Week of May 12, 2025
Uber drivers like to talk to me. Some of my most memorable encounters include the conspiracy theorist in Vegas (the rich are building bunkers to ride out "the apocalypse"), the crime scene biohazard clean-up specialist (lesson shared: be very sure your replica hand grenade is truly a replica), and the driver with kidney disease who I counseled about home dialysis between the Orlando airport and a Disney property. Last week, my driver was a retired Southwest pilot in Dallas who spent the trip explaining how religion sustained him through decades of sobriety and gently pushing me to explain the role of Jesus in my life ("I am not sure how you see our Lord, but..."). Fortunately, we arrived at the Southwest departure area at Love Field before I disclosed I am Jewish and before I explored the prevalence of alcohol use amongst pilots (some things are better left unsaid). I often wonder why there is no ride-share mobile therapy.
---
Listen to a Google Notebook LM A.I.-generated podcast of the newsletter with two virtual "hosts."
https://drive.google.com/file/d/1IYIP83xS5emMT6CuFo8NqG351MJwfWQH/
------
Science and Technology Trends
"The first nonverbal Neuralink patient (a patient with ALS and the 3rd person to receive the implant) is offering a glimpse into how he uses the technology – editing and narrating a YouTube video using signals from his brain." The narration is his cloned AI voice. The video walks through how he trained the system to map "attempted" movements into using his computer and how he calibrates the system over time. His use of the Grok3 LLM to amplify the interface into complex communication is fascinating.
https://www.sciencealert.com/world-first-neuralink-patient-makes-youtube-video-with-brain-implant
and
https://www.youtube.com/watch?v=CJn0WRKwg34
In last week's New England Journal of Medicine AI, Penn and the University of Maryland, radiologists published a checklist to guide medical authors and reviewers in reporting on the clinical uses of Large Language Models (like ChatGPT or Claude). The checklists allow for a rigorous approach to reviewing AI model elements such as specifications, data-handling practices, training procedures, evaluation metrics, and transparency standards to make them accessible and reproducible across clinical use cases. While the article itself is not exciting, it is an excellent example of why measuring the value and impact of clinical AI is so difficult. There are often few objective baseline comparators (what is a "good enough" medical record summarization?), and analysis requires a high degree of abstraction (I can't compare the outcome across different questions, but I can compare how reproducible different LLMs results are).
Article:
https://drive.google.com/file/d/1BuftZHLqMChwjznUpChRHAguJqiWp7mJ/
AI Summary and Analysis:
https://claude.ai/public/artifacts/cb4fc469-3d06-4eb3-94b7-87f977b36c23
Anti- Anti-Science Articles of Note
Dr. Kristen Panthagani offers an excellent analysis of the latest trend in logically fallacious anti-science - individuals who espouse that germ theory (yes, germ theory) is incorrect. This shift argues that "infections don't pose a risk to healthy people who have optimized their immune system. And if you want to prevent infections, vaccines aren't the solution; becoming healthier through nutrition, exercise, and dietary supplements are." This argument attempts to re-weight complex data - i.e., over-indexing on the importance of preexisting conditions and under-valuing the pathogenicity of infectious diseases (this notion is dangerously incorrect). This blog post highlights an intellectual layer cake of bad ideas- such as cherry-picking and false dichotomies combined with the typical fallacies of loss aversion and present bias (essentially over-valuing potential short-term side effects and low-probability outcomes while under-valuing the longer-term benefits of vaccination or other treatments).
https://yourlocalepidemiologist.substack.com/p/the-rise-of-a-new-form-of-germ-theory
TL;DR:
https://claude.ai/public/artifacts/aef8f0c8-f973-4fda-b972-3a511cd95e32
Living with AI.
Dwarkesh Patel offers a glimpse into the future of large corporations with only a handful of employees overseeing millions of AI agents. How will companies operate once fully automated with AGI (Artificial General Intelligence) leveraging the collective advantages of scaled digital entities?
https://www.dwarkesh.com/p/ai-firm
TL;DR: https://claude.ai/public/artifacts/d47542c2-8a72-4a5d-be37-c035c01a851e
A loyal reader shared this article from the Atlantic looking at the recent changes in ChatGPT that led it, for a brief period, to be overly sycophantic. The article explores the consequences of AI feeding us humans overly optimistic and validating answers.
https://www.theatlantic.com/technology/archive/2025/05/sycophantic-ai/682743/?gift=wRB5v0WyIdmVLiiSUSDcYe6KtnvM7R5mTXZP-afDEhI&utm_source=copy-link&utm_medium=social&utm_campaign=share
and
TL;DR: https://claude.ai/public/artifacts/1e0cb61d-b677-4075-94a9-c91ea1d5cc26
Related:
I learned the word "Infovore" and about the work of economist Tylor Cowen. Eric Topol interviewed him (and really seems to like him). Lovefest aside, Cowen's thoughts on AI and how to navigate a complex information landscape are worth the time to listen to.
https://erictopol.substack.com/p/tyler-cowen-the-prototypic-polymath
Infographics
Here is an interesting infographic highlighting the most common jobs by age cohort based on the 2024 US Bureau of Labor Statistics data. One wonders which of these jobs AI will impact.
https://www.qualtrics.com/m/assets/blog/wp-content/uploads/2024/10/common-occupations-by-age-3_reduced-97-scaled.jpg
from
https://www.qualtrics.com/blog/most-common-occupations-by-age-group/
Things I learned this week
Here is my favorite headline of the week: "'Unparalleled' snake antivenom made from man bitten 200 times." "Mr. Friede has endured more than 200 bites and more than 700 injections of venom he prepared from some of the world's deadliest snakes, including multiple species of mambas, cobras, taipans, and kraits." "It just became a lifestyle and I just kept pushing and pushing and pushing as hard as I could push - for the people who are 8,000 miles away from me who die from snakebit." I had never considered purposefully exposing oneself to snake venom a "lifestyle," but to each his own.
https://www.bbc.com/news/articles/cr5d0l7el36o
And, of course, there is a social media component to Mr. Friede's goodwill toward all snake-bitten men:
https://www.youtube.com/user/timfriede578/videos
Shampooty is a company that sells adult versions of kids' toys (for instance, a herse-version of the LittleTikes red and yellow plastic ride-on car). It is simultaneously disturbing and entertaining. While you can buy the products, the entire thing is art — the "Kids Toys, Adult Issues" project by artist Andy Sahlstrom. When I retire, I aspire to be like Andy, a self-described "Interactive Technology Engineer, Inventor, and Kinetic Artist."
https://shampoooty.com/collections/collectibles
and
https://andysahlstrom.com/design/kids-toys-adult-issues
AI art of the week
(A visual mashup of topics from the newsletter, using ChatGPT to summarize the newsletter, suggest prompts, and make the images).
https://drive.google.com/file/d/1FRP8TcT8D4-r0oKAs3UKE33SVw7QdCFe/
"A Byzantine-style tempera painting on a gold leaf background, divided into three distinct iconographic panels. On the left, two bearded men sit in a modern vehicle, depicted with stylized flatness and strong outlines, as one gestures passionately while the other listens—resembling a sacred conversation. In the center, a youthful figure in red and blue robes holds a replica grenade solemnly, framed by a decorative arch with geometric and floral motifs, and a skull in the background. On the right, a wise elder in similar robes administers a strange healing ritual to a boy seated at a table, using an ornate glass vessel and tubing. The entire piece is framed with intricate decorative borders and uses symbolic gold throughout, evoking medieval religious iconography with modern narrative irony."
(Who amongst us does not love evoking medieval religious iconography with modern narrative irony?)
---
There was a slight increase in COVID infections to 1 in 170 (from 1 in 196) last week. Looking at the wastewater data, Norovirus and COVID are the only two infections in circulation in recent weeks.
The Pandemic Mitigation Collaborative (PMC) website uses wastewater levels to forecast 4-week predictions of COVID 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
Comments
Post a Comment