AI pioneers clash over the future of open-weight models amid safety debate
At the Ai4 conference in Las Vegas, three prominent AI researchers — Geoffrey Hinton, Fei-Fei Li and Andrew Ng — offered differing views on how open AI development should be. All three agreed, however, that some level of regulation is needed.
As major AI labs increasingly position themselves as the safer alternative to open-weight models, three influential researchers used the recent Ai4 conference in Las Vegas to make the case for keeping AI development open. The panel included Nobel laureate Geoffrey Hinton, World Labs CEO Fei-Fei Li, and Coursera co-founder Andrew Ng.
All three expressed concern that a small number of large companies could end up controlling the pace of AI progress, drawing a comparison to how Apple and Google dominate mobile operating systems. Ng said he did not want to see "gatekeepers" emerge in AI, arguing this would limit how widely people could access the technology, and called for maintaining multiple competing providers and models.
A distinction between open source and open weights
Hinton drew a sharp line between open-source software, where code can be inspected and fixed, and so-called open weights, where the trained parameters of a model are released publicly. In his view, this makes it easier for people to take expensive, already-trained foundation models and repurpose them for harmful uses such as cyberattacks. Still, he acknowledged that this battle has effectively already been lost, since open-weight models are now a permanent feature of the AI landscape.
Ng focused more on the geopolitical stakes, warning that if China's open-weight models gained wide adoption across Asia, Africa and the developing world, they could shape how billions of people encounter ideas about democracy and human rights. He argued that excessive lobbying and fear-driven caution in the U.S. are holding back domestic open-source AI development relative to China's offerings.
Li rejected the idea that the choice must be between total openness and total closure, comparing the situation to nuclear physics, where research is published openly even as materials like uranium remain tightly regulated. She pointed to the Human Genome Project as an example of open collaboration that became a foundation for further scientific and commercial progress.
Despite their disagreements, all three researchers agreed that some degree of regulation will be necessary to steer AI development in a beneficial direction for society.
