Week of September 22, 2025
As I was reviewing content for this week's newsletter, the "AI anxiety" theme seemed to permeate. The frequency with which AI plays a role in my clinical time is increasing. Ambient audio transcription and medically oriented LLMs (think ChatGPT attached to an enormous volume of medical articles) are my most frequent tools. And while I am cognizant of superintelligence’s trade-offs, the technology is now giving me more time to focus on the patient (rather than my laptop screen) and the ability to more precisely answer questions I never had an easy way to explore. (Like the recent, "Hey doc, how much green tea is too much and does it interact with my medications?"). I am grateful to experience this shift in my lifetime. While I share the anxieties and concerns expressed in the articles below, I am, on balance, excited rather than fearful.
In case you are wondering, here is the prompt and answer regarding green tea: "Generate a table of the studies that looked at green tea consumption, the benefits, rate the study quality, the patients included, the dosing of green tea, noted side effects, and proposed mechanism of action. Also, please list the contraindications and typical drug-green tea interactions a nephrologist may be concerned about." FYI - Green tea does interact with medicines metabolized through the liver (the CYP450 enzyme pathways).
https://www.openevidence.com/ask/c043b473-674c-4793-834a-028c5f8b5e29
---
Listen to a Google Notebook LM A.I.-generated podcast of the newsletter with two virtual "hosts."
https://drive.google.com/file/d/1JLbeabzA07zCQGK3rL-YDJcTCflIYoAY/view
------
Science and Technology Trends
I found comfort in learning that astronomers are studying how to 'appropriately' smash a spacecraft into the surface of a hazardous asteroid. However, it is not as simple as one might imagine - "Slamming into its surface indiscriminately runs the risk of knocking the asteroid through a 'gravitational keyhole' that sends it back around to hit Earth later." Despite the "early 2000s Bruce Willis movie vibes" this article engenders, there are still a lot unknowns (like how do we test, is it generalizable, and how do we account for various asteroid characteristics (mass, composition, rotation, etc). I had never considered the practical problems of moving asteroids without an Aerosmith soundtrack.
https://www.sciencedaily.com/releases/2025/09/250909031457.htm
and
https://meetingorganizer.copernicus.org/EPSC-DPS2025/EPSC-DPS2025-77.html
and AI summary
https://claude.ai/public/artifacts/c3707bbf-e2e2-449f-88f6-6f3ce80f6227
Assuming we can effectively deflect asteroids, human health will continue to matter. In that spirit, I found the Phase 3 study data (double-blinded, randomized, placebo-controlled, with 3000+ adults over 72 weeks) from Eli Lilly's oral GLP-1 demonstrated "near equivalent to injectable" efficacy for weight loss. The study, published in the New England Journal of Medicine this week, showed that "orforglipron is an effective oral alternative for obesity treatment, potentially addressing barriers to GLP-1 therapy such as injection requirements, storage limitations, and accessibility. While lower than injectable options, the 11.2% weight loss with the 36mg dose crosses the clinically meaningful 10% threshold associated with cardiometabolic benefits."
Article
https://www.nejm.org/doi/pdf/10.1056/NEJMoa2511774
AI summary of the study
https://claude.ai/public/artifacts/7e7e8981-5126-49a0-b88d-55866cc54884
X discussion
https://x.com/cremieuxrecueil/status/1968096846548812105?s=42
Anti-Science Articles of Note
Several science bloggers, all physicians and epidemiologists, captured last week's Kafka-esque meetings of the CDC's ACIP (Advisory Committee on Immunization Practices). I offer a Perplexity-summarized analysis of 4 blog posts that recount the details of the newly empaneled CDC advisors' discussions, many of whom have neither the credentials nor the experience to set national vaccine policy. "[And yet,] some key decisions—especially related to Covid-19, Hepatitis B, and MMRV vaccines—were made, albeit uncertain about the practical impact for patients and clinicians."
AI Summary
https://www.perplexity.ai/search/summarize-across-all-the-tabs-L.m9VpdbTdGtAWxvgQiRWQ#0
The blog posts I used as the basis of the summary:
https://theunbiasedscipod.substack.com/p/day-1-of-the-acip-meeting
https://theunbiasedscipod.substack.com/p/acip-day-2-when-i-became-all-seven
https://yourlocalepidemiologist.substack.com/p/covid-19-vaccine-changes-what-it
https://insidemedicine.substack.com/p/im-still-confused-member-says-during
Living with AI.
I have seen several discussions about the need to detect and mitigate "scheming" behavior in AI models—where models pursue hidden, misaligned goals while concealing their intentions. One paper is from OpenAI, where the authors stress the need for deliberative alignment, a training method where models explicitly reason about safety specifications, as an anti-scheming intervention. A second is Sam Harris's podcast Making Sense, which introduced me to AI-Neoluddites, Eliezer Yudkowsky and Nate Soares, the authors of 'If Anyone Builds It, Everyone Dies: The Case Against Superintelligent AI.' (A light uplifting title, for sure.)
https://mlq.ai/news/openai-and-apollo-research-unveil-methods-to-detect-and-reduce-ai-scheming/
https://chatgpt.com/s/t_68cf0ab8933c8191bf03ba1c26232551
Sam Harris Podcast
https://open.spotify.com/episode/3nJvUtidCziB2tibaGlKMM?si=xt5WPIZ4R7uVPA5yiuov2A
Bonus material from a loyal reader on thinking about and using big data - an interview with Alex Karp of Palantir:
https://open.spotify.com/episode/3c8dUIKC3kb91xrkWy6ocu?si=dL_f4HZUSFmlSklKE2kdFg
OpenAI sponsored research (published this week) offering "a first-of-its-kind view into how [ChatGPT] creates economic value through both increased productivity at work and personal benefit. The [data] show that consumer adoption has broadened beyond early-user groups, shrinking the gender gap in particular; that most conversations focus on everyday tasks like seeking information and practical guidance; and that usage continues to evolve in ways that create economic value through personal and professional use." (Per the previous articles, is this just the AI warming us up to take over - are we being manipulated into trusting the LLMs?)
Professor Ethan Mollick covered this paper in detail:
https://x.com/emollick/status/1967800804301283452
The OpenAI website on the paper:
https://openai.com/index/how-people-are-using-chatgpt/
An AI summary of the 60+ page paper published by the National Bureau of Economic Research, a non-profit, non-partisan Think Tank.
https://chatgpt.com/s/t_68d01ea6a9cc8191b4536498addd6995
Infographics
Sometimes, infographics communicate facts. Sometimes, infographics succinctly convey overwhelming complexity. This image captures the latter of the two:
https://www.thehighestofthemountains.com/images/thehghestofthemountains_brain_map_135a-185px.jpg
I suggest perusing the author's website. He is (self-reportedly) a space shuttle guidance system engineer who "applies his engineering and electronics background to reverse-engineer the human brain." Moreover, he is deeply religious, using complex engineering to "explore" faith, intelligent design, and refute evolution. It is a fascinating mélange of topics. Irrespective of his mix of ideas, he makes some fantastic infographics.
Things I learned this week
I didn't know I had an opinion on human-AI "relationships." This week, I learned I do, thanks to this editorial debate between writers Kat Rosenfield and Kate Lindsay on whether engaging in "intimate acts" with AI constitutes cheating on your non-AI spouse (or your other AI spouse?)
https://post.substack.com/p/is-it-cheating-if-your-illicit-affair
In case you were wondering, it is not legal to hold a Viking funeral (like burning a longboat on the Chesapeake Bay) in Maryland (and almost any other state)—the many unanticipated barriers center around environmental pollution and fire risk, not the ceremony itself.
https://thefisherlawoffice.wordpress.com/2013/02/09/estate-planning-for-a-viking-funeral/
Orcas are again attacking boats off the coast of Europe. (See the first time I wrote about this phenomenon: https://www.whatadamisreading.com/2024/08/what-adam-is-reading-week-of-8-5-24.html). "According to experts, orcas typically attack boats through playfulness, boredom, or a learned 'fad' among a specific subpopulation, with individual orcas like White Gladis potentially starting the behavior after a traumatic event like a collision with a boat or nets." Given the range of explanations, experts seem uncertain about the Orcas' intentions, although the idea of Orca peer pressure is amusing.
https://www.mirror.co.uk/news/world-news/footage-shows-terrifying-moment-yacht-35939383
It seems that Apple's new AirPods translation tools (if they could include sperm whale and Orcas) could help ascertain their intentions. (I wonder if they are anti-capitalist whales raging against those who own yachts?)
From 2024: "AI reveals Sperm Whale Alphabet" https://www.bbc.com/future/article/20240709-the-sperm-whale-phonetic-alphabet-revealed-by-ai
AI art of the week
A visual mashup of topics from the newsletter.
I use ChatGPT to summarize the newsletter, suggest prompts, and make the images.
"Banksy-style urban mural on a concrete wall: a group of orcas perched on the deck of a sleek yacht. Each orca wears oversized headphones. One whale speaks into a glowing translator device that sends out neon streams of code, medical charts, and planetary orbits. At the bow of the ship, a DJ orca scratches on a turntable shaped like a medical examination table. Spray-painted AI speech bubbles and floating journal pages hover overhead. Behind them, enormous asteroid diagrams and tangled infographics spill across the wall in chaotic spray-painted layers, dripping with vibrant neon color. The whole mural has the layered, rebellious feel of street art, with overlapping stencils, tags, and abstract sonar wave patterns. The style should emphasize stark black and white stencils, strong silhouettes, and a minimalist yet impactful composition, characteristic of Banksy's street art."
ChatGPT
The LLM would not generate an image based on the above prompt due to "content restrictions."
Gemini:
https://drive.google.com/file/d/1qAkos-gERtQMVlP79d2BVhLiQUZsCXYb/view
Grok:
https://drive.google.com/file/d/1vWHrlwRij051GoHlZ9u6Ry0Z15URsWcL/view
---
COVID rates are still rising. I continue to recommend the updated mRNA (Pfizer and Moderna) and protein-based (Novavax) to my patients and family. Details from this blog may help clarify the craziness:
https://yourlocalepidemiologist.substack.com/p/covid-19-vaccine-changes-what-it
The Pandemic Mitigation Collaborative (PMC) utilizes wastewater viral RNA levels to forecast four-week predictions of COVID-19 rates.
https://pmc19.com/data/
based upon https://biobot.io/data/
Wastewater Scan offers a multi-organism wastewater dashboard with an excellent visual display of individual treatment plant-level data.
https://data.wastewaterscan.org/
----
Clean hands and sharp minds,
Adam
Relive all the thrills and excitement - The What Adam is Reading Archive
http://www.whatadamisreading.com/
CONFIDENTIALITY NOTICE: THIS MESSAGE IS CONFIDENTIAL, INTENDED FOR THE NAMED RECIPIENT(S) AND MAY CONTAIN INFORMATION THAT IS (I) PROPRIETARY TO THE SENDER, AND/OR, (II) PRIVILEGED, CONFIDENTIAL, AND/OR OTHERWISE EXEMPT FROM DISCLOSURE UNDER APPLICABLE STATE AND FEDERAL LAW, INCLUDING, BUT NOT LIMITED TO, PRIVACY STANDARDS IMPOSED PURSUANT TO THE FEDERAL HEALTH INSURANCE PORTABILITY AND ACCOUNTABILITY ACT OF 1996 ("HIPAA"). IF YOU ARE NOT THE INTENDED RECIPIENT, OR THE EMPLOYEE OR AGENT RESPONSIBLE FOR DELIVERING THE MESSAGE TO THE INTENDED RECIPIENT, YOU ARE HEREBY NOTIFIED THAT ANY DISSEMINATION, DISTRIBUTION OR COPYING OF THIS COMMUNICATION IS STRICTLY PROHIBITED. IF YOU HAVE RECEIVED THIS TRANSMISSION IN ERROR, PLEASE (I) NOTIFY US IMMEDIATELY BY REPLY E-MAIL OR BY TELEPHONE AT (855.472.9822), (II) REMOVE IT FROM YOUR SYSTEM, AND (III) DESTROY THE ORIGINAL TRANSMISSION AND ITS ATTACHMENTS WITHOUT READING OR SAVING THEM. THANK YOU.
-DaVita Inc-
Comments
Post a Comment