Stanford Institute for Human-Centered Artificial Intelligence (HAI)

Stanford HAI and the AI for Public Benefit Lab are hiring a postdoctoral fellow to study how AI can make a measurable difference in people's lives, across democracy, social services, emotional well-being, and persuasion. The role spans research design, evaluating AI systems for accuracy and neutrality, large-scale field and survey experiments, and collaboration with civic partners, with room to pursue independent work. Learn more and apply: https://lnkd.in/gKse5Bm

A benchmark score is a prediction about how an AI model will behave in the world. Almost no one checks that prediction later. In two new studies supported by Stanford HAI, Stanford Assistant Professor Sanmi Koyejo, graduate student Sang Truong, and their collaborators borrow tools from psychometrics to test whether AI benchmarks measure what they claim to, and find they often don't. Read more: https://lnkd.in/d-WVzqfS

Who shapes AI, and how do you keep people at the center of its development? Stanford HAI Denning Director James Landay spoke with John Werner of Imagination in Action about that question, from educating lawmakers to working with industry while preserving academic independence. They also discuss how Stanford is adopting AI, the competition for research talent, what the AI Index reveals beyond the headlines, and how safety, sovereignty, and open source connect. Watch here: https://lnkd.in/gPAMWXWF

What does responsible AI regulation look like in practice? In a new report, Stanford HAI Associate Director Rob Reich and UC Berkeley School of Information’s Deirdre Mulligan offer the California Department of Technology concrete recommendations for updating the Transparency in Frontier AI Act, from redefining "frontier models" to expanding incident reporting requirements. Read the report: https://lnkd.in/gwcyqFSr

In 2018, Stanford experts said driverless cars were 90 percent of the way there. The remaining 10 percent proved stubborn: seeing through fog and rain, reading road markings of varying styles and quality, and making safe decisions. AI has since helped move self-driving cars onto city streets. But some of the hardest problems remain, including how to prepare vehicles for situations no one can fully anticipate. "As we try to design for more and rarer situations, it becomes increasingly expensive and harder for human designers to conceptualize those," says Stanford HAI Faculty Affiliate Marco Pavone, associate professor of aeronautics and astronautics and director of Stanford's Autonomous Systems Laboratory. In a conversation with Stanford Report, Pavone explains what got autonomous vehicles this far and what could help them go further. Read more in Stanford Report: https://lnkd.in/gwWdaY7B

It was wonderful to see so many Stanford HAI faculty and leaders take part in Imagination in Action this month at the Google Bay View campus. We heard so many important conversations and research updates in AI, from scientific discovery and robotics to governance and the future of work. HAI Leaders and Affiliated Faculty at Imagination in Action: James Landay, Erik Brynjolfsson, Rob Reich, Vanessa Parli, Marco Pavone, Diyi Yang, Surya Ganguli, Melissa Valentine, Arvind Karunakaran, Mehran Sahami.

Empirical Methods in the Age of AI Conference is sold out. If you were hoping to hear from Anastasios Angelopoulos, there's still a way. This Friday, the co-founder and CEO of Arena will present at a free Stanford HAI Seminar on measuring AI in the real world. His work sits at the intersection of theoretical statistics and machine learning, with a focus on model evaluation and reliability guarantees for AI systems whose inner workings can't be directly inspected. Friday, October 2 | 1:00–2:15 PM PT | Gates Computer Science Building, Room 119 or virtually via Zoom Register here: https://lnkd.in/gs36a_Rb

Stanford HAI's AI Index is looking for a Chapter Project Manager who combines project management experience with working knowledge of AI in biomedicine and healthcare. The role spans keeping a multi-author chapter on schedule, checking technical content for accuracy, and identifying emerging trends in AI's use in medicine and science that should be covered. This is a one-year, hybrid position, with 16 hours per week on the Stanford campus. Details: https://lnkd.in/g8TPpGwp

From the genome to the brain, AI is changing how research gets done across the sciences. Each Wednesday this fall, the AI and Data for Science seminar series brings a different Stanford researcher to share how AI and data-driven methods are being applied in their field. Many of the speakers leverage Stanford’s high-performance computing infrastructure, including the Marlowe GPU cluster operated by Stanford HAI in partnership with University IT and the Office of the Vice Provost and Dean of Research. Talks are open to the public and held in person, Wednesdays from 4:30–5:30 PM PT in Stanford's Computing and Data Science (CoDA) building, room E160. September 30 | Gordon Wetzstein | From World Models to World Agents October 7 | Olivier Gevaert | Multimodal modeling for precision medicine (Packard 101) October 14 | Curtis Langlotz | Training Methods for a Medical Imaging Foundation Model October 21 | Anshul Kundaje | Decoding regulatory DNA with deep learning models October 28 | Dan Yamins | Large-Scale Computational Models at the Neuroscience-AI Intersection November 4 | Ruijiang Li | Multi-Modal Foundation Models for Precision Medicine November 11 | Alex Pentland | Detecting coordinated behavior across multiple fully encrypted domains See the full schedule: https://lnkd.in/gYK7uTc4

Policymakers, governments, and business leaders around the world rely on Stanford HAI’s AI Index for rigorous, objective insights into AI’s technical progress, economic influence, and societal impact. The 2026 AI Index report is now available in Simplified Chinese: https://lnkd.in/gVgKeqem