开启时会模糊预览图，关闭后正常显示
RT @glangley: Oakland debated, and the facts won out. • Fact — after Flock tech was deployed in Oakland in 2024, violent crimes were solv…
GP @a16z — Building American Dynamism 🇺🇸 — Anthropologist — Formerly Founder/CEO @OpenDNS — Lokah Samastah Sukhino Bhavantu
if your enjoy what we share. you should definitely join our free newsletter. 👇 https://t.co/R4AnveK3V6
Discovering "boring" businesses at https://t.co/VrB2vWopEc. Sharing "boring" business opportunities at https://t.co/1qQOVZrUXW.
That they did this after NSA, which argued for post-trained sparsity being a dead end, also speaks to their ability to ignore sunk costs. I do wonder what happened to NSA though. It's clear that blockwise attention is suboptimal, but also that DSA is a stopgap.
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
Chatting #AI with @anilananth over ravva dosa/poori/idlee/kapi at Tivoli garden in B'lore.. 😎
AI researcher & teacher @SCAI_ASU. Former President of @RealAAAI; Chair of @AAAS Sec T. Here to tweach #AI. YouTube Ch: https://t.co/4beUPOmf6y Bsky: rao2z
DSA is such a zany idea actually. People only appreciate it for efficiency gains (if at all) but it's a big philosophical deviation from the industry pattern of optimizing layerwise linear/SWA+full softmax hybrids. Only a very confident lab can put something like this in prod.
That they did this after NSA, which argued for post-trained sparsity being a dead end, also speaks to their ability to ignore sunk costs. I do wonder what happened to NSA though. It's clear that blockwise attention is suboptimal, but also that DSA is a stopgap.
More coming in tomorrow's edition of https://t.co/BvTc8nQQW5: Get access to 100+ AI Agent, RAG, LLM, and MCP tutorials with open-source code - All for FREE.
Open-source ecosystem for high-leverage AI builders. GitHub's #1 AI Agents Repo with 75k+ stars. Join 200k+ active AI builders.