Interested in chatting at ICLR?forms.gle/dbtmF5EeyM5H1d…

We are really excited to host Aleksander Madry from OpenAI at the PASS seminar on 3/26, 2pm ET! Submit your questions about the Preparedness team: tinyurl.com/pass-question, and join our mailing list to receive notifications about talks: tinyurl.com/pass-mailing

The secret is out 👀

Very proud to be part of this Amazing MURI Team with Christos Papadimitriou Columbia University, Erik Demaine, Constantinos Daskalakis, Aleksander Madry Massachusetts Institute of Technology (MIT), Zico Kolter , Russ Salakhutdinov, Katia Sycara Complete Music Update, Kamalika Chaudhuri UC San Diego, and our own Soheil Feizi UMD Department of Computer Science.

Congratulations Shibani Santurkar, this is extremely well deserved!

We are pleased to announce the 2021 AAAI/ACM SIGAI Dissertation Award Winner.

Congratulations to Shibani Santurkar, Massachusetts Institute of Technology for her work entitled Machine Learning Beyond Accuracy: A Features Perspective On Model Generalization.

And congratulations…

Gary Lupyan OpenAI Hi, just wanted to clarify - these are just supposed to be very basic “don’t sue us if you’re rejected” type terms for the grant application. They’re not at all supposed to be a barrier to applying. If your university office has an issue, please tell them to reach out to…

Great news! The US AI Safety Institute is an extremely important effort and looking forward to its thriving under Elizabeth's and Elham's leadership.

#NEWS : Secretary Gina Raimondo announces key members of the executive leadership team to lead the U.S. AI Safety Institute, which will be established National Institute of Standards and Technology.

Elizabeth Kelly to lead the Institute as Director & Elham Tabassi to serve as Chief Technology Officer. commerce.gov/news/press-rel…

New from me: OpenAI says GPT-4 poses “at most” a slight risk of helping people create biological threats. This is the first study from the “preparedness” team (and its leader, Aleksander Madry, told me there's more work TK from this team on other topics). bloomberg.com/news/articles/…

latest from preparedness @ openai: gpt4 at most mildly helps with biothreat creation.

method: get bio PhDs in a secure monitored facility. half try biothreat creation w/ (experimental) unsafe gpt4. other half can only use the internet.

so far, gpt4 ≈ internet… but we’ll…

Our latest update: quantifying how LLMs impact bioweapon creation. Now part a growing set of frontier model evaluations to track and forecast catastrophic risks from AI!

I joined OpenAI, new results from my team just dropped!

AI's impact on biosecurity has been a major aspect of discussion for catastrophic risks.

To quantify this risk, Preparedness @openai is releasing a new study: bio PhDs, a secure facility, 5 hours for 5 harmful tasks, access to spicy (research-only) GPT-4!