DeepSeek accused of 'pretending linear attention doesn't exist' in new architecture
teortaxesTex · x · 2026-09-10
A user asked why DeepSeek's impressive new model still doesn't use linear attention, prompting teortaxesTex to quip that DeepSeek 'pretending linear attention doesn't exist is glorious' — a jab at the lab's deliberate avoidance of that architecture路线.
More from Fun
- AI meme: a scaled causal encoder-decoder model is 'more historically unique than Navier-Stokes' — max_paperclips · 2026-09-10
- Ex-OpenAI policy VP Miles Brundage: rival AI leaders are trading compliments now, 'the vibes are a-shifting' — Miles_Brundage · 2026-09-10
- Someone paid $15,000 of their site's monthly profit to outbid Higgsfield — marclou · 2026-09-10
- "I think I downloaded the wrong Muse agent chat" — cto_junior · 2026-09-10
- Beff Jezos: 90% sure Anthropic will 'synthesize bioweapons' to spur regulatory capture — AIFlow_ML · 2026-09-10
- FIFA as a misaligned mesa-optimizer: AI alignment concepts applied to the US Open — gleech · 2026-09-10