📢 Curious how AI can enhance—rather than replace—scientific principles? Join us this fall for the second iteration of our course! Know someone who might be interested? Tag them or share this post! #AIforScience #AI4Science cc: @um_midas @UM_MICDE @UMichCSE

Michigan AI faculty Michael Wellman, Mithun Chakraborty, @_arodriguezca & @BondiElizabeth played a key role in organizing #AAMAS2025 last week, bringing together 600+ #AI researchers. A fantastic week of collaboration & cutting-edge research! 🌍🤖 https://t.co/Kh6B2oSAcC

🚀 Day 2 of #AAMAS2025 has officially kicked off! Thrilled to announce Prof. Michael P. Wellman as the Local Chair & Mithun Chakraborty as the Local Co-chair. Ready for a day full of insights, innovation, and inspiring conversations!

📌Thrilled to announce that Profs. @_arodriguezca and @BondiElizabeth will give keynote talks at the upcoming Autonomous Agents for Social Good (#aasg2025) workshop @AAMASconf! More details: https://t.co/ZBu3mZNZgK

Cheers to the #AI in Science Fellow Haotian Chen (mentored by Prof. @_arodriguezca) & team fellows for being awarded $100K by the Schmidt Sciences Community Initiative Fund to organize a workshop focused on Knowledge-Guided #MachineLearning (KGML)!🎆🎊 https://t.co/nxs3netdPH

Topics will include time series, uncertainty quantification, and AI for science (physics-informed ML, differentiable simulators). We talk about both methodologies and real-world applications that have been successfully deployed. Join us in Zoom if you’re around.

🎙️Excited to speak today at 1 PM EST at the @ColumbiaDBMI Columbia Department of Biomedical Informatics Seminar Series! I’ll be discussing how we can integrate AI, data, and scientific models in epidemiology for public health prediction. https://t.co/ORCA75cmbN

cc to our team: @kaiwang_gua @AnanyaAJoshi Aparna Taneja

Some topics to cover include anomaly detection at scale, spatio-temporal ML, scientific ML, AI for agent-based models, decision-focused learning, multi-armed bandits. Keep an eye on our site for the tutorial slides! https://t.co/xOR172wtxl

Tomorrow 2/26 at #AAAI2025 we are delivering a new tutorial focused on #AI for decision making in #PublicHealth. We will talk about AI technical challenges across the entire decision making pipeline: data, prediction, decision, and deployment

🚨 Don’t miss out at #AAAI25! Join @_arodriguezca for an exclusive tutorial on Data-driven Decision-making in Public Health and its Real-world Applications.⬇️ Details: https://t.co/P7BukzD0jM 👉 https://t.co/KBX7OVu8Kp

We hope this sparks further research in #multimodal uncertainty quantification. Come see our #AAAI2025 paper a the poster session this Friday at noon! Kudos to my PhD student, Ruipu Li @RuipuL, for his excellent work on his first PhD project!

We design new control-inspired loss functions to enable the NN to effectively act as a controller for the intervals while ensuring theoretical coverage guarantees. Once the NN functions as a controller, we can integrate multiple data modalities to further enhance UQ.

To achieve this, we use an online setup that monitors instances where the uncertainty bounds fail to cover the actual ground truth. Our goal is to keep both realized (long-term) and running (short-term, within a window) coverage as close as possible to the target coverage.

Come see our #AAAI2025 paper at the poster session this Friday at noon! 📑 https://t.co/X4tI6ItYGM Our work shows that NNs can effectively adapt predictive intervals in #ConformalPrediction for #TimeSeries from a control perspective + incorporate #multimodal data.

Join us tonight!

Flu season isn’t here just yet, but today’s FridayNightAI “AI vs the Flu” promises to be a timely & engaging conversation—with @_arodriguezca & Adam Lauring. Ann Arbor locals: skip braving the 10°F cold and join us inside the cozy @aadl library for a “seasonal” AI discussion! ⬇️

You can find the talk slides here: