
Everything related to machine learning and data science in general. Follow @fastml for notifications about new posts and @fastml_viz for data visualizations.
Joined June 2013
- Tweets10.2K
- Following162
- Followers6.2K
- Likes38K
@fastml_extra retweeted
It cost Anthropic $1.63m to destroy ~$1.5T of Nvidia’s market cap and >$5T of the entire AI technology sector.
‘Anthropic paid $1.63 million to two lobbying firms tied to Secretary of State Marco Rubio and former White House chief of staff Susie Wiles.’
@fastml_extra retweeted
"We are all trying to find the guy who did this"
We will stop measles outbreaks, protect American families, and work to regain U.S. measles elimination status. I continue to recommend MMR vaccination as the most effective way to prevent measles.
Worldwide COVID-era disruptions interrupted routine childhood vaccination, leaving gaps in protection here and abroad. The multi-year global outbreak has caused many nations to lose their elimination status including, Armenia, Austria, Azerbaijan, Bolivia, Canada, Mexico, Spain, United Kingdom, and Uzbekistan.
Under Dr. Schwartz’s leadership, CDC has provided approximately $100 million in additional funding since August to strengthen prevention and response to measles and other vaccine-preventable diseases.
CDC has worked alongside state teams in Texas, South Carolina, Virginia, Kansas, New Mexico, and Wisconsin to help bring their outbreaks rapidly and successfully under control. CDC is also supporting Pennsylvania and New York with epidemiologists and other experts.
We are working to close vaccination gaps and rebuild public trust. We will answer families’ questions honestly about vaccine benefits and risks and work with trusted health care providers and community leaders to prevent measles, limit its spread, and support timely treatment.
We will support Americans who become ill, regardless of vaccination status, and for the first time equip health care providers with guidance to diagnose and treat measles and its complications.
@fastml_extra retweeted
It's sadly appropriate that the founding father of the GPU seems to have been banging mistresses in parallel, rather than sequentially.
@fastml_extra retweeted
Researchers have used reinforcement learning to teach a 5-tonne John Deere skid-steer loader to perform complex acrobatic manoeuvres autonomously. Trained in simulation and transferred to a real machine, the system can balance on two wheels, navigate rough terrain, spin, and control its bucket before landing safely.
The research demonstrates the potential of AI-driven automation for heavy construction and industrial machinery.
Video Credit: @leggedrobotics / @ETH
Paper: wevlv.co/4kltYys
#engineering #technology
@fastml_extra retweeted
BREAKING: Jensen Huang is in the datacenters ripping the Nvidia cards out - causing worldwide outages - saying “they’re mine anyway”. He removed his leather jacket, revealing he’s all greased up, and they can’t get a hold of him because of the grease
@fastml_extra retweeted
I think it was this one that did it:
The California Department of Motor Vehicles has ordered Waymo to stop operating its autonomous vehicles on California freeways, citing concerns that the vehicles pose an imminent public safety hazard.
abc7news.com/story/californi…
@fastml_extra retweeted
"Jensen was in a relationship with a CCP spy!"
No evidence. Walked back.
"Jensen was illegally selling chips to China"
Nope.
"I don't like that Jensen is lawfully spending so much capital facilitating the open source ecosystem"
Oh, okay. Finally found the real grievance.
@fastml_extra retweeted
So they didn’t know what was wrong but they definitely know $50B would fix it.
@fastml_extra retweeted
I found that Windows starts to collect your coding agent usage information. Windows 11 build 26200.9974 checks codex.exe, claude.exe, copilot.exe, cursor.exe, antigravity.exe, github.exe and agy.exe through MissionControlAggregator, then reports agent bitsets in ETW.
@fastml_extra retweeted
It is sheer madness that even the story about her being a spy turned out to be a lie, and all she did was handle routing for China through Taiwan. Essentially, this is the most blatant public character assassination of this century, as they took down a man for their own interests without any legitimate reason. – at Chiyoda-ku, Tokyo
@fastml_extra retweeted
NY Times: U.S. intelligence indicates that Jensen Huang had an affair with a spy.
U.S. intelligence: She is not a spy. We do not know what you are talking about.
NY Times: Well, what is a spy really anyway?
@fastml_extra retweeted
Wow that escalated quickly. Was gone for a few weeks and now pretty much everyone launched their own decision model copycats:
- Sep 23: Together AI, Tev1-4B
- Sep 24: Alibaba Cloud, decision-model
- Sep 29: Liquid AI, d1
- Sep 30: Databricks, ai_decide
- Oct 1: Amazon AWS, Strands Decider 2B
- Oct 1: Cloudflare, Clef and Clef-flash (Clef-omni on Oct 9)
- Oct 1: Perplexity, Decider (v1.3 on Nov 12)
- Oct 6: OpenAI, Decisions API (GPT-6 Luna)
- Oct 7: H2O.ai, H2O-Lightning-4B
- Oct 7: Snowflake, Snowflake Decision
- Oct 9: Microsoft, Microsoft-Decision-1
- Oct 15: IBM, Granite Switch 4.1 with Decision adapter
- Oct 21: Baseten, Mammoth-1 and Elephant-1
- Nov 6: Salesforce, Koa Decisions
- Nov 11: Google DeepMind, Gemini Decision 1
According to a Vercel report, 26% of requests are now routed through decision models 📈
@fastml_extra retweeted
Every time I post this my mentions are full of people saying "nooo waaaay".
Yes way.
faa.caseyhandmer.com/
I hope this amuses you as you wait in airports overnight.
@fastml_extra retweeted
Labs disagree on vision encoders for MLLMs! ⚔️
Gemma 4 12B scrapped the encoder.
Kimi K3 dropped encoder pretraining.
Inkling 2 Small revived the encoder Inkling killed.
We trained 200+ models in a controlled study to settle it:
✅ Keep vision parameters, skip the pretraining. And they don't have to be an encoder! 🤔
Paper: arxiv.org/abs/2611.17334 🧵
@fastml_extra retweeted
This is exactly what I suggested last Thursday. Nvidia was planning to optimise a new internal open-weight AI based on Chinese advances this year. Anthropic kneecapped Huang to forestall the release.
“In June this year, Nvidia announced Nemotron 3 Ultra, an open-weights model with 550 billion parameters. In the interview below from four months ago, Jensen even said outright that Nemotron could already go toe-to-toe with Claude Opus, with capabilities approaching those of frontier large models, and at a much lower cost.”
现在基本可以确定,Anthropic就是这次针对黄仁勋抹黑行动的幕后黑手。向纽约时报爆料老黄和Alice私事的,正是Anthropic联合创始人Jack Clark。而他手里持有的Anthropic股票比Dario还多200万股。除了我之前分析过的双方在对华芯片出口问题上的分歧之外,还有一个几乎没人讨论过的关键问题。
英伟达其实早就准备亲自下场,跟Anthropic和OpenAI抢饭碗了。今年6月,英伟达宣布了Nemotron 3 Ultra,一个拥有5500亿参数的开放权重模型。老黄在四个月前的这段采访里甚至亲口说,Nemotron已经可以跟Claude Opus相提并论,具备接近前沿大模型的能力,而且成本还便宜得多。
你站在Anthropic的角度想想,如果这时候还坐以待毙,那不就是等死吗?一方面,你整个公司的命脉都掌握在英伟达手里,因为训练和运行模型都离不开它的GPU。另一方面,既然英伟达自己掌握着AI模型最核心的瓶颈——算力,那它凭什么不自己训练大模型,直接跟你竞争?AI产业就像一个五层蛋糕,而英伟达完全有潜力把这五层全部吃下来。






































