Ted Moskovitz retweetledi
Ted Moskovitz
306 posts
Ted Moskovitz
@ted_moskovitz
science of scaling at @AnthropicAI
London, UK Katılım Ekim 2019
244 Takip Edilen1.1K Takipçiler
Ted Moskovitz retweetledi
Ted Moskovitz retweetledi
Ted Moskovitz retweetledi
I strongly recommend signing up - Anthropic is an amazing place to work at, I love the culture, the impact is massive and the pace is incredible!
I’ll be at the event, happy to answer your questions there 🤗
Anthropic@AnthropicAI
We're hosting a social hour in London in early August for quant traders and developers. If you'd like to join us, meet researchers from our London office, and learn about the technical problems we're working on, please sign up at the following form: gem.com/form?formID=3d…
English
Ted Moskovitz retweetledi
We're hosting a social hour in London in early August for quant traders and developers.
If you'd like to join us, meet researchers from our London office, and learn about the technical problems we're working on, please sign up at the following form: gem.com/form?formID=3d…
English
This is a great opportunity!
Jack Lindsey@Jack_W_Lindsey
We're launching an "AI psychiatry" team as part of interpretability efforts at Anthropic! We'll be researching phenomena like model personas, motivations, and situational awareness, and how they lead to spooky/unhinged behaviors. We're hiring - join us! job-boards.greenhouse.io/anthropic/jobs…
English
Ted Moskovitz retweetledi
📣 Anthropic Zurich is hiring again 🇨🇭
The team has been shaping up fantastically over the last months, and I have re-opened applications for pre-training.
We welcome applications from anywhere along the "scientist/engineer spectrum". If building the future of AI for the general good and working in a highly collaborative environment resonates with you, take a look:
job-boards.greenhouse.io/anthropic/jobs…
English
Ted Moskovitz retweetledi
Anthropic's hosting a social in London in early August for quants & quant engineers interested in a career jump!
Especially engineers: if you spend all day optimizing trading system performance, we are excited to chat to you.
gem.com/form?formID=3d…
English
Ted Moskovitz retweetledi
Excited to share this work has been accepted as an Oral at #icml2025 -- looking forward to seeing everyone in Vancouver, and an extra thanks to my amazing collaborators for making this project so much fun to work on :)
Aaditya Singh@Aaditya6284
Transformers employ different strategies through training to minimize loss, but how do these tradeoff and why? Excited to share our newest work, where we show remarkably rich competitive and cooperative interactions (termed "coopetition") as a transformer learns. Read on 🔎⏬
English
Ted Moskovitz retweetledi
Anthropic's hosting a social in NYC in mid-June for quants interested in a career jump. Sign up link below!
I was a quant trader for 4y, retired 2y, then Anthropic for 4y. Quant trading and retirement were good; Anthropic has been great 🎉
gem.com/form?formID=ed…
English
Ted Moskovitz retweetledi
@alinagbnva, Lars Rollik, Jasvin Kaur, @ted_moskovitz, Joseph Cohen, @EmmettJThompson, Troy Margrie, @ClopathLab & @StephensonJones.
With @GatsbyUCL, @imperialcollege
Read the full paper in @Nature ⤵️
nature.com/articles/s4158…
English
Ted Moskovitz retweetledi
Fascinating new paper on learning dynamics of in-context learners: "Strategy Coopetition Explains the Emergence and Transience of In-Context Learning" by @Aaditya6284, @ted_moskovitz, @sara_drag, @FelixHill84, @scychan_brains, and @SaxeLab
arxiv.org/abs/2503.05631
🫶
English
New work led by @Aaditya6284! This was a really fun and interesting project, and I think there are a lot of cool insights to be had on the interplay between in-context and in-weights learning on the circuit level.
Aaditya Singh@Aaditya6284
Transformers employ different strategies through training to minimize loss, but how do these tradeoff and why? Excited to share our newest work, where we show remarkably rich competitive and cooperative interactions (termed "coopetition") as a transformer learns. Read on 🔎⏬
English
Ted Moskovitz retweetledi
And I am excited to announce that I have joined Anthropic, Switzerland! 🇨🇭
Anthropic is setting up a new office in Zurich, expanding its global presence. I am super excited to build a team here, where we will be working on LLM training, focussing on developing multimodal capabilities...
English
Ted Moskovitz retweetledi
If you’re interested in interpretability of LLMs, or any other AI safety-related topics, consider applying to Anthropic’s new Fellows program! Deadline January 20, but applications are reviewed on a rolling basis so earlier is better if you can. I’ll be one of the mentors!
Anthropic@AnthropicAI
We’re starting a Fellows program to help engineers and researchers transition into doing frontier AI safety research full-time. Beginning in March 2025, we'll provide funding, compute, and research mentorship to 10–15 Fellows with strong coding and technical backgrounds.
English
Ted Moskovitz retweetledi
Apply to join the Anthropic Fellows Program!
This is an exceptional opportunity to join AI safety research, collaborating with leading researchers on one of the world's most pressing problems.
👇
alignment.anthropic.com/2024/anthropic…
English
Was really lucky to help out on this project at the end of my PhD—huge congrats to @Aaditya6284 @AlbertSYue @djstrouse @louvishh on getting it over the line!
Aaditya Singh@Aaditya6284
Math reasoning benchmarks keep getting saturated… Excited to introduce HARD-Math: Human-Annotated Reasoning Dataset for Math. Consisting of 4,780 short answer problems, based on the AHSME, AMC, & AIME contests, HARD-Math still poses a challenge for frontier LLMs. Read on 🔎⏬
English
Ted Moskovitz retweetledi
Claude 3.5 Haiku is now available on our API, Amazon Bedrock, and Google Cloud's Vertex AI.
Haiku is fast and particularly strong at coding. It outperforms state-of-the-art models—including GPT-4o—on SWE-bench Verified, which measures how models solve real software issues.
English
Ted Moskovitz retweetledi
Ted Moskovitz retweetledi
We finally formalize the RASP-L conjecture in this work!
- theoretical guarantee on generalization for C-RASP tasks
- validated on 8 algo and 17 finite-state languages
- C-RASP is based on communication complexity, can only transfer O(logN) bits b/w inputs
Thread for details👇
Michael Hahn@mhahn29
When do transformers length-generalize? Generalizing to sequences longer than seen during training is a key challenge for transformers. Some tasks see success, others fail — but *why*? We introduce a theoretical framework to understand and predict length generalization.
English