Six of the most influential figures in artificial intelligence took the stage at the Queen Elizabeth Prize 2025 ceremony to weigh in on recent breakthroughs, ongoing challenges, and whether current AI models are really as advanced as they seem.
It was an all-star lineup, as Bill Dally, Yoshua Bengio, Fei-Fei Li, Yann LeCun, Jeff Hinton, and Nvidia CEO Jensen Huang gathered for a panel on the breakthroughs and hype shaping today's AI landscape.
The conversation touched on the GPU boom, large language models, self-supervised learning, and the impact of datasets like ImageNet. While the panel agreed that AI will change daily life, they also made it clear that progress won't be immediate and many of the hardest problems remain unsolved.
Bengio argued that today's reasoning-based models have moved so far from their original function that "LLM" is no longer an accurate name. With multi-modal inputs, more interactivity, and tool use, these systems are evolving into agents. "It's not at all the same thing as what it was three years ago. I don't think we can predict where the technology will be in two years, five years, ten years," Bengio said.
Huang, of course, maintained that AI isn't a short-lived bubble, but the backbone of a new industry with a growing need for data centers. Others on the panel were more skeptical about the hype. Li and LeCun warned against expecting anything close to human-level intelligence soon, pointing out the major scientific roadblocks still ahead.
"We're missing something big still," LeCun said, adding that LLMs won't reach human intelligence, let alone anything like superintelligence. "That's why AI progress is not just a question of more infrastructure, more data, more investment, or further development of the current paradigm," LeCun continued. "It's actually a scientific question of how we make progress toward the next generation of AI."
