Stanford Institute for Human-Centered Artificial Intelligence (HAI)

Building an AI chatbot for healthcare isn't like building ChatGPT. You need real-time data, strict privacy controls, and integration with complex EHR systems. Teams across Stanford Health Care’s Technology organization came together to build “ChatEHR”, a privacy preserving and practical GenAI tool that could serve as a model for other health systems: https://lnkd.in/eDEAMdwf

Join us online for the 5th annual Stanford AI+HEALTH Conference on December 9-10, 2025. Now in its fifth year, AI+HEALTH continues to be Stanford's largest gathering focused on advancing the responsible use of AI in health and medicine. Co-hosted by Stanford AIMI, Stanford HAI, and Stanford CME, this CME-accredited conference brings together clinicians, researchers, and leaders from academia, industry, non-profit and government to explore real-world applications and strategies driving meaningful impact in healthcare. This year's program offers updated content, the latest use cases, and leading thought leadership across four tracks—covering current clinical applications, implementation strategies, industry insights, and foundational lectures designed to strengthen understanding of AI in healthcare. Early registration ends this Friday, October 31st! Register here: https://lnkd.in/e8nvyBPr #AIplusHealth #StanfordAIHealth #StanfordAIMI #StanfordHAI #StanfordCME

HAI Senior Fellow Michele Elam is one of the new senior associate vice provosts for undergraduate education who will co-lead the AIMES (AI Meets Education at Stanford) initiative. Read more about how Stanford University educators are bringing AI into the classroom: https://lnkd.in/giSWeGFC

During medical visits, doctors are increasingly using AI-powered "ambient scribes" that automatically transcribe conversations and populate electronic health records. While these tools streamline documentation, HAI Faculty Affiliate Nigam Shah recommends patients to ask how their data will be used and what scope of use they're consenting to beyond immediate medical care. In our latest blog, Dr. Shah explains what the healthcare AI landscape looks like from the patient’s point of view and where the hidden risks may lie: https://lnkd.in/g7dB_w2f

Ready to lead the AI revolution in your industry? Our custom-built programs work with world-class Stanford faculty to deliver human-centered AI strategies that drive innovation, while addressing your organization's unique needs. Check out the possibilities here: https://lnkd.in/gsxNwnsS

Whether you're making policy decisions, conducting research, or curious about economic trends, join us for tomorrow's webinar hosted by the Stanford Digital Economy Lab which will discuss the ADP National Employment Report and the importance of data in understanding changes in the economy. The talk will feature a moderated Q&A between HAI Senior Fellow Erik Brynjolfsson and ADP's Chief Economist Nela Richardson. The ADP National Employment Report (NER) is an independent and high-frequency view of the private-sector labor market based on the aggregated and anonymized payroll data of more than 26 million U.S. employees. This monthly report is a free resource for business leaders, researchers, policymakers, and the public. Register to join the webinar tomorrow: https://lnkd.in/gwcuzWWe

Should parents and teachers ban AI tools in education? HAI Faculty Affiliate Hari Subramonyam says no — but with an important caveat. "Help kids use them more intentionally," he advises, "like asking for help with brainstorming or getting feedback." His recommendation: Stay curious alongside students. Ask what choices they made, why they prompted AI in specific ways, and explore different approaches together. Listen to the full conversation in Stanford University Graduate School of Education's School's In podcast: https://lnkd.in/gCUTu9wB

Join us tomorrow for a Stanford HAI seminar with Google's CTO Blaise Agüera y Arcas. He'll challenge the myth of AI as alien intelligence and reframe it as a social phenomenon born from language — the "DNA" of collective human intelligence. https://lnkd.in/gW5xWShk

Over 3.4 million American children need speech therapy, but clinicians are stretched thin — often serving hundreds of students each. Could AI ease the burden? Stanford researchers, including CS graduate student Sang Truong and HAI Faculty Affiliates Sanmi Koyejo and Nick Haber, put 15 leading language models to the test, including GPT-4, Gemini, and Whisper. The results were sobering: none came close to the 80-85% accuracy threshold needed for clinical deployment. But when the team fine-tuned open-source models, performance improved significantly on diagnostic tasks. The results offer exciting promise: https://lnkd.in/g8ZSyFsP

What are some of the biggest challenges of new products like AI browsers? Stanford HAI Co-Director James Landay says anticipating use cases and preferred interfaces is going to be difficult. Unlike traditional tech, AI is unlikely to have one dominant interface; instead, users will likely want to interact through multiple modes, combining both familiar and novel approaches. Read more about his insights via Fast Company: https://lnkd.in/gYMMQnhU