WaveNet turns 10: audio-as-language-modeling is now the standard, dilated convs still everywhere
heiga_zen · x · 2026-09-09
A researcher reshared DeepMind's original WaveNet announcement to mark the 10th anniversary of the landmark paper.
- Audio generation framed as language modeling has become the standard paradigm, and the paper's dilated convolutions still appear throughout modern codecs and vocoders.
- The author recalls being struck a decade ago that this would transform speech synthesis forever, while noting the paper was written very 'frugally' — details and ablations felt scarce because, per one author, the team rushed the writeup to share it as widely and quickly as possible.
More from Research
- Goodfire commits $10M to launch a 'Human Genome Project for AI' interpretability effort — mathildepapillo · 2026-10-27
- Researchers call for a Human Genome Project for AI to fully understand neural networks — mathildepapillo · 2026-10-27
- Jon Barron: every softmax and cross-entropy implementation has two subtle numerical bugs — jon_barron · 2026-10-27
- New theory paper explains what Anthropic's Jacobian lens actually detects in LLMs — wgilpin0 · 2026-10-27
- Interpretability researcher joining Goodfire AI post-PhD, calls for speeding up the field — mathildepapillo · 2026-10-27
- EdgeRock-4B holds KV cache at 72 MiB across 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27