'Don't cede control of future to superintelligence': Altman
What's the story
OpenAI CEO Sam Altman has stressed the importance of human control over the future development of artificial intelligence (AI). He believes that the rise of superintelligent AI systems should be limited if they endanger our ability to make decisions about our own lives. In a post on X, Altman dismissed the idea of a future where humans do tasks based on what AI deems worthy.
Control assertion
AI must serve humans, not vice versa: Altman
Altman stressed that "people should always be in control of the future," and that "people should not cede control of the future to superintelligence."
He further added that this requirement should ultimately limit AI progress.
His comments echo similar sentiments expressed at the Adipec energy conference in Abu Dhabi on November 2, where he said AI must serve humans, not vice versa.
Development debate
Altman's comments echo Amodei's September essay
Altman's comments come nearly two months after Anthropic CEO Dario Amodei asked AI developers to slow down their model capabilities.
In a September 12 essay, Amodei warned that more capable AI systems could outpace researchers' understanding and control over them.
Altman supported this proposal, saying OpenAI would also give independent evaluators employee-level access to assess the company's safety practices.
Control concerns
Altman warned against concentrating advanced AI capabilities in few firms
Altman raised similar concerns at the United Nations Security Council on September 23, warning against developing AI systems without sufficient evidence they could remain under human control.
He also warned against concentrating advanced AI capabilities in a few companies or countries, saying no one entity should be able to impose its worldview through such technology.
Policy stance
OpenAI calls for international technical standards covering human oversight
In a policy document published on September 21, OpenAI said fully autonomous recursive self-improvement should not be pursued unless it could be done safely.
The company called for international technical standards covering human oversight, model evaluations, and incident reporting.
To further bolster its cybersecurity efforts, OpenAI has partnered with Cathedral Systems for an internal security exercise to identify vulnerabilities and test defenses against emerging AI-related threats.