Figure 1. Schematic Diagram of MedAgentBench Framework.

MedAgentBench, a comprehensive benchmark designed to evaluate the agentic capabilities of large language models in realistic EHR environments, provides a critical foundation for advancing and integrating LLM agents into clinical workflows. Learn more: nejm.ai/45AUXwS

#AI #MedSky #MLSky

Red teaming ChatGPT in medicine to yield real-world insights on model behavior - npj Digital Medicine npj Digital Medicine - Red teaming ChatGPT in medicine to yield real-world insights on model behavior

It’s finally out! We brought a multidisciplinary team of physicians, computer scientists, and engineers to red team LLMs for healthcare uses. And we have shared the dataset! www.nature.com/articles/s41...

Francis Collins, the NIH Director for 12 years, led the Human Genome Project and other NIH efforts for 32 years, resigned today. Key words from his resignation letter
www.nytimes.com/2025/03/01/u...

Today's @usatoday.com front page
First US death from measles in 10 years
First child to die from measles in 22 years in the US
All of this fully preventable

Developing ICU Clinical Behavioral Atlas Using Ambient Intelligence and Computer Vision While computer vision has gained traction in medical applications, models specifically engineered for intensive care unit (ICU) activities are limited. We present Clinical Behavioral Atlas (CBA), a...

Dai and colleagues present the Clinical Behavioral Atlas, a computer vision system that can identify 40 clinically relevant activity categories and 55 object categories solely through RGB video data. Read the full article: nejm.ai/4g8D23N

#MedSky #AI #MLSky

It's not "Breaking News" now, since it was announced by NIH yesterday. But it is breaking.......
washingtonpost.com/health/2025/...

The rise of agentic AI teammates in medicine Medicine is in the dawn of a fundamental shift from using artificial intelligence (AI) as tools to deploying AI as agents. When used as a tool, AI is passive and reactive. Even powerful medical AI fou...

Our new piece in The Lancet on the potential of AI agents in medicine
www.thelancet.com/journals/lan... @jameszou.bsky.social

Tackling Algorithmic Bias and Promoting Transparency in Health Datasets: The STANDING Together Consensus Recommendations Without careful dissection of the ways in which biases can be encoded into artificial intelligence (AI) health technologies, there is a risk of perpetuating existing health inequalities at scale. O...

The authors of a Perspective set out recommendations enabling those creating and/or using health care datasets to identify and transparently acknowledge underlying biases. Read the full Perspective: nejm.ai/429WKJj

#MedSky

This is figure 1, which gives an overview of AI applications in the drug development pipeline.

A Review in Nature Medicine explores the state-of-the-art applications of artificial intelligence in small-molecule drug development, from target identification and drug synthesis up to clinical trial design and conduct. https://go.nature.com/4jFyuVG 🔒

When your lab decides that this quarter’s activity is goat yoga.

Alibaba releases Qwen2.5-Max, an AI model that they claim is better than DeepSeek, ChatGPT, and Meta. Their stock is soaring since its release. Here is link to it qwenlm.github.io/blog/qwen2.5...

This is part of the CGT Access Model, first announced by the Biden-Harris administration in early 2023.

Super exciting application of new outcomes-based pricing models for ultra-expensive cell & gene therapies!

Under this model, manufacturers of Casgevy ($2.2M) and Lyfgenia ($3.1M) have agreed that cost to Medicaid will be a function of drug efficacy in patients.

DeepSeek is a side project 🔥

Thanks Jackson!

Empowering and Serving a Community Imagine A World · Episode

Excited to share the most recent episode from Stanford Knight-Hennessy's "Imagine A World" podcast!

Sydney and I spoke with Christian Tanja and Jordyn Sanico (both members of the Knight-Hennessy admissions team) about their work, guiding values, and improbable facts. Enjoy :)

A vision–language foundation model for precision oncology - Nature Trained on unlabelled, unpaired image and text data, the Multimodal transformer with Unified maSKed modeling excelled in outcome prediction, image-to-text retrieval and visual question answering, pote...

A vision-language foundation #AI model for cancer from 50 million path images, a million path-text images, with high performance for predicting clinical outcomes, relapse, response to immunotherapy and other benchmarks
nature.com/articles/s41...
@nature.com (bad name for model—MUSK)

The GLP-1 family of drugs have anti-inflammatory effects independent of weight loss.
2 new studies shed light on the mechanisms
—brain-gut-GLP axis blocks inflammation induced by TLRs (toll-like receptor)
www.cell.com/cell-metabol...

Week 1 in the DLab done! Just this week I went to Half Moon Bay for their holiday party, played poker with the team, held mice for the first time, saw probes enter & record the brain, and more. So much fun!

Personalized progression modelling and prediction in Parkinson’s disease with a novel multi-modal graph approach - npj Parkinson's Disease npj Parkinson's Disease - Personalized progression modelling and prediction in Parkinson’s disease with a novel multi-modal graph approach

This publication today on multimodal #AI to accurately and precisely predict an individual's heath trajectory, in this case #Parkinson's disease, is a frontrunner for many to come
www.nature.com/articles/s41...

🧬🖥🧪 Here are some healthy reminders on just how much Illumina still dominates the #genomics space... 😐

It would be a real dereliction of duty to our nation's children to invite the return of vaccine-preventable diseases. Bringing polio, measles, diphtheria, under control is something we have been working on for a century -- successfully.