LLMs Are 'Pervasively Polyphonic': New Paper Rethinks What AI Understanding Means
burny_tech · x · 2026-10-15
A new arXiv paper by Matthieu Queloz and Pierre Beckmann argues LLMs are 'pervasively polyphonic': outputs emerge from coalitions of parallel mechanisms of uneven reliability, none indispensable. The inherited 'monophonic' concept of understanding — one mechanism underpinning all capacities — is ill-framed and leads to hazardous inference. The authors propose grounding trust in 'sound circuitry' that is reliably recruited and controls outputs, making understanding attributions tractable claims about internal organization.
More from AGI Musings
- "You distilled humanity": devs slam Anthropic's double standard on training data and model output — HankYeomans · 2026-10-15
- Blaise Aguera: Human intelligence is societal—and top AI works as societies of models — blaiseaguera · 2026-10-15
- Security researcher: doomers don't belong in security — HankYeomans · 2026-10-15
- New Paper: 500k Glassdoor Reports Show Interviews Are Two-Way Screening — Bad News for AI Interviews — danielrock · 2026-10-15
- Europe's edge may lie in applied AI, not frontier models — ingliguori · 2026-10-15
- Dev proposes annual week-long offline survival camps for kids after superintelligence — sytelus · 2026-10-15