Week of November 18, 2024
The future I imagined when I was five included robots, the "pew pew" of lasers, and space travel. "1980s Adam" did not imagine the trade-offs in moving from dial-up modems to dopamine-fueled doomscrolling, supervising my self-driving car, and enjoyably conversing with the voice-mode AI. And while I like my car, I am troubled that "a guy who acts like a Bond villain" owns the company (as one loyal reader puts it). Likewise, I am vaguely uncomfortable that ChatGPT now "remembers" my name and interests and that my wife likes peonies. Fortunately, an AI-generated dialogue between deceased futurists helps reconcile my concerns. I offer an imagined dialog between Alvin Toffler [author of the 1970 book Future Shock], Aldous Huxley, and George Orwell regarding their views on AI and self-driving cars.
https://claude.site/artifacts/f3995511-c3ab-4b49-9ec6-a749658de144
A little background on Toffler:
https://en.wikipedia.org/wiki/Future_Shock
and
https://www.npr.org/2016/06/30/484215904/encore-future-shock-40-years-later
If you haven't read Brave New World or 1984 lately, you should.
---
Listen to a Google Notebook LM A.I.-generated podcast of the newsletter with two virtual "hosts" (who did a great job handling the more "delicate" topics in this week's newsletter).
The algorithm continues to evolve. This week, despite reviewing all the articles by minute 11, the hosts fill another 6 minutes with random commentary, including a liberal arts-style analysis of the DALL-E art. For the moment, they still seem to like me.
https://drive.google.com/file/d/1K-T0Q8abNB-Bzwt4iyez_pId_r6fzxpb/vie
About NotebookLM: https://blog.google/technology/ai/notebooklm-audio-overviews/
------
Science and Technology Trends
Science highlighted the evolving role of CRISPR-based CAR-T therapy whereby a patient's cells are extracted, engineered, and re-introduced into the patient. These therapies can now treat patients with autoimmune diseases (like lupus). Though CAR- is costly and complex, the role of engineered immune cells is evolving quickly.
https://www.science.org/content/article/breakthrough-cancer-immunotherapy-now-taking-aim-autoimmune-disease
Anti-Anti-Science Articles
Steven Novella, a neurologist from New Hampshire (at Yale Medical School), writes editorials and deep dives into specific topics related to evidence-based science. His commentary this week is a well-written overview of why monitoring and disciplining physicians who spread medical misinformation is so tricky. He writes, "Licensed professionals are generally given a lot of leeway in exercising their professional judgment, and we certainly don't want to chill discussion and debate about what the evidence says and best practices. However, a licensed professional carries and benefits from the imprimatur of authority and legitimacy conferred by the state government. In other words, they can say (in some form) – listen to me and trust me because I have MD after my name, and I am a certified specialist." Put another way, the public's most significant protection from ill-informed or mendacious clinicians is medical malpractice claims.
https://sciencebasedmedicine.org/physician-misinformation/
A Secretary of Health and Human Services who lives in a bubble of science denialism and logical fallacies (like cherry-picking) will amplify the above problem.
https://www.theatlantic.com/health/archive/2024/11/rfk-jr-hhs-sanewashing/680663/
and
https://insidemedicine.substack.com/p/rfk-jr-to-be-trumps-nominee-for-
and
https://www.logicallyfallacious.com/logicalfallacies/Cherry-Picking
Living with AI
Stanford researchers published data suggesting that using LLMs does not improve physicians' diagnostic accuracy. In 2023, various physicians (family medicine, internal medicine, or emergency medicine, across different career stages) evaluated six clinical vignettes over 60 minutes, randomly assigned to use either an LLM or conventional resources. "The availability of an LLM to physicians as a diagnostic aid did not significantly improve clinical reasoning compared with conventional resources. The LLM also demonstrated higher performance than both physician groups, indicating the need for technology and workforce development to realize the potential of physician-artificial intelligence collaboration in clinical practice." Various studies have highlighted that AI alone performs diagnostic reasoning better than humans with or without AI support. Bear in mind that these were theoretical cases. The most expensive areas of healthcare are often not about the diagnosis but rather the ongoing coaching of managing chronic disease.
https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2825395
Last week, I shared some output from Anthropic’ s Claude summarizing trials of over-the-counter supplements used to treat abnormal cholesterol levels (high LDL, low HDL, etc.). No readers noticed (or at least commented) that the Claude-generated summaries included incorrect links (however, the articles are real, and the summarized data was accurate). Either way, asking AI to summarize medical literature is promising and concerning. A Nature article published this week precisely highlights the kinds of observations my Claude summary exercise illustrates - "Artificial intelligence could help speedily summarize research, but it comes with risks."
https://www.nature.com/articles/d41586-024-03676-9
Open AI will release computer-controlling agents in 2025. Prepare to use the word "agentic" more frequently.
https://www.bloomberg.com/news/articles/2024-11-13/openai-nears-launch-of-ai-agents-to-automate-tasks-for-users
and
https://www.aitoolreport.com/articles/openais-secret-ai-agent-revealed
Infographics
I offer "Battling Infectious Diseases in the 20th Century: The Impact of Vaccines" from a 2015 Wall Street Journal infographic.
https://graphics.wsj.com/infectious-diseases-and-vaccines/
Bonus infographic (when you like to dislike something). Here is a visualization comparing state obesity rates with the number of gyms per 100K inhabitants. It is a compelling graphic, but there are so many confounders with both weight and gym location (income levels, population distribution relative to gyms, the size of the gyms, and other factors related to obesity, etc.) that the implied correlation is fallacious. I included it because it exemplifies how appealing visuals can, sometimes, outstrip actual value. (I do note, however, that Hawaii and Colorado have the lowest obesity rates among all states at 25%).
https://www.reddit.com/r/dataisbeautiful/comments/1grxnz8/oc_gyms_per_100k_inhabitants_vs_obesity_rates_in/#lightbox
Things I learned this week
It is probably too medical, but I learned about a 2008 outbreak of a neurologic autoimmune disease induced in pork processing plant workers exposed to aerosolized pig brains. The NY Times article is an excellent read describing how epidemiologists need time to dig into problems and how they think through clusters of patients becoming ill.
https://www.nytimes.com/2008/02/05/health/05pork.html?unlocked_article_code=1.aU4.jTB_.ejo1jCi5dukd&smid=url-share
Great White Sharks are less likely to attack seal decoys (and presumably surfboarders) with attached LED light strips. "Using seal decoys fitted with LED lights and towed behind a boat, we explored the efficiency of different light configurations on the deterrence effect, showing that visual shape and motion cues are critical for prey recognition by Great White sharks. Counter illumination that is brighter than the background is most effective in deterring sharks, implying that, in this context, counterillumination works through disruptive camouflage rather than background matching. Our results reveal the importance of a dark silhouette against a lighter background [as a driver for] predatory behavior in Great White sharks and that altering the silhouette may form the basis of new non-invasive shark deterrent technology to protect human life." The research demonstrated that horizontal light configurations were the least likely to engender a shark attack (compared to random, vertical, or no LED light placements.) Two thoughts: 1) I recommend looking at unintentionally amusing illustrations and videos from the journal article. 2) If they extend this research to a real-world trial with surfboarders, do not get randomized to the vertical LED strip arm.
Article
https://www.cell.com/current-biology/abstract/S0960-9822(24)01431-3
Photo
https://cell.com/cms/10.1016/j.cub.2024.10.042/asset/0b77c6d9-908d-4700-a020-4eca77b49eaf/main.assets/fx1_lrg.jpg
Only a handful of English words can serve as nouns, adjectives, verbs, and adverbs. Only one of them is as provocative and versatile as "fuck." (Read that last sentence twice - see what I mean? For complete clarity, I am using the noun, not an adjective.) "Fuck has an enormous range of uses across many parts of speech, as this dictionary details: sexual and nonsexual, positive and negative, literal and figurative, funny and violent. For any situation, there's probably some sense, some expression or catchphrase, some proverb, some intonation that can be brought to the table." Literary Hub published a fantastic article on the history, usage, and cultural significance of "The Most Famous Swear Word."
https://lithub.com/a-brief-history-of-the-most-famous-swear-word-in-the-world/
Claude generated a list of less-charged English words used as nouns, adjectives, verbs, and adverbs:
https://claude.site/artifacts/44398fe4-2957-4204-9145-093047b8fd43
South Carolina has an animal containment problem - with now escaped research monkeys AND emus. However, if you name your emus Thelma and Louise, it seems you are tempting fate.
https://www.cbsnews.com/news/south-carolina-emus-missing-horry-county/
AI art of the week
(A visual mashup of topics from the newsletter, now using ChatGPT to summarize the newsletter, suggest prompts, and make the images).
"A scene of chaos in South Carolina with escaped research monkeys climbing trees and two emus, named Thelma and Louise, running down a rural road, evoking a cinematic chase vibe. The emus have surfboards with LEDs on their back. A shark is driving the truck."
https://drive.google.com/file/d/1yMfSrJqNonAoNFiF1uixGQEbU0arZBbX/view
--- COVID data
Possibly due to a warmer-than-expected fall, COVID cases are still declining (now down to 1 in 180 individuals infected), which decreases the likely number of cases through the holidays.
https://x.com/jpweiland/status/1857551356262826450
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