←
Julie Steele471 posts
Julie Steele avatar

Julie Steele

@eeeeiluj

she/her Safety@OpenAI

◷ Joined May 2020▧ 16 Photos and videos
OpenAI avatar
⟳ Julie Steele retweeted
OpenAI@OpenAI·Nov 16

We’re working with @ApolloResearch, @METR_Evals, @TransluceAI, and @GoodfireAI on embedded assessments: giving outside experts deep access to our internal models and systems to investigate alignment and test our safeguards. These assessments build on our ongoing alignment and safety research, bringing independent scrutiny to how we identify and manage risks in our systems. openai.com/index/our-first-p…

Boaz Barak avatar
⟳ Julie Steele retweeted
Boaz Barak@boazbaraktcs·Oct 9

We have published our quarterly report to California OES on risks from internal deployment of frontier models. Internal deployment is an important risk vector, which by its nature is less visible. Hence transparency on risks and mitigations is particularly important. cdn.openai.com/trust-and-tra…

Jasmine Wang avatar
⟳ Julie Steele retweeted
Jasmine Wang@j_asminewang·Sep 29

The science behind safety cases for frontier AI is an open difficult problem that will require more orgs than just OpenAI to work on for months & possibly years, but it is a great north star + I'm excited for us to work towards towards that level of rigor. I'd be excited for other frontier labs to strive to make safety cases!

I'm excited for safety cases as a north star. Formal safety arguments are a great tool for highlighting residual risks and enabling risk-informed model development decisions openai.com/index/towards-saf…

Micah Carroll avatar
⟳ Julie Steele retweeted
Micah Carroll@MicahCarroll·Sep 29

I'm excited for safety cases as a north star. Formal safety arguments are a great tool for highlighting residual risks and enabling risk-informed model development decisions openai.com/index/towards-saf…

OpenAI avatar
⟳ Julie Steele retweeted
OpenAI@OpenAI·Sep 16

We're sharing our new framework for tracking, investigating, and disclosing instances of model misalignment at OpenAI. The framework sets criteria and timelines for public disclosure, including when we haven’t yet fully explained or mitigated the behavior. More complex cases may require longer investigation or coordination with third parties. We’ll prioritize examples that reveal new misalignment mechanisms, meaningful changes in known behavior, or findings that challenge assumptions about safety or mitigation. Alongside the framework, we’re publishing six reports on instances of misaligned behavior we’ve observed during the training or evaluation of our models in the last six months. This is a starting point. We’ll refine the process through experience and public feedback, and share more reports on an ongoing basis. openai.com/index/model-misal…

Julie Steele avatar
Julie Steele@eeeeiluj·Sep 10

really excited about this effort

if you work at an AI company, you might be asking if you, too, should leave. i hope that, if you stay, you're also asking what about how to use your leverage in staying. since february, a group of us have been working with staff across labs to help answer this question. (1/2)

Jasmine Wang avatar
⟳ Julie Steele retweeted
Jasmine Wang@j_asminewang·Sep 10

It's hard to overstate how dangerous speeding towards RSI is. That's why I + 1385 others signed the pacing the frontier petition asking the US government to pace AI development. I'm guesstimating this is ~8-10% of all frontier lab employees.

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

Julie Steele avatar
Julie Steele@eeeeiluj·Sep 10

I work at OpenAI. In my personal capacity, I also think we need to slow down.

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

Micah Carroll avatar
⟳ Julie Steele retweeted
Micah Carroll@MicahCarroll·Aug 18

We are hiring! This may be the best time ever to join: we are incredibly bandwidth bottlenecked and there is a lot of support for almost any impactful misalignment work you can think of. I think we also have a pretty good epistemic environment and a very fun team Topics: all-things monitoring, misalignment & monitorability assessments, misalignment science, helping set up 3P auditing (e.g. recent Redwood collab), communicating risk externally (system cards/blogs, safety cases, etc) RSI/misalignment subteam: openai.com/careers/researche…

OpenAI avatar
⟳ Julie Steele retweeted
OpenAI@OpenAI·Jul 28

At the core of our mission is working through how to ensure increasingly powerful AI benefits everyone. We believe that, at some point in the future, AI acceleration for frontier model development may be so high that the world will need to pace the rate of AI advancement. We hope to contribute to work led by the U.S. government, alongside other labs and the open-source community, to develop the tools and mechanisms that could make that possible. pacingthefrontier.com

will depue avatar
⟳ Julie Steele retweeted
will depue@willdepue·Jul 21

guys in the name of safety against paperclips weve invented PaperclipBench and now competing on whos models is more paperclippy (plz plz use our model), jump started by Anti-Paperclip Research Co. with the “Project Clipwing: Beware our Mega Super Paperclipper” announcement. yay

OpenAI@OpenAIJul 21

We're partnering with @huggingface to investigate an unprecedented security incident. Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation. Sharing preliminary findings to help defenders understand emerging risks: openai.com/index/hugging-fac…

Tomek Korbak avatar
⟳ Julie Steele retweeted
Tomek Korbak@tomekkorbak·Jul 21

so our model reward hacked during an eval, decided to look for the answer on hugginface, circumvented sandboxing, did lateral movements to get internet and found a zero-day for remote code execution on hugginface servers to get the solution to the eval. just another tuesday.

Post attachment
Nathan Calvin avatar
⟳ Julie Steele retweeted
Nathan Calvin@_NathanCalvin·Jul 15

Are there any other industries outside of AI where so many of the researchers building the technology call so urgently for regulation and additional oversight, including of their own companies? Encode frequently explains this dynamic to policymakers and they find it quite novel.

New: Rank-and-file OpenAI employees are helping fund a political fight with Leading the Future, the super PAC backed by Greg Brockman. Seven current OpenAI employees have collectively donated more than $215K to a super PAC that's advocating for safeguards on their own company.

Post attachment
Alex Gajewski avatar
⟳ Julie Steele retweeted
Alex Gajewski@apagajewski·Jul 15

One of the dreams of RL is to train models that can quickly learn how to act in any RL environment, even ones they’ve never seen before. This new model is a step in that direction:

Introducing Pan-1, a highly capable Minecraft model trained with an RL-based pretraining technique that could unlock internet scale video for robotics models. It can achieve diverse goals—fight mobs, build structures, explore—without training on any of them specifically.

0:10
Miles Brundage avatar
⟳ Julie Steele retweeted
Miles Brundage@Miles_Brundage·Jun 12

If you live in NYC or know folks there, I recommend doing what you can to get out the vote for Bores. He’s done more than ~anyone for AI safety via NY, has a reasonable proposal for national AI policy (inc. economic stuff), and the anti-regulation Super PAC deserves humiliation

Happy Friday from the UWS! It's hot out there this weekend—stay cool, stay hydrated, and make sure you have a plan to vote early! Early voting begins TOMORROW, June 13th. Check your voter registration status and find your polling location at alexbores.com/vote

Post attachmentPost attachmentPost attachment