What they're not telling you: # google-deepens-anthropic-bet-with-up-to-40-billion-investment.html" title="Google Deepens Anthropic Bet With Up To $40 Billion Investment" style="color:#1a1a1a;text-decoration:underline;text-decoration-style:dotted;font-weight:500;">Google DeepMind Veteran Raises $1.1 Billion For AI That Doesn't Train On Human Data A former DeepMind architect is betting $1.1 billion that the AI industry's dominant approach—training systems on human-generated data—is a dead end. David Silver, the scientist who engineered AlphaGo's 2016 victory over world Go champion Lee Sedol, launched Ineffable Intelligence in January with a $5.1 billion valuation and a provocative thesis: future superintelligence won't emerge from large language models scraping human text, but from systems that teach themselves through trial and error. Silver's bet represents a direct challenge to the OpenAI and Google model that currently dominates venture capital and public discourse, yet mainstream tech coverage has largely treated it as a minor sidebar to the LLM boom rather than a fundamental disagreement about AI's trajectory.
# THE TAKE This "human data-free" narrative is marketing theater masquerading as technical innovation. David Silver's raise exploits a genuine problem—training data toxicity, copyright liability, GDPR exposure—then sells the solution as philosophical purity. The math doesn't work. Synthetic data generation requires *something* trained on human knowledge. You can't bootstrap a language model from thermodynamic first principles. Silver's approach likely means: human data laundering through intermediate models, synthetic augmentation of existing datasets, or narrowly-scoped systems where the limitation reads as feature, not constraint. The billion dollars flows because enterprise wants compliance theater. "No human training data" checks regulatory boxes while avoiding the harder conversation: what intellectual property frameworks actually govern AI commodification? It's technically defensible but ideologically dishonest. Silver understands this. That's why the framing sells.
What the Documents Show
Silver's core argument cuts against the narrative that bigger models trained on more human data equal smarter AI. "Human data is like a kind of fossil fuel that has provided an amazing shortcut," he told Wired. "You can think of systems that learn for themselves as a renewable fuel—something that can just learn and learn and learn forever, without limit." This distinction matters because it suggests the current generation of AI—ChatGPT, Claude, Gemini—represents a technological cul-de-sac, not progress toward genuine intelligence. Instead of building understanding, these systems pattern-match against billions of human-generated tokens, fundamentally constrained by what humans have already done and documented. Silver's track record lends credibility to the contrarian position.
Follow the Money
AlphaGo didn't simply memorize Go strategies from human masters; it combined human-provided knowledge with reinforcement learning and self-play, eventually discovering moves that surprised even top professional players. This hybrid approach demonstrated that AI could exceed human achievement in narrow domains by learning independently rather than merely reproducing human expertise. Silver has spent his career advancing this insight, and his $1.1 billion war chest suggests serious institutional backers believe the approach warrants the investment despite the current LLM gold rush. What makes this significant is what it reveals about AI development's hidden disagreements. While OpenAI, Google, and Anthropic pursue scaling laws—the assumption that bigger models trained on more data equal better outcomes—Silver is articulating a different vision of superintelligence. "I think of our mission as making first contact with superintelligence," he said.
What Else We Know
"By superintelligence, I really mean something incredible. It should discover new forms of science or technology or government or economics for itself." This isn't science fiction; it's a concrete technical roadmap that diverges sharply from the prevailing consensus. For ordinary people, the implications are profound. If Silver's approach succeeds, it means the current fears about AI are misdirected. Concerns about LLMs reproducing human bias, polluting information ecosystems, and concentrating power in companies controlling training data would become footnotes. Instead, we'd face an entirely different problem: autonomous systems that generate knowledge entirely outside human oversight or comprehension.
Primary Sources
- Source: ZeroHedge
- Category: Tech & Privacy
- Cross-reference independently — don't take our word for it.
Disclosure: NewsAnarchist aggregates from public records, API feeds (Federal Register, CourtListener, MuckRock, Hacker News), and independent media. AI-assisted synthesis. Always verify primary sources linked above.