MUST READ | Anthropic researcher calls for AI slowdown, warns superhuman systems could leave humans jobless
Sam Altman warns against handing over control to AI
Altman said he and Brockman strongly disagreed with the idea of a gradual handover of control to AI. The concern is linked to the possibility of recursive self-improvement (RSI), where AI systems could help develop newer and more capable versions of themselves, potentially making it harder for humans to retain control.
The OpenAI CEO said preventing such a transition would require companies to account for competitive pressures and the possibility of both accidental and intentional transfers of control.
DON'T MISS | Salesforce layoffs in India: Employees get garden leave, share exit terms and check severance
Altman also warned that the industry cannot approach the next phase of AI development in the same way it has approached the previous few years.
"If the world and the companies developing AI treated the next four years like they've treated the last four years, I think that would be very bad," he said.
AI safety and monitoring need to keep pace with development
The growing capabilities of AI models are also making it harder for developers to monitor how these systems operate. Brockman explained that existing chain-of-thought monitoring, which lets researchers examine the reasoning models write out while solving problems, may become less effective as advanced models do more computation without explicitly showing their thought process.
Altman also called for greater caution across the industry, arguing that companies need shared safety standards rather than operating without enough understanding of one another's development practices.
The concerns come after a security incident involving an OpenAI model and Hugging Face. Brockman said the company had paused 25% of its production engineering projects during a security sprint, redirecting engineers towards securing its systems using AI models.
Altman also made his position on human agency clear, saying, "I don't want to be a Labrador. I want my kids to be Labradors." He was referring to a future in which humans could lose their agency as AI becomes more powerful. He said people should retain the ability to make decisions, invent things and shape their own future.