Dean Ball: AI research taste may self-destruct as agents mass-produce 'slop'
Afinetheorem · x · 2026-10-05
Responding to the debate conflating "RSI" with "foom", Dean Ball argues automating AI R&D may produce slower, more jagged progress rather than discontinuous speedups.
His core point: research taste may be trickier than it looks. Even if next-gen AI finds genuinely high-taste research directions, they'll be exhausted fast — after which the ideas are no longer high taste, having been explored en masse. It mirrors the moving target of "slop": anything recognizably agent-produced becomes slop or commodified once promulgated at scale.
He cites examples: certain styles of math papers are becoming slop; certain web design styles already are. No media or intellectual output proliferated at such scale may remain high taste, even if we'd have considered it genius from a human a few years ago. The bar continually rises.
More from AGI Musings
- Economists Debate Whether AI Will Do Economic Research in Its Own Format — paulnovosad · 2026-10-05
- Professor's grim math: slowing AI costs 100,000 lives per week — DoctorJosh · 2026-10-05
- Countervailing Technologies: how distillation-style tools can diffuse AI power concentration — DavideCrapis · 2026-10-05
- NeuroAI Researcher: You Don't Need to Emulate the Whole Brain to Get Intelligence — aran_nayebi · 2026-10-05
- Consciousness debate sparks: 'neurons are just math' argument mocked — inductionheads · 2026-10-05
- Gary Marcus Mocks Lawmaker's Argument That LLM Randomness Rules Out Consciousness — GaryMarcus · 2026-10-05