Ineffable Intelligence adds six cofounders from DeepMind, InstaDeep

Ineffable Intelligence's rapid assembly of a high-profile team is a direct challenge to Google DeepMind's dominance in AI research. Four of the six new hires are Silver's former colleagues, and three worked on AlphaStar, a breakthrough that showcased DeepMind's ability to build AI that beats humans at complex games. This influx of talent is part of a broader trend of researchers leaving DeepMind and other older labs to join 'neolabs.'

Ineffable's stated mission is to achieve superintelligence through reinforcement learning, the technique Silver is known for. The hiring of Junhyuk Oh, who was the reinforcement learning lead on AlphaStar and has explored automating AI research, suggests a focus on 'recursive self-improvement.' Similarly, Wojciech Czarnecki's background in multi-agent systems will shape their science team.

The stakes for Ineffable are high, now that it has a full founding team and a $5 billion valuation. The success of this endeavor hinges on whether this specific group of experts can use reinforcement learning to achieve more than incremental progress. The company's ability to live up to the 'superintelligence' promise, backed by a historic seed round, will likely be judged by its measurable research milestones in the coming years.