Hot off the press: DCRI’s Sreekanth Vemulapalli presents ADAPT-CEC, an adaptive AI algorithm that outperformed GPT-4 in CV event adjudication & nearly matched human CEC results in the EUCLID trial. A hybrid AI + human approach proved most accurate. In Circulation: www.ahajournals.org/doi/10.1161/...
A photograph of a screen displaying answers to an online poll asking “in one word how would you describe your coding experience”. The highest response answers are beginner, intermediate, novice, but some other less common response include nonexistent, rusty, and peter principle.
A tip of the hat to the person who entered “peter principle” in response to the the question “in one word how would you describe your coding experience”
Three people sitting around a circular table, working on an Apple laptop. The setting is an office with soft, blue lighting and partitioned screens in the background.
Duke University School of Medicine researchers built an AI model that predicts youth mental health risk before symptoms escalate.
Help this Duke project advance in #STATMadness — vote daily: www.statnews.com/feature/stat-madness/bra...
A new perspective in JAMIA by Duke's Michael Cary, Nicoleta Economou, and colleagues examines the National Academy of Medicine’s AI Code of Conduct. The paper highlights a key gap in health AI governance: the lack of explicit nursing representation. academic.oup.com/jamia/advance-
Photograph of my copy of the book “A Giant Leap” by Robert Wachter, a hardcover book sitting on top of a beige fabric ottoman.
I finished reading Robert Wachter’s new book “A Giant Leap” and it was a wonderful and engaging read. Thank you @bobwachter.bsky.social for your overview of the current landscape and all of its complexities. (I especially enjoyed seeing your conversation with @robcaliff.bsky.social in chapter 6!)
I enjoyed working with an all-star team – the great Lisa Rotenstein of UCSF and David Bates of Brigham & Womens – on this paper describing how to measure return on investment (ROI) for clinician-facing AI tools.
jamanetwork.com/journals/jam...
“Even if you are skeptical about AI’s promise or agnostic about its relevance, the way others use it will shape higher education. Slowing down is not about becoming an AI advocate. It is about developing habits that protect judgment, highlight blind spots and reduce harm.”
Don't miss the opportunity to join our 13th Duke Machine Learning School, directed by Ricardo Henao. The student seats are almost sold out, and others are going fast.
The ML School will held March 9-11 in person at Duke: aihealth.duke.edu/2026-mlss-he...
Recommended post and thoughtful questions to ponder!
A great new story by Duke's student-led newspaper @dukechronicle.bsky.social about the new Discovery AI initiative! www.dukechronicle.com/article/duke...
Photograph of my copy of the book “Crossing the Chasm” with its corner crossing a seam in my kitchen table
Just finished reading the classic “Crossing the Chasm” by Geoffrey Moore (3rd edition). A wonderfully engaging read and helpful model in understanding disruptive technology. #booksky
Coffee and improved cognition, reduced dementia
>130,000 people followed 37 years
Benefit seen only with caffeinated coffee or tea and most pronounced ~2 cups/day
@jama.com ☕️☕️jamanetwork.com/journals/jama/fullarticl...
Dr. Califf’s latest substack is a compelling personal story and call for action to ensure that all children have the best possible chance, not just luck: open.substack.com/pub/robcalif...
I like seeing summaries like this one that integrate findings from practice, theory, empirical work, and neuroscience. Each domain provides a type of evidence and when they cohere, we can have greater confidence in making practice recommendations. #PsychSciSky #AcademicSky #EduSky
A great post on the current state of AI use in instructional videos by by colleague Steve Toback: open.substack.com/pub/stevetob...
This August as I sheltered indoors from wildfire smoke, I got a moving note from MIT statisticians - my disability story had informed a fundamental insight about wildfire smoke forecasting that could someday help AI systems protect public health.
Here's the story:
natematias.com/portfolio/20...
We are thrilled to announce that our NEW Large Language Model will be released on 11.18.25.
AI Grand Rounds Episode 34 From Clinician to Chief Health AI Officer: A Conversation with Dr. Karandeep Singh
New AI Grand Rounds!
Dr. Karandeep Singh explains how early experiments with code led him to biomedical informatics, why gaps between paper performance and clinical reality must be confronted, and how governance committees weigh ethics and safety. Listen to the full episode: nejm.ai/ep34
I’m looking forward to hearing @robcaliff.bsky.social’s views on the “Non-learning health system” this afternoon! rethinkingclinicaltrials.org/news/septemb...
"It's not just the technology -- the human element and how humans interact with the technology is also important." -@aaronecarroll.bsky.social said while moderating the #hdpalooza
Really good (gift) article on #GenAI and cheating. I appreciate how Victor Lee brings data, research history, and careful thought to this issue. We should be teaching individual responsibility and creating environments that decrease the likelihood of cheating. #PsychSky #EduSky #AcademicSky
Epic UGM 2025 FOMO generator. AI scribes by UCDavis and Ohio Health
AI scribes all the time! What are organizations thinking about this rapidly evolving area? how about a balanced scorecard? Cost is a big deal.
🧪 When doctors track heart disease, they need scans over time—but aligning them is tough, tedious, and subjective. We built an algorithm that automates alignment, saving time and effort.
Read more 👉 duke.is/align24
@dukeengineering.bsky.social #research #comphealth #CCDHI #cardiovascular
"The question is, again, what are the consequences of it?"
The Verge's Hayden Field interviews Duke AI Health Director Michael Pencina for a story investigating a possible hallucination in a paper highlighting the capabilities of Google's Med-Gemini LLM: www.theverge.com/health/71804...
Figure 1. Example of a Grounded Report from PadChest-GR.
A novel dataset designed to train and evaluate grounded report generation models from chest x-ray images includes comprehensive sentence-level bounding-box annotations for all clinically relevant findings in each image. Learn more: nejm.ai/44rk1a1
@dccastr0.bsky.social #AI #MedSky #MLSky
Duke University School of Medicine researchers have developed two pioneering studies that offer a new approach to ensuring that artificial intelligence systems used in clinical settings meet the highest standards of quality and accountability.
#DukeB&B #AcademicSky #MedSky #AI
Excited to announce we are launching a Graduate Certificate in Society-Centered AI at Duke w/coursework in computer science, social science, and philosophy/ethics and provide support to 12 current Duke students each year thanks to a new $2.4 million grant interdisciplinary.duke.edu/news/equippi...
In the overflowing room for the Triangle AI Summit, an enthusiastic show of hands for people who have used LLMs!
Artificial intelligence (AI) & generative AI has the potential to address challenges facing the health care system, including clinician burnout, operational inefficiencies, and concerns about patient safety. But barriers remain.
More: academic.oup.com/jamia/advanc...
@dukeaihealth.bsky.social
Flyer displaying an overview of the 12th Duke Machine Learning Summer School
Please join us for next month’s Duke Machine Learning Summer School in Generative AI! Learn more at aihealth.duke.edu/2025-mlss/
