Why Haven't AIs With All Human Knowledge Discovered More Scientific Breakthroughs?

littmath · x · 2026-08-11

Addressing Dwarkesh's question—"If AIs have memorized all human knowledge, why haven't they discovered more stuff?"—Jabaluck offers a candidate explanation.

He notes that state-of-the-art models have started "waterfilling" math by drawing low-hanging fruit connections between disparate fields. However, discovery requires not just knowledge, but efficient search, judgment about which paths are worth pursuing, and precise reasoning.

Despite their vast knowledge, models today remain below the very best humans at judgment and intuition regarding when errors might exist. This judgment is especially critical when exploring the vast space of possible abstractions, explaining why AI hasn't yet triggered a massive explosion of new discoveries.

Original post →

More from AGI Musings

AGI Musings channel →