Sequoia backs David Silver’s Ineffable Intelligence, a London lab pursuing reinforcement-learning superintelligence without human training data

Sequoia says it is partnering with David Silver and his new London-based AI research lab, Ineffable Intelligence, to build a superlearner that learns entirely through experience. The lab aims to scale reinforcement learning from scratch, avoiding pre-training and imitation of human internet data.

The effort reflects a growing push beyond today’s data-hungry foundation models toward agents that can discover strategies and knowledge independently. If successful, experience-based AI could open paths to more original reasoning in science, math, and technology, but it remains a highly ambitious and unproven research direction.