DeepMind Pioneer David Silver Departs, Betting LLMs Won't Reach Superintelligence
# DeepMind Pioneer David Silver Departs, Betting LLMs Won't Reach Superintelligence The architect of AlphaGo is going his own way — and he thinks current AI is heading down the wrong path. According to The Decoder, David Silver, one of DeepMind's first employees and the lead researcher behind A...
DeepMind Pioneer David Silver Departs, Betting LLMs Won't Reach Superintelligence
The architect of AlphaGo is going his own way — and he thinks current AI is heading down the wrong path.
According to The Decoder, David Silver, one of DeepMind's first employees and the lead researcher behind AlphaGo, has left to found a new company called Ineffable Intelligence. The startup was registered in November 2025, and Silver is now seeking investors and AI researchers.
According to Fortune, a Google DeepMind spokesperson confirmed Silver's departure: "Dave's contributions have been invaluable, and we're grateful for the impact he's had on our work at Google DeepMind."
Silver's bet: large language models alone cannot reach superintelligence. His reasoning, according to The Decoder, is that LLMs are fundamentally limited because they're built on human knowledge. Instead, he's doubling down on reinforcement learning — the same approach behind AlphaGo, where AI learns through trial and error rather than imitating human data.
In April 2025, Silver co-authored a paper with renowned AI researcher Richard Sutton calling for what they termed the "Era of Experience" — a fundamental shift away from training on human knowledge toward systems that learn from their own experience. Silver has described Ineffable Intelligence's goal as building an "endlessly learning superintelligence that self-discovers the foundations of all knowledge."
Silver's track record is notable. According to Fortune, he was instrumental in developing AlphaGo (which beat the world's best Go player in 2016), AlphaStar (which beat top StarCraft II players), AlphaZero (which mastered chess, shogi, and Go from scratch), and MuZero (which mastered multiple games without knowing the rules). More recently, he worked on AlphaProof, which answered International Mathematical Olympiad questions.
Our read: Silver's departure signals that even some of the researchers who built today's most impressive AI systems believe the Transformer architecture has hard limits. It's a notable contrarian bet from someone who has actually shipped AGI-adjacent results.
