🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

📣New preprint!📣We’ve long known humans tend to anthropomorphize computers. But with the rise of social AI applications, like AI companions, studying this is now more crucial than ever. We introduce a new method for *empirically evaluating* anthropomorphic behaviors in LLMs🧵

I'll be at ACM CHI in Yokohama, Japan next month! If you'll be there too, I'd love to connect to chat about all things at the intersection of HCI and AI for creating beneficial and safe experiences!

Early methods for studying affective use and emotional well-being on ChatGPT: openai.com/index/affectiv…

We made a long list of concrete projects and open problems in evals with 100+ suggestions! docs.google.com/document/d/1gi… We hope that makes it easier for people to get started in the field and to coordinate on projects. Over the last 4 months, we collected contributions from 20+

Red teaming paper that we published late last year is on arxiv! arxiv.org/abs/2503.16431

A lot of folks worked together to get this right. There’s real progress here on what the meat of preparedness should look like. I’m proud of it!

The First Workshop on Sociotechnical AI Governance at CHI 2025 (STAIG@CHI’25) is less than a week away! We’re super excited to hold a panel with some amazing speakers to discuss why a sociotechnical approach to AI governance is so important. Join us in Yokohama or online.

Evaluations are essential to understanding how models perform in health settings. HealthBench is a new evaluation benchmark, developed with input from 250+ physicians from around the world, now available in our GitHub repository. openai.com/index/healthbe…

The pinnacle of safety <> benefits <> expert input. Congrats Karan Singhal Rahul Arora and the whole team who made this happen!

Proud of our team for pushing forward transparency with this initial step!

FutureHouse's goal has been to automate scientific discovery. Now we used our agents to make a genuine discovery – a new treatment for one kind of blindness (dAMD). We had multiple cycles of hypotheses, experiments, and data analysis – including identify the mechanism of action.

“you better not be spreading yourself too thin” me:

We’ve activated our strongest safeguards for ChatGPT Agent. It’s the first model we’ve classified as High capability in biology & chemistry under our Preparedness Framework. Here’s why that matters–and what we’re doing to keep it safe. 🧵