andy jones
@andy_l_jones
Followers
7K
Following
1K
Media
274
Statuses
2K
engineering & research. #1 most bugs written at @AnthropicAI DC, SF
Washington, DC
Joined October 2012
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.
5
2
19
i use this chart a bunch in discussions of labor market impacts of AI. it captures two things:.* it's one thing to invent a better plough, and another to invent the combustion engine.* the success of a paradigm shift comes with a lot of survivorship bias
3
2
21
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 🎉.
4
5
146
RT @ElanaPearl: 🧬What are protein language models (PLMs) actually learning about biology? Our paper introduces InterPLM - a framework that….
0
132
0
RT @sleepinyourhat: A big part of my job these days is to think about what technical work Anthropic needs to do to make things go well with….
0
59
0
RT @ajeya_cotra: This is very important to appreciate about the AI risk discourse IMO. I liked this chart from @NateSilver538's new book. A….
0
30
0
Despite working on LLMs for going on four years now, Zed & Sonnet 3.5 is the first time I've found myself using a model all day every day for my work. There's some rubicon it crosses of 'smart enough model' and 'good enough UX' that everything I tried previously fell short on.
Fast Edit Mode: a breakthrough from @AnthropicAI that we're piloting in Zed. It allows Claude 3.5 Sonnet to echo its input far faster than generating new text. The result? Near-instantaneous refactoring. We're collaborating with their team in Zed's open source codebase.
10
23
342
armen (& @liliyu_lili!) lead one of my fave scaling law papers of recent times, excited to see where this goes.
I've left FAIR/Meta, it's time to build.
1
1
15
experiments with attn are a real pain in the ass and my tools-oriented view of how research is done makes me think that'll turn out be a big chunk of why 'figuring out how the heck models use attention' is still a standing problem.
0
0
3
when i asked horace what he was excited about a couple a weeks ago at ICML, he said this. sounded very neat, turns out: very neat!.
For too long, users have lived under the software lottery tyranny of fused attention implementations. No longer. Introducing FlexAttention, a new PyTorch API allowing for many attention variants to enjoy fused kernels in a few lines of PyTorch. 1/10
1
0
25
in the process of constructing a very daft joke today, i discovered daniel dennett's dad is the first star on the CIA memorial wall
0
0
5
I think I've gotten back to everyone now! I got, uh, ten times the inbound that I expected and so many apologies had to be made 😓. If we weren't able to schedule time but you see me around, come up and say hi! . recognize me by my snazzy tue/wed outfits
1
0
1
RT @NeelNanda5: Are you excited about @ch402-style mechanistic interpretability research? I'm looking to mentor scholars via MATS - apply b….
0
24
0
I'll be at ICML next week! I'm a long-time Anthropic engineer/researcher, & I want to meet new folk, so DM me if you wanna walk & talk!. * Academic engineers. If you maintain an open-source LLM lib, if you're the one at your lab who everyone asks to fix their CUDA issues, DM me!.
5
11
178
RT @mackhawk: Key Republican lawmakers asked the Biden admin to prepare a formal intelligence community assessment of Microsoft’s partnersh….
0
3
0
RT @NeelNanda5: Obvious caveat: These are MY takes, and I'm sure other researchers disagree on points, or think I'm missing a crucial paper….
0
21
0