Hinton, Li, and Ng urge nuanced openness in AI amid safety debates at Ai4

At the Ai4 conference in Las Vegas, three of the field’s most prominent researchers—Geoffrey Hinton, Fei-Fei Li, and Andrew Ng—spoke in favor of nuanced approaches to openness in AI, rather than absolutist positions. Their remarks came as concerns about AI safety and control intensify alongside rapid advances in the technology.

Andrew Ng, co-founder of Coursera, argued that gatekeepers in AI could replicate dynamics seen in mobile platforms, where a handful of companies control access and shape what gets built. “I don’t want there to be gatekeepers,” Ng said. “That limits how all of us can access AI.” He advocated for promoting openness to ensure broad access and prevent a small set of firms from dominating the field.

Geoffrey Hinton, a Nobel laureate and former Google researcher, distinguished between open-source software, which exposes code for inspection, and open-weight models, which release trained model parameters. While he expressed reservations about open-weight models due to misuse risks such as cyberattacks, he acknowledged their inevitability. “I think that battle’s been lost,” Hinton said. “We now have open-weight models, so the barrier to lots of people getting these big models has disappeared.” He also emphasized the need for regulation to guide AI development toward beneficial outcomes.

Fei-Fei Li, CEO and co-founder of World Labs, rejected a false dichotomy between complete openness and complete closedness. She pointed to nuclear physics—where research is public but uranium is regulated—as an example of layered control. Li also cited the Human Genome Project, where open scientific findings enabled private-sector innovation. “We need some levels of openness, both in scientific discovery, in education, in global partnership, as well as lucrative business models for entrepreneurs,” she said. “But we also will accept closed-source systems.”

All three speakers agreed that some form of regulation is necessary to steer AI toward socially beneficial outcomes. Hinton cautioned against leaving such decisions to a small group of tech leaders, stating, “You can’t leave it to people like Elon Musk and Mark Zuckerberg to decide how AI should be done.”