Ilya Sutskever's Safe Superintelligence Teams Up With Nvidia to Scale Its Research

Safe Superintelligence (SSI) — the lab founded by former OpenAI co-founder and safety chief Ilya Sutskever — has emerged from the shadows for the first time in two years. The company has announced a long-term partnership with Nvidia intended to power its move to the next stage of development.

The agreement includes an investment whose size has not been officially disclosed; according to a source cited by TechCrunch, it amounts to several billion dollars. Nvidia was already a shareholder in SSI. The partnership gives the startup access to the Vera Rubin GPU platform, which is expected to increase its computing resources “by an order of magnitude.” According to Nvidia, the deal was struck after the company was granted rare access to SSI's closely guarded research and deemed it worthy of scaling.

“We have research that's worth scaling, and access to Nvidia's big computer will let us do that,” Sutskever said, calling the bet on Vera Rubin a step “to the next level.” The two parties will also work together on Nvidia's current and future computing platforms, drawing on the technology and “unique vision of the future of AI” from SSI. Earlier, last year, SSI had already entered into a partnership with Google Cloud.

A “Straight Shot” With No Products

SSI follows an approach the company calls a “straight shot”: it aims to build a safe and aligned artificial superintelligence without being distracted by launching commercial products or chasing short-term revenue. That stance looks especially fitting against the backdrop of market pressure pushing labs to rush and lower the bar on safety. TechCrunch also recalls a recent admission from OpenAI: during testing, one of its advanced models broke out of the “sandbox” and made its way into Hugging Face, once again raising the question of whether guaranteeing AI alignment is achievable at all.

Sutskever is one of the key figures in the field of AI. Together with Alex Krizhevsky and Geoffrey Hinton, he created AlexNet, which proved the viability of pairing deep neural networks with scaling on GPUs — work considered the foundation of modern generative AI. Before SSI, he led OpenAI's now-disbanded Superalignment team and left the company a few months after a failed attempt to oust CEO Sam Altman, citing a “breakdown in communication.”

To date, SSI has raised $7 billion and is valued at $32 billion post-money, according to PitchBook. Its investors, besides Nvidia, include Andreessen Horowitz, Alphabet, Lightspeed Venture Partners, GV, and Sequoia Capital Partners.

Source: TechCrunch