According to a leak, SSI, founded by former OpenAI chief scientist Ilya Sutskever, is developing a small reasoning engine based on test-time training (TTT), which could be released as early as August. The news has attracted widespread attention, as Ilya is considered a pioneer in deep learning, and his new move could significantly impact the AI industry.
SynthePulse Insight · AI deep readingMembers
Version 1 · 1 source
SSI, founded by former OpenAI chief scientist Ilya Sutskever, is reportedly developing a small inference engine based on test-time training (TTT), which may be released in August. This news has attracted widespread attention, but its authenticity has not been confirmed. This article reviews public clues and analyzes its technical route and potential impact.
On August 13, X user 'Three Strawberries' leaked that SSI is exploring a small inference engine based on TTT (Test-Time Training). The model would learn 'how to learn' from specially curated data and update some weights during problem-solving, allowing it to compete with much larger models even at a small scale.
The leak claims the current version is ready and may appear as early as August, initially open to a few users; the next generation will be 10 times larger. However, subsequent reports suggest it may not be released this month, casting doubt on authenticity.
This report is based on coverage by QbitAI, with core information from an anonymous leak by X user 'Three Strawberries,' whose authenticity has not been confirmed by SSI or Ilya. The Nvidia partnership and investment amount come from public announcements and Reuters reports, but the investment amount is a leak. Ilya's podcast views are from a public interview. The paper information is mentioned in the leak. All inferences are based on these sources.
Primary report
Primary source