OpenAI co-founder Ilya Sutskever's new company has raised $1 billion and announced plans to build a team of "the world’s best engineers and researchers" that will focus on a single mission.
His firm is so dedicated to its goal that it is named after the only product it will develop. To achieve its ambition of creating an AGI that will benefit humanity, rather than destroying it, Safe Superintelligence Inc (SSI) will create an innovative working environment capable of solving "the most important technical problem of our time."
"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 firm wrote.
"SSI is our mission, our name, and our entire product roadmap, because it is our sole focus. Our team, investors, and business model are all aligned to achieve SSI."
Mountain: identified. Time to climb https://t.co/3iwzcbAdxw
— Ilya Sutskever (@ilyasut) September 4, 2024
The firm will address the safety of models and their capabilities "in tandem, as technical problems to be solved through revolutionary engineering and scientific breakthroughs".
"We plan to advance capabilities as fast as possible while making sure our safety always remains ahead,' they added.
It will build a "straight-shot SSI lab, with one goal and one product: a safe superintelligence". Its raise involved NFDG, a16z, Sequoia, DST Global, and SV Angel.
The history of Safe Superintelligence Inc.
Sutskever, 37, launched SSI in June with two co-founders called Daniel.
The first, Daniel Gross, worked on AI initiatives at Apple, and the second is former OpenAI researcher Daniel Levy, a former OpenAI researcher. Sutskever works as chief scientist, and Levy holds the position of principal scientist, while Gross leads fundraising and building up computing power.
Sutskever left OpenAI after sitting on the board that voted to boot out CEO Sam Altman before quickly reinstating him. In May, he officially left the company after a decade, describing its trajectory as "nothing short of miraculous" in an X post and saying he was "confident that OpenAI will build AGI that is both safe and beneficial."
However, it will do this without drawing on the resources of Sutskever's "Superalignment" team, which aimed to "steer and control AI systems much smarter than us" but has now been shut down.
He told Reuters the new venture began after he "identified a mountain that's a bit different from what I was working on."
Describing the people he'd like to work alongside, he said: "Some people can work really long hours and they'll just go down the same path faster. It's not so much our style. But if you do something different, then it becomes possible for you to do something special."
How is SSI different from OpenAI?
SSI will run as a normal for-profit business, as opposed to the unusual “capped profit” structure of OpenAI, which is a partnership between the original nonprofit and a new capped profit arm.
At OpenAI, all profit allocated to investors and employees, including Microsoft, is capped so that all residual value created above and beyond the limit is "returned to the nonprofit for the benefit of humanity". Read more about OpenAI's structure here.
In a briefing note, Deutsche Bank Research said SSI was "the antithesis" of OpenAI.
"By not planning for now to release commercial products, Safe Superintelligence enters an entirely different class of start-ups from the usual ‘build the product—sell the product’ type," it wrote. "This may clash with funding, the lifeblood of AI innovation because of the high cost of training and scarce talent. OpenAI, for instance, started as a research company."
The bank added: "There is a heated debate in AI between those who fret about the potential risks and are calling for slower progress, and those who want to go full steam ahead. With his own platform, Sutskever may gain a larger voice and emerge as something of an opposite to his former boss Sam Altman, the CEO of OpenAI."
SSI is looking for people who will fit into its culture and who have the skills and capabilities the company needs to grow, rather than a shiny shelf of certificates and credentials.
It plans to partner with chipmakers and cloud providers to satisfy its thirst for compute, but it hasn't decided to work with Nvidia, Microsoft, or another cloud competitor.
Sutskever is a proponent of scaling, a position which states that the performance of AI models can be increased by using more data, resources and neural network architectures. SSI will approach scaling differently to OpenAI, but it is not yet clear how.
If you can work in Palo Alto and Tel Aviv, you can apply for a position here.
"We offer an opportunity to do your life’s work and help solve the most important technical challenge of our age," SSI wrote.
