"People should not cede control of the future to superintelligence"
OpenAI Chief Executive Officer (CEO) Sam Altman posted on X on the night of Nov. 3 (the morning of Nov. 4 Japan time) that "people should always be in control of the future." He stated that OpenAI is against the notion that it would be acceptable for people to be relegated to a role of doing things that "AI systems think makes sense to reward in humans." He also wrote that people should not cede control of the future to superintelligence (AI that far exceeds human intelligence) and that this requirement should ultimately constrain AI progress.
People should always be in control of the future. OpenAI is against the notion that it would be acceptable for people to be relegated to a role of doing things that “AI systems think makes sense to reward in humans”. People should not cede control of the future to superintelligence, and that requirement should ultimately constrain AI progress.
The post is three sentences long and does not give specific measures or the conditions under which development would be held back.
Quoted passage, no source given
In the second sentence, "AI systems think makes sense to reward in humans" is enclosed in quotation marks in the original. Altman did not say whose words they were and did not name any specific person or company.
Some users on X have pointed out that the passage comes from "Machines of Loving Grace," an essay published in October 2024 by Anthropic CEO Dario Amodei. The claim is unconfirmed, and Altman has not disclosed the source of the passage.
An argument continuing from September's U.N. Security Council
Altman has said before that AI should be kept under human control. At a meeting of the U.N. Security Council on Sept. 23 (U.S. Eastern Time), he said that as AI capabilities grow, people must be at the center of AI decision-making. Amodei, who joined the same meeting by video, warned that if AI improvement continues at the current pace, people could lose control of it, and expressed support for efforts to manage the speed of development in a coordinated way.
According to India's ANI news agency, Amodei called on the industry in September to slow development, and Altman agreed, saying he shared Amodei's view that the pace of frontier development needs to be calibrated. He also supported allowing independent evaluation bodies to verify the safety of AI systems, it said.
Fortune reported on Sept. 12 that Altman said in an interview that he would accept pausing or stopping development even if investors objected. The magazine reported that OpenAI's IPO would be in 2027 or later.
Squaring it with the speed of development is the focus
OpenAI continues to develop new models. Altman was reported to have said at an exhibition in Abu Dhabi on Nov. 2 that an internal model that produced a proof of the Navier–Stokes equations would soon be released as "the next GPT" (related article).
Citing a report by The Washington Post, ANI pointed out that OpenAI, Anthropic and other major companies have continued to release high-performance models even after calls for a slowdown. It said the question is how safety commitments are reflected in actual development.
OpenAI and Anthropic also differ in their views on regulation. In an Oct. 4 interview with Politico, Altman said he would not accept truly catastrophic risks or a serious loss of control to AI, while saying there is a pretty big gap between OpenAI and Anthropic on regulation.
