MIT LIDS

Congrats LIDS PI & @IkigaiInc cofounder Devavrat Shah! Ikigai's decision intelligence, simulation, and forecasting capabilities will become part of the @Celonis Platform! Learn how they will continue to build the next generation of Enterprise AI together. bit.ly/4wszdQg

Is your #weather app good enough for deciding whether and when to go outdoors? A recent study co-authored by several MIT CS3-affiliated researchers introduces a new method for evaluating the accuracy and reliability of today’s #airquality forecast tools. tinyurl.com/w7vvsbk3

Games people — and machines — play 🎲 LIDS PI @gabrfarina combines concepts from game theory with such tools as machine learning, optimization, and statistics to advance theoretical and algorithmic foundations for decision-making. Read a profile: bit.ly/4nm4X5G

What’s the TRUE cost of #AI? LIDS PI Priya Donti joins the India Opportunity Show to discuss AI’s growing energy demands, data center expansion, sustainability challenges, and how AI can still drive meaningful climate impact when deployed responsibly. bit.ly/4emauGX

⚖️ 🔨 When removing one bias from an AI model, it can create or amplify other biases. This can lead to what AI researchers call the "whac-a-mole dilemma." A new paper from @aihealthmit researchers proposes a smarter debiasing approach. news.mit.edu/2026/smarter-w…

Inside Efficient AI: From GPUs to GPTs Why is AI so energy-hungry—and how can we fix it? LIDS Affiliate PI Song Han joins MIT President Sally Kornbluth on Curiosity Unbounded to discuss efficient AI, lighter models, and optimizing GPU systems. 🎧 bit.ly/3QYPxYI

Watch the full conversation here: youtu.be/D6f1IMPHCuY?si…

If you are in Rio for @iclr_conf , check out Vincent’s work!!!

For @iclr_conf, I’ll be presenting FlashAttention on a Napkin today (Friday)! Pavilion 3-622 3:15-5:45

Check out our ICLR poster for our TMLR paper which got J2C certification!!!

She's on the TIME 100 AI list, alongside Sam Altman, Jensen Huang, and Satya Nadella. She is building at the intersection of AI and climate, where the real battle for the future is being fought. She is using AI to decarbonise power grids, integrate renewables, and solve real-world climate problems at scale She is none other than @priyald17, an Assistant Professor at MIT, co-founder of Climate Change AI, and a PhD from Carnegie Mellon University. Her work sits at the cutting edge of technology and sustainability. But what truly stands out is how she thinks While the world is obsessed with building bigger models and larger data centres, she questions whether that is even the right path We might be building and thinking about AI in the wrong direction: >> The AI race consists of three races. AGI, profit, and societal good. Most of us are only watching one. >> Data centre ROI isn't really ROI yet. The I is happening. The R is still speculation. >> AI literacy isn't optional anymore, not for founders, not for students, not for policymakers. She highlights that much of today’s AI buildout is speculative, driven by the assumption that value will come later, rather than being grounded in real, present-day problems In a world where AI is increasingly being built for profit and consumption, she is pushing for a version of AI that is responsible, sustainable, and aligned with society’s long-term needs. If there is a new generation of thinkers shaping how AI should truly be built and deployed, she is right at the centre of it. Watch the full conversation only on The India Opportunity Show. Link in comments

We’re hiring a Postdoctoral Associate in AI for hyper-local weather forecasting! 🌍 I’m looking for a postdoc to join the Earth Intelligence Lab at @MITMechE and @mitidss, working on next-generation, uncertainty-aware weather forecasting supported by a NASA-funded project.

Civic Discourse track: Research that strengthens public reasoning, deliberation, and access to reliable information. 📣Deliberation at National Scale, @Harvard @MIT @MITLIDS Meaningful deliberation has never been able to scale. This team is building the infrastructure to change that. @ArielProcaccia @baileyflan @Arfung @lessig @bakkermichiel 🗣Rebuilding Trust in Civic Discourse, @Cornell Not content moderation, but rather trust infrastructure, with a cryptographic auditor layer that makes the whole system inspectable. Jon Kleinberg @nataliebazarova @Cristian_DNM Robert Kleinberg Mor Naaman @DG_Rand

This project builds on an existing open-source video deliberation platform to bring structured, AI-mediated civic conversation to millions of people simultaneously. The five-year goal is a national deliberation involving at least 0.5% of the U.S. population, starting with city-level pilots in Boston and New York. Deliberation at National Scale @Harvard @MIT @MITLIDS @ArielProcaccia @baileyflan @Arfung @lessig @bakkermichiel laude.org/moonshots/one/…

Before autonomous vehicles scale, the science behind them needs an upgrade. Researchers from @MITLIDS, @MIT_CEE, @MITdusp, and collaborators find mobility-on-demand studies often lack transparency and reproducibility—key for real-world decisions. More: bit.ly/4cCrAig

Evaluating the ethics of autonomous systems A new framework from @MITLIDS, @MITAeroAstro & @Saab––SEED-SET—can pinpoint when #AI decision-support systems make unfair recommendations in complex settings like power grids. More: bit.ly/41dEFbr

Smarter traffic control for warehouse robots 🤖 @MITLIDS + Symbotic researchers developed an #AI system that learns how congestion forms and decides which robots should go first—rerouting them in advance to avoid bottlenecks. More: bit.ly/4lWdawE

Check out this great article written by Stephanie Martinovich about our work on AMoD reproducibility. Work led by Xinling Li and Meshal Alharbi 💥 recently published in IEEE Transactions on Robotics!!! ieeexplore.ieee.org/document/11397…

Working on the SLAM Handbook has been one of the highlights of my career — I’m grateful to have collaborated with such an incredible group of co-editors and contributors. The handbook is now free and open-source: dspace.mit.edu/handle/1721.1/… [1/n]

Spotting overconfident #LLMs New work from LIDS Kimia Hamidieh, Marzyeh Ghassemi, and collaborators introduces a more reliable uncertainty metric to flag when models are confident but wrong. This could help users better judge when to trust #AI outputs. bit.ly/4sobRZN