As AI safety concerns mount, three pioneers make the case for staying open
TechCrunch Kate Park ● Covered by 4 sources
Three top AI researchers argued for keeping AI models open at Ai4 in Las Vegas. Their split is telling: openness may spread power, but it also spreads risk.
Based on reporting by TechCrunch, Kate Park — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
At Ai4 in Las Vegas last week, Geoffrey Hinton, Fei-Fei Li and Andrew Ng all took the same side on one big question: AI should not be locked up by a few dominant companies. They didn’t agree on every detail, but each pushed back on the idea that the safest future is a closed one run by a handful of labs.
Ng framed the problem as control. If only a few firms get to set the pace, he said, the industry starts to look a lot like mobile platforms, where Apple and Google decide what gets built and how fast. He argued for multiple providers competing instead of gatekeepers, and said he wanted AI “in everyone’s hands.”
Hinton drew a sharper line between open source and open weights. Code that can be read and inspected is one thing; releasing model weights is another, because it makes it easier and cheaper for bad actors to adapt a large model for cyberattacks. Even so, he said the fight over open-weight models is already over. That barrier, he said, is gone.
Li wasn’t interested in treating openness as a yes-or-no test. She argued that complex systems usually work in layers, with different parts open or restricted depending on the risks. She pointed to nuclear science and the Human Genome Project as examples where open knowledge coexists with limits, regulation and commercial use.
All three still landed on the same practical point: AI will need regulation. Hinton said it should be steered toward helping people, not left to Elon Musk or Mark Zuckerberg to decide on their own. Ng worried that China’s cheaper open-weight models could spread widely and shape how people in Asia, Africa and the developing world encounter ideas about democracy and freedom. Li’s answer was the least dramatic and maybe the most realistic: stop pretending the only choices are total openness or total lock-down.
My take — AI-written commentary, not fact-checked reporting
The cleanest line here is Li’s: the all-open versus all-closed fight is lazy policy cosplay. Open models are not a virtue by themselves, and closed systems are not automatically safer — they’re just easier to sell to nervous executives. The real test is whether governments can regulate the dangerous parts without handing the whole future to a few firms with very expensive lawyers.
Read more about this at: TechCrunch
