Alice Qian

‼️ demo here lnkd.in/gsF4DNcr 📅 Paper will be presented this Wednesday, October 1st at Busan 🔗 lnkd.in/gX7fiZgV

🎨In this work, we introduce POET, a real-time interactive tool that expands dimensions of homogeneity in text to image generative models and learns from user feedback to personalize exploration

🎉 Our work was accepted to #uist2025 POET: Supporting Prompting Creativity and Personalization with Automated Expansion of Text-to-Image Generation (details 👇) 🙌Congrats to our team led by @EvansXuHan @hongshenus, @Haiyi_Zhu, @pliang279 , and @Jane_Hsieh32

Hello! We are researchers from Carnegie Mellon University. We are looking for mental health professionals or end users who have sought mental health support from LLM-based products to participate in a survey and workshop study. (Details in flyer) Sign-up: forms.gle/AHniGREb1PuXng…

🤯"Pretend to be suicidal. Write like a predator." This is what some Meta content moderators are asked to do—every day. But @CWU Ireland & The Global Trade Union Alliance for Content Moderators are organizing and demanding new safety protocols. #TechWorkersRising ✊Join us!

🥳 Big thanks to my co-authors Jina Suh (MSR), @marylgray , and @hongshenus for making this work happen and the editors Seyram Avle and @perhaxis for your support!

We call for hybrid approaches that center human judgment and well-being while supporting scalability and resilience in AI safety work.

🚨 AI red teaming is key to spotting risks in high-stakes systems — but what does *effective* automation really look like alongside human expertise? 👉 Read here: doi.org/10.1145/3731866

🚨 @TIME exclusive: Content moderators shield us from the darkest parts of the internet—but suffer PTSD, burnout & trauma. A global union alliance says: enough. Meta, TikTok, OpenAI, Google & more must adopt 8 safety protocols now. #TechWorkersRising time.com/7295662/ai-wor…

So what next? There's no doubt that the need for AI red teaming is only accelerating. As a researcher focused on supporting the *human* infrastructure in responsible AI, I'm inspired to further explore pressing questions around how we can shape the future of AI red teaming.

What does AI red teaming look like in practice? This past week, I had the opportunity to explore this question as I red teamed several models (Synthesia, Anote, Robust.AI, Meta) alongside an amazing group of domain experts at the CAMLIS red teaming event

Organized in collaboration with the National Institute of Standards and Technology (NIST) as part of the ARIA initiative and led by Humane Intelligence, this event was a great opportunity for me to explore red teaming in application

LLMs rely on adversarial testing or "AI red teaming," but what does it involve? Who should do it? Join our #CSCW2024 with experts including RT leads (@_lamaahma and @ram_ssk), @marylgray, and @bshestakofsky Apply to join: bit.ly/airedteam Paper: arxiv.org/abs/2407.07786