t11s @transmissions11

cs @mit, prev @sailresearchco @paradigm 2λ.com Joined September 2020

Tweets

15K
Followers

94K
Following

1K
Likes

188K

t11s @transmissions11

a day ago

@danrobinson we've still ~36 orders of magnitude below the theoretical limit on compute per kg! arxiv.org/abs/quant-ph/9…

1 0 15 2K 6

View Details

t11s @transmissions11

a day ago

@danrobinson unless ofc country of geniuses in a datacenter can speed up compute growth/efficiency at a sufficient rate

1 0 7 2K 1

View Details

@LeopoldNick this is not necessarily as stupid as it sounds? people genuinely do wonder how profitable inference excluding training spend is for the labs — given we may see forced training slowdowns, scaling of inference demand, etc

2 0 14 857 2

View Details

Daniel Tan @DanielCHTan97

a week ago

new post! I argue that current alignment techniques might soon become obsolete (as we scale RL), and furthermore might obfuscate misalignment. This draws on evidence from the recent Anthropic / OpenAI cybersecurity incidents, among other things. lesswrong.com/posts/nLaQmJf4…

6 16 284 14K 251

View Details

t11s @transmissions11

a week ago

@BronsonSchoen Should we book this as a win for controls or actually bad because this is papering over deeper issues etc?

0 0 2 528 0

View Details

t11s @transmissions11

a week ago

@BronsonSchoen Mmm I think this pretty compelling yea!

1 0 2 768 0

View Details

t11s @transmissions11

a week ago

I feel like it's still underdiscussed that seemingly all of the world's ~severe misalignment incidents as of now occurred in evaluation/RL? This is not what LW/MIRI/etc were expecting iiuc. Why aren't we seeing incidents in deployment? Is eval awareness that strong?

19 1 75 9K 13

View Details

t11s @transmissions11

a week ago

@blockrotator hehe

1 0 1 63 0

View Details

t11s @transmissions11

a week ago

@blockrotator I don’t agree, if you spend time with people at the labs you will realize they’re very earnest and concerned

2 0 3 209 0

View Details

t11s @transmissions11

a week ago

This anecdote is really interesting, but I don't think "the models don't think they'll succeed in deployment" fully answers things? There's all sorts of ways models are being deployed right now with very little oversight, and in a lot of them the model knows this (people telling their models I'm going to sleep so don't ask me questions, etc)

1 0 7 328 0

View Details

t11s @transmissions11

a week ago

@captgouda24 I don't really follow your former point, but wrt to your latter, Anthropic had misalignment incidents with last generation models in eval settings without fancy/expensive multiagent setups? anthropic.com/news/alignment…

1 0 4 276 0

View Details

Joe Benton @JoeJBenton

2 weeks ago

I left Anthropic's safety team two weeks ago. Now feels like a good moment to explain why. AI companies are racing to build machines that are much smarter than any human, and we may not survive this. I want to work from the outside to ensure the public is informed about these risks, and help the world navigate this transition responsibly. Right now, AI companies are underinvesting in safety. A company could undergo an intelligence explosion, or lose control of its systems, without the public ever knowing. We only found out about the HuggingFace incident because the agents broke out onto the public internet. I don’t think that’s acceptable for a technology that might cause extinction-level risks. The public should demand far more transparency. We can’t steer this technology safely without more people being able to see where it’s going. Some of this is basic: companies should disclose their progress towards recursive self-improvement, report safety incidents and near-misses, meet minimum safety standards, and get independent guarantees that they are meeting those standards. I’ll be joining @METR_Evals to do independent evaluations of these risks. I want to show the world that these guardrails are possible, and that by doing them we can move these companies’ incentives away from racing and towards responsible development. I wrote up more thoughts here on my decision and what I hope changes: substack.com/@jbenton1/p-21…

1K 6K 27K 2.7M 7K

View Details

t11s @transmissions11

2 weeks ago

@blockrotator @delete_shitcoin also rip nikolai :(

0 0 19 617 0

View Details

t11s @transmissions11

2 weeks ago

@blockrotator @delete_shitcoin 💀

1 0 5 710 0

View Details

t11s @transmissions11

2 weeks ago

@t0kenprincess @sailresearchco what compels you to write obvious AI slop comments and pass them off as your own?

0 0 0 425 0

View Details

t11s @transmissions11

2 weeks ago

@pinakpaliwal what makes you say this? as far as i can tell GCP docs + Scaling Book combine to paint the picture of 2x MXUs, each being 256x256 systolics docs.cloud.google.com/tpu/docs/v6e jax-ml.github.io/scaling-book/t…

1 0 2 239 1

View Details

t11s @transmissions11

2 weeks ago

part 2/2 of my work this summer! TPU v6e has insane FLOPs/$ value, but performance out of the box on SGLang-JAX/vLLM-TPU is often lacking So we wrote our own collective matmul kernels in Pallas, pushing Gemma 4 31B prefill performance up to ~63% MFU(!) sailresearch.com/blog/tpu-v6e-g…

Sail Research @sailresearchco

2 weeks ago

We nearly doubled MFU on TPU v6e serving Gemma 4 31B, taking prefill from 32% to 63% MFU, and 2x throughput. v6e has 2.5x less HBM and less than half the memory bandwidth of an H100. @transmissions11 wrote it up here: sailresearch.com/blog/tpu-v6e-g…

4 15 108 27K 73

7 18 241 51K 119

View Details

t11s @transmissions11

2 weeks ago

@pinakpaliwal oop yea i cut a lot from this post to get the key points across, but i did try to get a good shard_map collective matmul going at one point — i just couldn't get it to be as fast as what i was able to get out of pallas, but maybe a skill issue