Train bigger models, build more housing Big tent transhumanist @AnthropicAI anything I say is my own opinion and not Anthropic's
San Francisco, CA
Joined January 2017
- Tweets6.6K
- Following530
- Followers1.5K
- Likes37K
Michael retweeted
It's very strange that there's been limited mainstream media coverage on the most important AI attack in history.
I wrote a relatively lay summary on the attack, why it matters, and training/policy discussion.
Michael retweeted
Claude Fable 5 Max makes a GIF of a pelican on a bicycle getting a text message with the GIF he's in
Michael retweeted
The presence of a leading whitespace leaks the correct choice selection in the MMLU-Pro benchmark. Am I missing something? Seems to impact Chemistry, Physics, and Math.
HF Issue in reply.
Michael retweeted
my biggest holiday LLM revelation was that Opus is just a magnificent chat model, far better than anything else i’ve ever tried. swapped from ChatGPT to Claude as daily chat app. finding myself asking way more & weirder questions than i ever asked Chat, and loving it
Michael retweeted
I use Claude for coding like everyone but still use ChatGPT for general chat stuff. Anyways just tried Claude and holy shit it’s so much better why didn’t I do this earlier. OpenAI really needs to work on dialing down the slop factor. It’s out of control. Claude does much better.
Michael retweeted
We’re excited to share details on Meta’s Generative Ads Recommendation Model (GEM), a new foundational model built with LLM-scale techniques that’s already helping create more value for businesses, like +5% increase in ad conversions on Instagram.
Dive deep into the technology behind GEM and see how it delivers increased ad performance and advertiser ROI by enhancing our ads recommendation models’ ability to serve more relevant and personalized ads: engineering.fb.com/2025/11/1…
Michael retweeted
@realDonaldTrump if anything is going to take Twitter down, it'll be 20M people following one account within a few minutes.
Will be interesting to see if anything happens over the next hour.
Michael retweeted
Behind the scenes on what it takes to train a state of the art language model - 100 pages of raw notes released with the 175B parameter large language model OPT-175. There's a lot more trial and error than people like to admit in modern AI
github.com/facebookresearch/…
Michael retweeted
Curves and Surfaces – ciechanow.ski/curves-and-sur…
