Ilya Sutskever forms new AI startup to pursue 'safe superintelligence'

Former OpenAI chief scientist promises efforts will be insulated from commercial pressures

Ilya Sutskever, the Israeli-Canadian deep learning pioneer and former co-founder and chief scientist at OpenAI, has launched a new startup company, Safe Superintelligence Inc.

In an announcement on a simple web page and on X on Wednesday, Sutskever said the new company would be solely focussed on creating a "safe superintelligence," a superintelligence being an AI that surpasses human intelligence in some recognised definition of that word.

"We plan to advance capabilities as fast as possible while making sure our safety always remains ahead," Sutskever wrote.

"Our singular focus means no distraction by management overhead or product cycles, and our business model means safety, security, and progress are all insulated from short-term commercial pressures."

The latter statement is perhaps telling. Sutskever was one of the OpenAI board members who sought to oust CEO Sam Altman in November, a management coup that ultimately failed after a majority of employees, including eventually Sutskever himself, demanded his reinstatement following an intervention by Microsoft, a major investor in OpenAI.

Sutskever finally left OpenAI in May, without ever going public on the reasons behind his decision to quit, or for his original support for moves to replace Altman.

OpenAI began in 2015 as a non-profit with the goal of conducting research and development into artificial general intelligence (AGI) with the goal that it be built in a way that benefits humanity as a whole, but the company changed its focus over the years. Now very much a for-profit company, former employees have accused OpenAI of sidelining AI safety in pursuit of financial reward.

SSI currently consists of three people: Sutskever, Daniel Levy, another former OpenAI employee, and serial AI investor Daniel Gross. The post says that the startup is an American company based in Palo Alto, California, and Tel Aviv, and that it is "assembling a lean, cracked (sic) team" of the world's "best engineers and researchers dedicated to focusing on SSI and nothing else".

There is no public information about any investors in the company.

SSI is not the first startup to be founded by ex-OpenAI employees. In 2021 Anthropic was founded as a public-benefit company to deploy safe AI models by Daniela Amodei and Dario Amodei. The company, which produces the Claude series of AI chatbots, later attracted multi-billion-dollar investments from Google and Amazon.