Sarah Catanzaro (@sarahcat21) started following @TheGregYang on Sep 28, 2024

:( Guts

I'll vote for whichever candidate first dresses up as goku and fights this Majin Buu

RT @DongfuJiang: Excited to share GenAI-Arena and Wildvision-Arena are accepted to NeurIPS 2024 Dataset & Benchmark Track! VideoScore has a…

RT @TheGregYang: The future belongs to those who do evals

Damn. @allen_ai is crushing it. I have no affiliation or incentive to compliment them; I’m just genuinely floored.

You probably should read this but… TLDR: Your users will never discover the power of your tool if it’s not simple to use and understand

debating whether this halloween i should dress up as object being pointed by molmo

it's spreading 😊 https://t.co/CPEv7laL6t

RT @demishassabis: Feedback loop: train SOTA chip design model (AlphaChip) -> use it to design better AI chips -> use them to train better…

RT @GoogleDeepMind: Join host @FryRSquared as she speaks with @AncaDianaDragan, who leads safety research at Google DeepMind. 🌐 They explo…

+1; foundation models for weather forecasting are so damn cool. If you’re working on this and want to nerd out, DM me.

So what they’re all saying is that they had to raise from hip, under the radar investors who deeply empathized with developers and built a firm specifically to support technical founders? #vcbrags I guess

This is great advice. Figure out who the hiring manager is and send them a short but personalized note (by email; not LinkedIn - spend the time to find it; it’s not that hard). So few people (proportionally) do this that it stands out when they do.

Excited to share what we’ve been working on at CSM!

RT @CSM_ai: We have built CommonSim-1, a neural simulation engine controlled by images, actions and text. This is a new era in generative A…

RT @altryne: About to kick off the @NousResearch NousCon! I hear there going to be a few announcements! 🔥 Will try to keep yall updated…

Playground v3: Improving Text-to-Image Alignment with Deep-Fusion Large Language Models abs: https://t.co/3EZbWVWOfu New arch: image DiT + pretrained Llama-3-8B text transformer with image-text joint attention. Similar to, but simpler than MMDIT. Trained on synthetic https://t.co/xWQPL8hz2u

AI is not about automation but augmenting. Creativity and preserving creative control. So excited to see @runwayml develop custom models for @Lionsgate to help their storytellers continue to advance their oeuvre.

So many gen-Z asking what Oracle does and yet we wonder why data management for AI apps is hard…

Machines are becoming more complex but these are still systems built on data, code, and models. Great to see tool builders like the team from @OuterboundsHQ focused on managing these assets as their users and their use cases evolve: https://t.co/pq2hZIOrSK

Why do most new LLMs sound like peptide hormones? And more importantly, why do we call it post-training and not post-translational processing? The biopharma people got that one right…

>LeNet >AlexNet >AidanBench this field was built by people with the audacity to name things after themselves, love to see it

typical DeepMind-er: distinguished cognitive science professor at Cambridge typical OpenAI-er: IOI and IMO gold 2014 typical Anthropic-er: Ivy-league EA physics major on lethal dose of stimulants

Test time compute is great and all but so are products that run fast.