When #AI operates in the physical world, “pretty close” may not cut it. LIDS researchers developed HardFlow, a method that helps pretrained generative AI satisfy hard constraints while improving solution quality — without retraining. bit.ly/4xydxBN

Thinking about adding machine learning to your engineering toolkit? Start by hearing from the person who teaches it. On Wednesday, September 9 at 12pm ET, MIT professor Youssef Marzouk will walk through MIT xPRO's machine learning, modeling, and simulation: engineering problem-solving in the age of Ai. You'll learn what to expect from the program and get to ask the course team directly. Register now: bit.ly/3UD0lxE

🗳️ What is AI telling voters about elections? MIT’s new LLM Election Observatory, from LIDS PI @charapod and collaborators, is tracking how major AI models answer questions about candidates and issues — and how those answers evolve throughout the midterms. bit.ly/4dwYpx7

When it comes to AI-assisted medical diagnosis, one size may not fit all. A new study from @MarzyehGhassemi and team found that AI explanations affect clinicians and non-experts differently—highlighting the need to design AI support around user expertise. bit.ly/4gv0jR7

📣 Planning to submit our #NeurIPS2026 workshop? we are using OpenReview this year, & account creation can take up to 2 weeks. We recommend creating your OpenReview account by Aug 15 to ensure you can submit on time for the Aug 29 deadline. 👉Learn more: climatechange.ai/events/neur…

Hiring a postdoc at @MIT 🚨 Compositional design, applied category theory, optimization → autonomy stacks, control architectures, multi-agent systems, agentic AI, mobility & energy networks. Rolling review, first pass Sept 15 👉

The 2026 volume of the Annual Review of Economics is now online. Take a look at the most read article, "Large Language Models: An Applied Econometric Framework" by @jensottoludwig, @m_sendhil, and @asheshrambachan bit.ly/4wI35an

Congratulations to LIDS PI Sasha Rakhlin, named director of MIT’s Statistics and Data Science Center! @rakhlin aims to strengthen connections across MIT and advance the rigorous foundations of data science and #AI. MIT News: bit.ly/4gcm08m

MIT Researchers combined an efficient algorithm with dedicated hardware to rapidly generate 3D maps for navigation using minimal memory and power. news.mit.edu/2026/new-chip-c…

Inspired by pinecones, tree bark, and seedpods, MIT researchers developed a mathematical framework for designing manufacturable, adaptive materials that mimic complex behaviors found in nature — with applications from robotics to aerospace. bit.ly/4zr0VOK

Deep learning runs on math, but we've never had a formal language for architectures! Now we do. Our new work is out on @TmlrOrg , led by @Vincent Abbott 🎉 arxiv.org/pdf/2604.07242

Using reason, again and again A profile explores the work of LIDS PI Brian Hedden, whose research connects ethics and computing. From "Time-Slice Rationality" to MIT's SERC initiative, Hedden is helping shape a more thoughtful future for AI and computing. bit.ly/4wadCuy

Ashia Wilson, a professor in @MITEECS, is a recipient of the 2026 Junior Bose Award. The award is given annually to an outstanding contributor to education. 📸: Conor McArdle engineering.mit.edu/news/ash…

MIT LIDS PI Dimitri Bertsekas, whose groundbreaking work shaped optimization, control, reinforcement learning, and AI, has passed away at 83. A prolific author, educator, and mentor, his impact will continue through generations of students and researchers. bit.ly/3Rg1fip

Following questions where they lead. A new profile highlights Bailey Flanigan's interdisciplinary journey—from medicine and economics to computer science and political science. Today, she develops computational tools to strengthen democratic participation: bit.ly/452IlyR

A faster way to sample from #AI diffusion models from @MITLIDS @FanChen0111, PI @rakhlin, with @MIT_CSAIL PI @KonstDaskalakis and Sinho Chewi. Read the paper that won the Outstanding Paper Award at #icml2026 @icmlconf

Helping AI meet the real world @MITLIDS Devavrat Shah develops AI that supports continuous decision-making under real-world constraints. His work led to Ikigai Labs, creator of a foundation model for tabular and time-series data, now part of @Celonis More: bit.ly/3Rnx60p

🚢🌊 Our latest paper introduces ORCA (Observation-informed Real-time Correction with Attention), a transformer for correcting numerical weather prediction (NWP) over the ocean. 🧵

Keeping kids safer from illegal AI-generated content. @MITLIDS, BU, @thorn researchers developed a way to evaluate genAI models for harmful capabilities—without generating illegal outputs. The method could help auditors identify risky open-source models. bit.ly/4vAqCto

In game theory, generalists sometimes beat specialists. @MITLIDS researchers found that policy gradient methods can outperform specialized game-theoretic algorithms in certain imperfect-information games—challenging conventional wisdom about strategic #AI. bit.ly/4faViv1