Data toxicity can lead to harmful model outputs — and since most evaluations focus on English datasets, we’re underestimating multilingual toxicity in state-of-the-art LLMs. Our team partnered with researchers from Carnegie Mellon University and University of Virginia to highlight this gap.
Allen Institute for AI (AI2)’s Post
More Relevant Posts
-
When should AI not comply? That's what our latest work aims to propose — in addition to safety considerations, Faeze Brahman, Sachin Kumar & their collaborators outline the taxonomy of model noncompliance, and offer CoCoNot, a resource for training and evaluating models’ noncompliance.
Broadening the Scope of Noncompliance: When and How AI Models Should Not Comply with User Requests
blog.allenai.org
-
AI models are more integrated into our daily lives than ever before — making user safety paramount. Introducing AI2's Safety Toolkit, featuring red-teaming framework, a high-quality, large-scale safety training dataset, and content safety moderation tools. #AISafety
The AI2 Safety Toolkit: Datasets and Models for Safe and Responsible LLMs Development
blog.allenai.org
-
It’s time to talk about open — our CEO Ali Farhadi joined open model builders Jonathan Frankle from Databricks, Laurens van der Maaten from AI at Meta, and Jon Turow from Madrona to talk about the future of open-source AI, which is a set of models and artifacts made publicly available to modify and reuse.
-
-
Yejin Choi takes the stage this Thursday at Databricks’ #DataAISummit. If you’re in person or tuning in virtually, watch with us 👉: https://lnkd.in/dgTb63-s
-
-
Our CEO Ali Farhadi joined Ina Fried at the Axios AI Summit in New York this week, speaking about openness in research, how to build trust between AI and the public, and more. Watch the entire conversation here: https://bit.ly/AliAxios
-
-
“Without actual openness, it’s hard to be scientific about the evaluation.” Ali Farhadi weighed in on openness and the future of AI at the Axios AI Summit. Read more about why we believe in a truly open-source future for AI: https://lnkd.in/gVCvneug
Allen Institute CEO: How AI "broke trust" with the public
axios.com
-
Tune in now for Ali at Axios AI NY!
Axios AI NY
https://www.youtube.com/
-
Happening today! Join the Axios AI Summit to hear from our CEO Ali Farhadi about openness in AI.
Axios is bringing our AI Summit to NYC for Tech Week in partnership with Tech:NYC. Hear from leaders who are shaping the future of AI, from finance to media and healthcare & more.
Axios AI NY
www.linkedin.com
-
Axios’ AI Summit is happening tomorrow in NYC and via livestream 🚀 Tune in at 12:10 pm PT, June 5th for a fireside chat with our CEO Ali Farhadi and Ina Fried. We’re excited to join fellow leaders who are shaping the future of AI across industries! Register: https://lnkd.in/e-UVsjDN
Axios AI NY 2024 Livestream
axiosainy2024livestream.splashthat.com
