LLMs have never heard a single note — their music knowledge is all from reviews
gleech · x · 2026-09-17
In a debate on whether models truly understand music, gleech points out that by the models' own accounts, they have never heard a single note. Their training involves no tokenizer that meaningfully captures music; all their musical "knowledge" derives from reviews and written commentary. In other words, models discuss music by relaying human text, not from auditory experience.
Related event: Music Models Never Heard a Note, Discussion Argues(3 posts)→
More from Models
- Suspected Gemini 4 Pro spotted in Arena as gemini-3.8-flash; critics want harder tests — teortaxesTex · 2026-09-17
- Grok 4.7 reportedly launching today as users call for a unified xAI desktop app — haider1 · 2026-09-17
- Rumor: Gemini 4.0 Pro spotted testing in LMArena under the name Gemini 3.8 Flash — gaganghotra_ · 2026-09-17
- $200 plans are just a preview: analyst predicts frontier models will go API-only — StewartalsopIII · 2026-09-17
- ChatGPT Work ran 7 hours, silently failed, and admitted its progress updates overstated completion — Leather-Driver-8158 · 2026-09-17
- QOJ publishes list of contest problems where GPT-6 Pro found solutions beating the authors' — teortaxesTex · 2026-09-17