
ai scaling + simplifying. founder @LatentSpaceAI spoilers from my favorite story: “Overhead, without any fuss, the stars were going out”
San Francisco, CA
Joined February 2014
- Tweets82
- Following1.5K
- Followers417
- Likes122
Ian Thompson retweeted
THRILLED to share that ChatGPT, our new model optimized for dialog, is now public, free, and accessible to everyone. chat.openai.com/
Try talking with ChatGPT, our new AI system which is optimized for dialogue. Your feedback will help us improve it. openai.com/blog/chatgpt/
Ian Thompson retweeted
Avalon is the world's fastest 3D simulator for RL agents. All baselines train on 1 GPU in ~1 day.
We want academic researchers to be able to study aspects of intelligence missing from today’s models, even w/o access to large-scale compute.
Get started: generallyintelligent.com/ava…
Ian Thompson retweeted
In 2020, @michael_nielsen & I began a 2-month project to write: "how would we fund science?"
2 years & 40,000 words later, it's become: "how can the culture & institutions of science actually change, and ultimately become self-improving?"
Our answer: scienceplusplus.org/metascie…
Ian Thompson retweeted
Bluesky is building a social protocol. We released “ADX” (the X stood for Experiment) in May. Now that the design is starting to solidify, we’re renaming it to the “Authenticated Transport Protocol” — the “AT Protocol.”
blueskyweb.xyz/blog/10-18-20…
Ian Thompson retweeted
Could a DAO build the next great city? Honored to give the first TED Talk on DAOs!
ted.com/talks/scott_fitsimon…
Ian Thompson retweeted
It takes a lot of human ratings to align language models with human preferences. We found a way to learn from language feedback (instead of ratings), since language conveys more info about human preferences. Our algo learns w just 100 samples of feedback. Check out our new paper!
Can we train LMs with *language* feedback?
We found an algo for just that. We finetune GPT3 to ~human-level summarization w/ only 100 samples of feedback
w/ @jaa_campos @junshernchan @_angie_chen @kchonyc @EthanJPerez
Paper: https://teapaw.org/t.co/BBeQbFMtVi
Talk: https://teapaw.org/t.co/48uwmwakOH
Ian Thompson retweeted
Excited to share a peek at what I've been working on recently: we're building networks of little sensors and using ML to help wind farms produce more energy!
Ian Thompson retweeted
It’s technically possible for our online identities and relationships to belong to us the way an email address, phone contact book, or password keychain does. So I think it will happen, one way or another. It’s just a matter of time.
Ian Thompson retweeted
Our first AI alignment paper, focused on simple baselines and investigations: A General Language Assistant as a Laboratory for Alignment arxiv.org/abs/2112.00861
Ian Thompson retweeted
Excited to share a bit about what @vmcheung and I have been working on this year!
At Gantry, we build infrastructure to help you go from making ML projects to shipping ML-powered products.
I wrote about some of the main challenges we see here: gantry.io/blog/toward-contin…
TL/DR 👇
Ian Thompson retweeted
Excited to share what I've been working on for the last few months! If you're interested in scaling laws and safety (or scaling laws *𝗳𝗼𝗿* safety) then check out our careers page: anthropic.com/#careers
Ian Thompson retweeted
Language models are amazing few-shot learners with the right prompt, but how do we choose the right prompt? It turns out that people use large held-out sets(!). How do models like GPT3 do in a true few-shot setting?
Much worse: arxiv.org/abs/2105.11447
w/ @douwekiela @kchonyc
1/N
Ian Thompson retweeted
excited to share what I've been working on the past few months - Scaling Laws for Language Transfer Learning youtube.com/watch?v=lpe5Gwuq…
Ian Thompson retweeted
There's a lot of work on probing models, but models are reflections of the training data. Can we probe datasets for what capabilities they require? @kchonyc @douwekiela & I introduce Rissanen Data Analysis to do just that: arxiv.org/abs/2103.03872
Code: github.com/ethanjperez/rda
1/N
Ian Thompson retweeted
I and @joshalbrecht wanted to hear more about other ML researchers' hunches, unpopular opinions, & research processes, so we decided to start a podcast! The first episode is out today, with the fascinating @kelvin_guu :)
Let me know if you have any guest requests!
We're thrilled to release our first-ever episode of “Generally Intelligent,” a podcast made specifically for ML researchers by us folks at Untitled AI!
Our first guest is the amazing @kelvin_guu from Google AI, co-author of REALM.
Episode + transcript: untitled-ai.com/podcast-epis…
Ian Thompson retweeted
Showing that an AI system causes harm isn't the same as showing that we shouldn't build it or that we shouldn't deploy it. Some thoughts on AI ethics.
askell.io/posts/2020/12/bad-…
Ian Thompson retweeted
What's next after open-source and open-access? We believe it's *open collaboration*. The field is fairer, more inviting, diverse and inclusive only when we make mentorship and collaborators accessible. But how? We are discussing it on Tue at #NeurIPS2020: mlcollective.org/neurips-202…
Ian Thompson retweeted
This underscores the importance of measuring likelihood in the right space, which I briefly discussed in a recent blog post (Section 6): benanne.github.io/2020/09/01…
That way, we can ensure most of the model capacity is used for things that actually matter. (4/5)
Ian Thompson retweeted
Turns out batch norm might cause contrastive learning! 🤯 Fascinating blog post demystifying @DeepMind's new BYOL approach here: untitled-ai.github.io/unders…























