Sam Altman Aims to Slow Down AI, but Promises Steady Progress

The head of OpenAI advocates for a measured slowdown and a national safety framework for advanced AI, while ensuring that advancements will remain rapid. Meanwhile, major players are discussing an independent self-regulatory body, a prospect that alarms smaller competitors.

Proposed Self-Regulation and Fear of a Cartel

OpenAI, Anthropic, and Google have been discussing the creation of an independent oversight body for the industry for several months, according to a source. Smaller companies, such as Cohere, fear that this initiative could lead to the formation of a cartel.

OpenAI Tightens Its Safeguards and Altman Outlines His Red Lines

Before each training phase that could lead to major advancements in capabilities, OpenAI introduces clearly defined safety protocols. Sam Altman advocates for the establishment of a uniform safety framework at the national level to regulate advanced AI, while maintaining that the sector should not wait for legislative action. He believes that international coordination requires support from the U.S. government and hopes that other companies will take inspiration from OpenAI's methods to develop their own. In his view, slowing down development does not mean stopping it: advancements will continue to be rapid, but at a pace lower than what would be possible. He argues that this slowdown would yield far more benefits than costs, that no internal competition in the United States should lead to negligence, and that capabilities must remain below the alignment and oversight levels required.

Several Leaders Support More Controlled Development

Over the weekend, Dario Amodei, Demis Hassabis, Satya Nadella, and Elon Musk all expressed their support for more controlled AI development.