li, whose ImageNet dataset has been instrumental in the deep learning revolution, offered a crucial contrast to the human-versus-machine debate. “Parts of machines will supersede human intelligence, and part of machine intelligence will never be similar to human intelligence. They’re built for different purposes,” she explained.
She pointed out that machine capabilities already exceed human performance in specific domains: “How many of us can recognise 22,000 objects in the world with granularity and fidelity? How many adult humans can translate 100 languages?”Story continues below this ad
Earlier in the talk, Li also highlighted a critical frontier, spatial intelligence. “Even today’s most powerful LLM-based models fail at rudimentary spatial intelligence tests,” she noted, highlighting that human intelligence extends far beyond language to encompass “the ability to perceive, reason, interact with, and create worlds.” This shows that achieving true human-level AI requires breakthroughs beyond scaling current language models.
Is AI a bubble or beginning?
Before discussing human-level AI, the panel tackled a more immediate concern – whether the current AI boom represents a sustainable revolution or an inflated bubble. In this context, Huang drew a stark contrast with the dotcom era. “During the dotcom era, the vast majority of fiber deployed was dark, the industry deployed far more than it needed. Today, almost every GPU you could find is lit up and used.”
He emphasised a fundamental shift in computing: “For the first time, AI is intelligence that augments people. It addresses labor, addresses work, does work.” This transformation, he claimed, justifies massive infrastructure investment. “We need hundreds of billions of dollars of these factories to serve the trillions of dollars of industries that sit on top of intelligence.”
Regardless, Yann LeCun identified a potential bubble within the broader revolution. “There is a sense in which there is a bubble – the idea that the current paradigm of LLMs would be pushed to the point of having human-level intelligence. I personally don’t believe in that.”Story continues below this ad
While the conversations were rife with disagreements on timelines and definitions, the pioneers unanimously recognised AI’s transformative impact. “This is a civilisational technology that’s going to impact every single human individual and sector of business,” Fei-Fei Li observed.
Though the question of when AI will match human intelligence may never have a single answer — or may not even be the right one to ask, what remains clear, as the pioneers of the AI revolution emphasise, is that the technology’s impact on civilisation is already undeniable.