London: When will artificial intelligence truly rival the human mind if ever? That question stirred sharp debate among the world’s leading AI researchers gathered at the FT Future of AI Summit in London, where pioneers of the field offered strikingly different visions of the future.
The event, featuring Geoffrey Hinton, Yoshua Bengio, Fei-Fei Li, Yann LeCun, and NVIDIA leaders Jensen Huang and Bill Dally, showcased not only divergent predictions but also deeper disagreements about what it means for machines to “match” human intelligence and whether such a goal even matters.
All six experts, recipients of the 2025 Queen Elizabeth Prize for Engineering, stand as architects of modern AI. Yet their perspectives spanned a wide spectrum from optimism that artificial general intelligence (AGI) could arrive within two decades to outright dismissal of the question’s relevance.
Geoffrey Hinton, often dubbed the “Godfather of AI,” made the boldest forecast, asserting that AGI could emerge within 20 years. “If you define it as a machine that could out-argue you in a debate we’ll likely get there in less than two decades,” he said.
Reflecting on his 1984 creation of an early neural model trained on just 100 examples, Hinton noted, “It took 40 years to reach where we are because we lacked computing power and data. Now we have both.”
His optimism stems from the exponential pace of AI development and the momentum of generative models that increasingly simulate complex reasoning and communication.
Yoshua Bengio offered a more cautious outlook, recognising AI’s astonishing progress while warning against overconfidence. “AI’s ability to plan across time horizons has exploded in the last six years,” he said. “If that trajectory continues, AI could reach the level of a human employee in about five years.”
Yet Bengio added a note of humility: “There’s too much uncertainty to make grand claims. We should stay agnostic there are many possible futures.” His remarks highlighted that while the pace of innovation is dazzling, the path toward true cognition remains unpredictable.
Yann LeCun, Meta’s Chief AI Scientist, struck a skeptical tone, arguing that AI’s evolution will be gradual rather than revolutionary. “It’s not going to be an event,” he said. “Capabilities will expand progressively over the next five to ten years, but real progress will take longer than people expect.”
He pointed out a stark limitation: “We don’t have robots that are as smart as a cat.” LeCun stressed that AI’s future depends not on more computing power or data, but on scientific breakthroughs in understanding perception, learning, and reasoning.
NVIDIA CEO Jensen Huang dismissed the notion of AGI timelines altogether. “It’s an academic question,” he said. “What matters is applying AI to solve real problems it’s already transforming industries.”
He contrasted today’s AI boom with the dotcom era, noting that “every GPU in the world is lit up and working.” For Huang, the future lies in AI infrastructure powering trillion-dollar industries, not philosophical debates about intelligence.
NVIDIA’s Chief Scientist Bill Dally echoed this pragmatic approach. “Our goal isn’t to replace humans,” he said. “It’s to build AI that augments human capability enabling people to be more creative, empathetic, and collaborative.”
Fei-Fei Li, Stanford professor and visionary behind the ImageNet dataset, took a more nuanced position. “Parts of machines will surpass human intelligence, but they’ll never replicate it,” she said. “They’re built for different goals.”
Li highlighted how AI already exceeds human ability in some domains “no human can recognise 22,000 objects or translate 100 languages” yet still falters in others. She pointed to “spatial intelligence” as a frontier that large language models fail to master, underscoring that human cognition is rooted not only in language but also in perception, creativity, and embodied understanding.
Before addressing human-level AI, the panel debated whether today’s AI surge marks a lasting transformation or an unsustainable bubble. Huang argued that unlike the dotcom era, “this time, everything we build is being used.” He emphasized that AI is reshaping productivity itself: “AI is intelligence that augments people it works.”
LeCun, however, cautioned that there is indeed a “bubble” around the hype of large language models (LLMs). “Pushing the current paradigm to achieve human-level intelligence is misguided,” he said.
Despite their disagreements, all six experts agreed on one point AI has already become a civilisational force. “This is a technology that will impact every individual and every sector,” said Fei-Fei Li.
Whether machines will ever think or reason like humans remains uncertain and perhaps irrelevant. As Jensen Huang put it, “The answer doesn’t matter.” What does matter, the panelists agreed, is that artificial intelligence has already begun reshaping what it means to work, create, and live in the 21st century.