With the free, twice-monthly NEJM AI Email Newsletter, you'll learn how AI will change clinical practice and healthcare, how it will impact the patient experience, and about the people are who are pushing for innovation.
Sign up here:
Follow
NEJM AI
@NEJM_AI
NEJM AI, a dialogue on medical artificial intelligence and machine learning from . #ArtificialIntelligence #AIinMedicine
NEJM AI’s Tweets
Illuminating and fun (and at times high entropy!) conversation with Michael Abramoff and coming to Grand Rounds soon
1
3
18
Large benchmark data sets have been central in accelerating progress in the #MachineLearning community. Although a few notable examples exist, there is a great need for this in medicine. Listen to the latest NEJM AI Grand Rounds podcast for more:
1
5
We’ll be able to have an input image and output a radiologist’s report to the point at which I hope it will be indistinguishable or even better in quality to the physician that needs the report.
– on the future of radiology
Hear more:
1
“We've talked about reading radiology reports as a way to learn. I want to get to writing radiology reports in their full entirety.” – Pranav Rajpurkar of . Listen more: ai-podcast.nejm.org/e/dr-pranav-ra
1
1
How well does your medical #MachineLearning task align with the problem you are trying to solve? In this clip, explains what can go wrong. Listen to the full episode: ai-podcast.nejm.org/e/dr-pranav-ra
4
11
I asked ChatGPT to explain what the kidneys do in the style of Dr Seuss. This response is 🏆🥇
61
566
3,317
Show this thread
Pranav Rajpurkar: “We really need clean data sets that have labels that we can generally trust.” Listen to the full interview with on AI and radiology: ai-podcast.nejm.org/e/dr-pranav-ra
2
5
Introducing ClimaX, the first foundation model for weather and climate. A fast and accurate one-stop AI solution for a range of atmospheric science tasks.
Paper: arxiv.org/abs/2301.10343
Blog: microsoft.com/en-us/research
Thread🧵
#ML #Climate #Weather #FoundationModel
30
198
778
Show this thread
Read the article "Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologists" referenced in the clip:
2
Show this thread
Will #ArtificialIntelligence replace radiologists? In this clip, offers a nuanced answer.
Hear more in the latest episode of NEJM AI Grand Rounds: ai-podcast.nejm.org/e/dr-pranav-ra
1
1
7
Show this thread
Large language models (LLMs) have become a major theme for the broader #MachineLearning community. A recent paper by Google systematically tested LLMs on medical question answering tasks: arxiv.org/abs/2212.13138
5
Recurring theme of our interview with : the importance of shared and useful benchmark datasets. Listen to full episode:
1
3
Stay on top of the latest developments in medical #ArtificialIntelligence and #MachineLearning with the NEJM AI Email Newsletter:
2
6
The latest episode of NEJM AI Grand Rounds features , a faculty member of , who has been at the forefront of medical AI his entire career. His group continues to push the frontier of medical AI. Listen to the full episode:
2
7
Google has created an AI that can accurately answer medical questions nearly as well as human doctors, but the company doesn't think the technology is safe enough yet for use in health care settings
11
19
48
An AI tool known as Sybil accurately predicted the risk of lung cancer for individuals with or without a significant smoking history
2
19
49
Pranav Rajpurkar: “It’s very important to understand the data generating process.” Listen to the full interview with as he discusses the challenges of #MachineLearning and medicine:
4
7
It's story time!
Find me on the other side of the microphone as a guest on the Podcast, hosted by & .
I share lessons learned from my journey as a researcher in AI, and my bets for the future of medical AI.
2
15
39
Episode 2 of NEJM AI Grand Rounds is here!
and are joined with , an assistant professor who leads a lab focused on developing AI capable of highly complex medical decision making.
Listen now:
7
151
1,266
Excited to share our new Grand Rounds episode, an enlightening conversation with of who takes us behind the scenes of a few seminal medical AI papers over the past decade and also gives us a preview of what's next:
7
26
On the next episode of AI Grand Rounds, we sit down with the prodigious Pranav Rajpurkar () to talk about AI in radiology, what he learned about mentorship from , and the impact of large self-supervised models on medicine.
Link:
3
18
"Identification of adverse events in EHRs in the future will probably be performed by means of computerization of triggers and also through leveraging of artificial intelligence."
2
4
ChatGPT is rising rapidly through the academic ranks - from middle author to corresponding in just a few days!
sciencedirect.com/science/articl. h/t
This Tweet is unavailable.
3
9
29
In an episode of , Dr. discusses his experience using #AI for #genomics and #cardiology, and thoughts on the impact of AI on #medicine
Link: podbean.com/ew/pb-sfs7c-13
Hosts:
#machinelearning #healthtech #Health
3
4
Extraordinary new paper from Google on medicine & AI: When Google tuned a AI chatbot to answer common medical questions, doctors judged 92.6% of its answers right … compared to 92.9% of answers given by other doctors.
And look at the pace of improvement! arxiv.org/pdf/2212.13138
138
2,033
8,462
Show this thread
With the NEJM AI Email Newsletter, you'll get:
☑️ Highlights from the NEJM AI Grand Rounds podcast
☑️ What the editors are reading in medical AI
☑️ AI articles from across NEJM Group
☑️ Trending conversations in social media
Get on the list:
3
2
A model named PubMedGPT 2.7B was trained on millions of scientific articles. On a dataset of exam prep questions for USMLE Step 1, the model answered 50% correctly. Read more about this autoregressive language model:
2
25
58
Dr. is a pioneer. In 2010, he led the team that conducted the first clinical interpretation of a human genome. Hear about his experiences applying AI to genomics and to cardiology:
5
Researchers built a #DeepLearning system that can detect anemia, elevated B-type natriuretic peptide (BNP), troponin I, and blood urea nitrogen (BUN), as well as values of 10 additional lab tests directly from echocardiograms. Read the paper in :
3
12
Sign up for our companion newsletter featuring content on how AI will change healthcare, its impact on the patient experience, and the people pushing for innovation:
5
7
Researchers have created a #DiffusionModel capable of generating chest radiographs matching a user prompt. This research is a continuation of diffusion models for image generation popular in graphic design and AI-enabled art. Learn about this research:
1
1
4
At some point in the future can we have pan-diagnostic imaging tests enabled by AI? Dr. shares why he's optimistic that AI and machine learning can help maximize preventive care to intervene before disease occurs.
Full episode: ai-podcast.nejm.org/e/dr-euan-ashl
9
14
What does the clinical adoption of #genomics teach us about the uptake of #ArtificialIntelligence in medicine? Listen to Dr. ’s thoughts here:
1
Dr. : recalls a project where his team created an AI system that could understand cardiophysiology by training it on a dataset of over 1 million echocardiograms.
Listen to the full episode: ai-podcast.nejm.org/e/dr-euan-ashl
1
2
3
A key bottleneck for sequencing a patient’s genome is how long the AI algorithms take to analyze the sequencing data. Hear how Dr. and his team solved this problem and set a Guinness World Record () in the process.
Full episode: ai-podcast.nejm.org/e/dr-euan-ashl
1
3
10
Do clinicians need to understand machine learning to contribute to machine learning projects? : “I think they should understand the basic principles but they do not need to know the details.” Listen to the full episode:
2
8
1
2
“AI somewhat paradoxically—and maybe this is surprising to some—can take the computer out of the consulting room.” – Dr. . Listen to the full interview:
6
13
#MachineLearning models are increasingly able to interpret and generate high-quality images and text. For example, researchers are using “diffusion models” to generate chest radiographs from text prompts (e.g. “big right side pleural effusion”). Read more:
1
3