OpenAI details its plan for an AI "for the benefit of all"

Sam Altman and Jakub Pachocki outline OpenAI's plan to build a personal AGI for every individual and automate research with humans by March 2028.

Sam Altman and Jakub Pachocki co-authored a programmatic text on OpenAI's direction, centered on an idea: the benefits and power associated with AI must be widely distributed rather than concentrated among a few players, a condition, they believe, for a safer future and a more resilient society.

This plan is structured around three objectives. First, an automated researcher, a system intended to accelerate and then partially automate research while remaining controllable and connected to humans, with an internal conviction: by March 2028, a significant portion of OpenAI's work could be conducted by AIs alongside its researchers. Next, accelerate the economy and science by sharing the gains. Finally, equip each individual with a personal AGI.

The text also sets a limit: full automation is not the intended horizon. Systems must remain safe, aligned with human intent, and under control, with the role of people shifting towards arbitration and choosing what is worth doing. OpenAI reiterates its commitment to an international coordination body, capable, if necessary, of slowing down cutting-edge development to ensure safety keeps pace.

The company presents this as its third phase, following research and product: making AI abundant, affordable, and accessible to everyone.