Google's Knowledge Graph Could Decide Who Wins AGI

Demis Hassabis laid out his vision for AI's future in a recent YouTube interview. But the real conversation happened after the video. Hacker News commenters focused on a specific technical shift: merging large language models with knowledge graphs. Combine the two and you get a pipeline from research to production that runs on compute instead of human effort.

DeepMind's AlphaGeometry proves this works. The system solves olympiad-level geometry by pairing a neural language model with a symbolic deduction engine. Hassabis calls this "System 2" thinking, AI that reasons through problems rather than pattern-matching. It works.

Google owns the largest structured knowledge base on the internet. After folding Google Brain into DeepMind, they're building models like Gemini that blend multimodal understanding with actual reasoning. Hacker News commenters said they'd prefer Hassabis to win the AGI race over Sam Altman, citing ethics concerns elsewhere. But preference isn't protection. When one company holds the data and the compute, independent researchers lose regardless of who's in charge.