Keep on to blur preview images; turn off to show them clearly
Fascinating discussion with gork on the nature of “manufacturing” and how American manufacturers are 7x more productive. Basically: because EVERYTHING Americans are involved in is manufacturing. iPhones made in Shenzhen? “Upstream value”. Licensing Shanghai drugs? You guessed it!
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
i think this is beautiful - doing ViT from raw pixels means you need to jointly train everything - this poor model must independently solve MNIST, and THEN/ALSO learn to be a perfect calculator in its weights. then keep going.... only constrained by the data you give it. it's why @percyliang's concept of "foundation models" in 2021 was so disruptive/sacrilegious in the Google vs OpenAI sprint to the GPT: instead of 1000 different small models all specialized in their tasks, concentrate all that budget/data/resources in one supermodel that has the capacity to model 1000 tasks; along the way you get 1) transfer learning, 2) capabilities you never explicitly trained for, 3) emergent abilities that only unlock at a given param/depth/data exposure rate.
achieve ambition with intentionality, intensity, & integrity - @dxtipshq - @sveltesociety - @aidotengineer - @latentspacepod - @cognition + @smol_ai
An open-source vibe coding platform that helps you build your own vibe-coding platform, built entirely on Cloudflare stack
https://t.co/Hg7K1z0bcC
发现boss上和v2ex上不定期有黑灰产招爬虫，这个Jd描述得特别定向。 譬如想爬拼多多、想爬微信公众号、想爬淘宝商品、淘宝评论、想爬抖音视频... 这些黑产统一特点是，给的工资不高（1万一个月），福利也隐晦不谈，要求倍儿高。 如真有傻的程序员信了这个邪，操着卖白粉的心拿着卖白菜的钱，回头还可能遇到跑路（欠钱不给）的情况。 这些发招聘信息的黑灰产真当人家公司安全团队是废材，回头锅会全甩到程序员头上，这种程序员工作真心面向监狱编程-_-。
从投资领域转到创业：找工作、找面试题、改简历、模拟面试. 创业（冷启动）｜AI , AIGC ｜ 安全技术｜RAG | 时空智能 | 认知心理学｜智能体 ｜ 生命科学 ｜ 强化学习 I built open source software at https://t.co/b69DXZhcyR
The new paid plan is now available on https://t.co/nCmwcvlJIY! Now, you can access professional product review blogs on the DR 74 site through this plan.
If you previously paid for the Turbo0 Pro plan, you can DM me to get a $69 discount code.
It is more than what meets the eye :KIMI-2-Thinking QAT. It has to also do with supporting more/Chinese AI Chips. Must read brilliant short blog below on why Kimi (Moonshot AI) chose QAT (quantisation aware training). Here is my read. TL:DR: Not just that it reduces latency for inference and memory requirement in memory bound scenarios (which MoE with Kimi-2's sparsity scale is) and speeds up RL training by 10-20%; it also enables alternate hardware ecosystems like Cambricon and Ascend from Huawei due to INT4 format. Quotes from the blog: ================= 1) Why INT4, not MXFP4? Kimi chose INT4 over "fancier" MXFP4/NVFP4 to better support non-Blackwell GPUs, with strong existing kernel support (e.g., Marlin). 2) Kimi-2-Thinking weights are 4 bit and activations are 16 bit (denoted as W4A16) They sate further W4A8 and even W4A4 are on the horizon. As new chips roll out with FP4-native operators, Kimi's quantization path will continue evolving.... INT4 support in made in China chips: ================= Cambricon GPUs explicitly support INT4 quantization, including for AI inference workloads, as seen across several models such as the MLU270, MLU370-X8, and newer chips, as well as in recent open-source releases with INT4 integration for large models like GLM-4.6. Huawei Ascend NPUs also support INT4 quantization for inference, as confirmed by documentation and kernel releases related to GEMM and quantized model deployments.
AI @amazon. All views personal!