Partnering with Ineffable Intelligence: A Superlearner for the Era of Experience
Summary
- Sequoia has partnered with David Silver & Ineffable Intelligence, a new AI research lab based in London, to pursue first contact with superintelligence.
- Ineffable is developing a "superlearner" system that discovers knowledge through its own experience without pre-training or human data imitation.
- This reinforcement learning approach aims to rediscover & transcend human inventions like language, science, mathematics, & technology from first principles.
- Unlike current AI trained on human internet data, Ineffable scales from a clean base to enable agents to develop non-human reasoning strategies.
- David Silver previously led defining deep reinforcement learning breakthroughs at DeepMind, including AlphaGo, AlphaZero, AlphaStar, & AlphaProof.
- Silver's work on self-play demonstrated that removing human pre-training allowed systems to achieve decisively superhuman performance with non-human mannerisms.
- The partnership represents a contrarian bet on scaling reinforcement learning despite the uncertainties of the timeline to superintelligence.
- Sequoia is co-leading Ineffable's first round of funding & supporting this ambitious scientific mission.
David Silver's new research lab is racing to build AI's next chapter with an RL 'superlearner.'