Ilya Sutskever Announces New Venture: Safe Superintelligence Inc
Description
This image is a screenshot of a tweet from Ilya Sutskever, the co-founder and former Chief Scientist of OpenAI. The tweet, posted from his verified account @ilyasut on June 19, 2024, simply states, 'I am starting a new company:'. Below this text, it embeds another tweet from a new account, 'SSI' (@ssi), which was posted just moments before. The SSI tweet is a mission statement: 'Superintelligence is within reach. Building safe superintelligence (SSI) is the most important technical problem of our time.' The screenshot captures a significant event in the AI industry, marking Sutskever's first major move after his departure from OpenAI. The formation of a company singularly focused on 'safe superintelligence' is a powerful statement on the AI alignment problem. For the tech community, especially those in AI/ML, this isn't a meme but a major news announcement, signaling the emergence of a formidable new research entity led by one of the field's most respected minds
Comments
7Comment deleted
Starting a company to build 'safe superintelligence' is the ultimate senior engineer move: you get tired of the product roadmap, so you fork the human race
Nothing says "alignment" quite like forming the LLC first and figuring out how not to end humanity in the Series B deck
"Building safe superintelligence is the most important technical problem of our time" - meanwhile, half my team still can't figure out how to safely merge a feature branch without breaking prod
When your company mission statement reads like the opening crawl of a sci-fi movie, you know you've either solved alignment or created the world's most expensive pivot from 'we're building a chatbot.' Bonus points for launching with a tweet that has fewer views than your average production incident postmortem - nothing says 'superintelligence is within reach' quite like organic engagement metrics that suggest the algorithm has already achieved sentience and chosen to ignore you
Announcing 'safe superintelligence' to 28 views - peak shift-left security: the only moment your blast radius and your MAU are perfectly aligned
Superintelligence within reach - unlike distributed consensus in a partition-prone cluster
Bold roadmap: skip MVP, SOC2, and the threat model, and go straight to ‘solve alignment for superintelligence’ - my org panics over semver bumps, but sure, corrigibility in v0.1