DeepMind’s David Silver just raised $1.1B to build an AI that learns without human data
What Happened
Ineffable Intelligence, a British AI lab founded a mere few months ago by former DeepMind researcher David Silver, has raised $1.1 billion in funding at a valuation of $5.1 billion.
Fordel's Take
David Silver, the DeepMind researcher behind AlphaGo and AlphaZero, raised $1.1B at a $5.1B valuation for Ineffable Intelligence. The pitch: models that learn from self-generated experience instead of human data. The lab is months old and has shipped nothing.
If this works, the RLHF and synthetic-data pipelines most teams are bolting onto Llama and Claude fine-tunes become obsolete overhead. Self-play worked in Go because the reward was unambiguous; language and agents have no such scoreboard. Most founders chasing 'agentic RL' are cargo-culting AlphaZero without a verifiable reward function, and Silver knows it better than they do.
Researchers tracking post-training should watch closely. Product teams shipping this quarter can ignore — nothing here touches your roadmap before 2027.
What To Do
Keep your eval harness deterministic and reward-checkable now, because that is the only moat self-play methods will respect.
Builder's Brief
What Skeptics Say
$5.1B pre-product for a thesis that has failed outside narrow game domains for a decade. Self-play needs a ground-truth reward; open-ended reasoning has none, which is exactly why DeepMind itself never shipped this.
Cited By
React
Get the weekly AI digest
The stories that matter, with a builder's perspective. Every Thursday.
