Three AI Pioneers Clash Over Open Models at Las Vegas Conference

Key points

• Geoffrey Hinton, Fei-Fei Li, and Andrew Ng each argued against total corporate control of AI at the Ai4 conference in Las Vegas.
• Hinton said the battle over open-weight AI models, ones where the trained model itself is released free to the public, is already lost.
• Andrew Ng warned that Chinese open-weight models could become the dominant AI infrastructure across Asia, Africa, and the developing world.
• Fei-Fei Li rejected the framing of open versus closed as a false choice, calling for a nuanced, layered approach.
• All three agreed some form of regulation is necessary, with Hinton saying the decisions cannot be left to tech billionaires alone.

Three of the most influential names in artificial intelligence history shared a stage in Las Vegas last week, and they could not fully agree. But on one point they were united: a small group of powerful companies should not get to decide how AI develops, or who gets to use it.

The venue was the Ai4 conference, first reported by TechCrunch AI. The speakers were Geoffrey Hinton, who won the 2024 Nobel Prize in Physics for his foundational work on AI; Fei-Fei Li, the Stanford professor and CEO of World Labs who helped build the image-recognition research that powered today's AI boom; and Andrew Ng, co-founder of Coursera and one of the most widely followed voices in the field.

What is the open-weights debate actually about?

Open-weight AI means a company trains a large AI model and then releases it free for anyone to download, modify, and run. That is different from open-source software, where you can read and edit the underlying code. With open weights you get the finished brain, not the blueprint.

Hinton drew that line clearly. "Open source is great. You show people the code, and lots of people look at the lines of code and say, 'Oh, there's a bug,'" he said. "Open weights means you train a big model and then you give people the weights. That's very different."

His worry: once a powerful model is free to download, anyone can spend a fraction of the original training cost to teach it to do harmful things, such as helping plan a cyberattack.

Yet Hinton also conceded the debate is over. "That battle's been lost," he said. Open-weight models exist, the barrier of enormous training costs has gone, and no announcement from any lab will change that.

Why does this matter for ordinary people?

The practical stakes are large. Ng framed it as a geopolitical contest. Chinese open-weight models are already spreading across Africa and Asia. If they become the default AI infrastructure for billions of people, he argued, those models could shape how entire populations encounter ideas about democracy and human rights.

"AI is a tremendous source of soft power," Ng said, adding that lobbying and fear-mongering in the United States are making it harder for American open-source AI to compete on cost.

Li pushed back on framing the whole question as a binary fight. She compared AI to nuclear physics: scientific papers are published openly, raw uranium is tightly regulated, and laboratory work sits somewhere between the two. Wholesale openness and complete lock-down are both the wrong answer.

"This debate, especially at the sweeping level of 'we can only tolerate one,' is a false debate," she said. "We need to get to a level of nuance."

What happens next?

Regulation, all three agreed, is unavoidable. The disagreement is over the details. Hinton was blunt about who should not be in the room when those rules get written. "You can't leave it to people like Elon Musk and Mark Zuckerberg to decide how AI should be done," he said.

For anyone who uses AI tools at work, at school, or at home, the outcome of this debate will decide which models are available, at what price, and with what restrictions baked in.