We Finally Know What Ilya Is Working On... | AI Copium

The Gist

Nvidia invested five billion dollars into Ilya Sutskever's safe artificial intelligence startup Safe Superintelligence after receiving a rare exclusive look at their research focused on human brain functions. This massive funding round and strategic partnership provide the startup with ten times more compute power, confirming that their secretive methodology centers on overcoming AI generalization limits.

Quick Overview

Nvidia invested five billion dollars into Safe Superintelligence, a secretive artificial intelligence lab founded by former OpenAI chief scientist Ilya Sutskever, Daniel Gross, and Daniel Levy. After previewing the lab's closely guarded research regarding overlooked aspects of human brain functions, Nvidia committed massive capital and hardware resources. This partnership provides Safe Superintelligence with ten times more compute power through Nvidia's Vera Rubin platform, bridging the gap between massive scaling and fundamental machine learning breakthroughs.

Key Points: Nvidia invested five billion dollars into Safe Superintelligence after reviewing their secret AI research. Safe Superintelligence was founded in June 2024 by Ilya Sutskever, Daniel Gross, and Daniel Levy with a valuation reaching thirty billion dollars. The startup operates without products, chatbots, or public model releases, focusing entirely on safe superintelligence. The partnership grants Safe Superintelligence ten times more compute power using Nvidia's next-generation Vera Rubin platform. Ilya Sutskever stated that their research focuses on overlooked aspects of how human brain functions operate. The startup previously relied primarily on tensor-processing units made by Alphabet's Google. Ilya Sutskever posted on Twitter that it is time to scale Safe Superintelligence following the Nvidia partnership.

Context: Ilya Sutskever left his role as chief scientist at OpenAI in 2024 to found Safe Superintelligence alongside Daniel Gross and Daniel Levy. Unlike competitors building commercial chatbots, the lab remained entirely secretive with no released models or benchmarks while raising billions in early funding.