What Adam is Reading

Week of January 30, 2023

Mental real estate is precious and artificial intelligence has rapidly moved into mine for the last few weeks. While coronavirus occupies less space in my reading (and writing), it is still a concern – causing avoidable illness, disability, and death and bringing not-yet-appreciated changes to the world. I wonder if A.I. is another "flavor of the week" topic or a sustained, powerful driver of societal change. Either way, it is, quite literally, what I am reading.

-----

Fourteen-day trends in COVID cases, hospitalizations, ICU admissions, and deaths are falling for a second week.

N.Y. Times Tracker

https://www.nytimes.com/interactive/2021/us/covid-cases.html

Virologist Marc Johnson's comments are typical of the reactions I've seen this week about the current XBB.1.5 spike (or lack thereof).

https://twitter.com/solidevidence/status/1619729627257401345

------

There were several COVID articles of interest in the last week:

A review and meta-analysis of 17 studies, including 11 million vaccinated and 2.5 million unvaccinated children aged 5 to 11 years, found the vaccine protects from infection (early on), severe illness, hospitalizations, and long COVID symptoms. Reactions to the vaccine were minor and self-limited. These data are very reassuring.

https://jamanetwork.com/journals/jamapediatrics/fullarticle/2800743

Conversely, being unvaccinated puts individuals at higher risk of numerous adverse long-term outcomes:

"Amongst this study of 1832 US adults, the risk of reporting symptoms for 28 or more days after COVID-19 onset was significantly higher in participants who were unvaccinated at the time of infection and those who reported moderate or severe acute illness symptoms. At six months after onset, [unvaccinated] participants had a significantly higher risk of pulmonary, diabetes, neurological, and mental health encounters vs pre-infection baseline."

https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2800554

The Chief Medical Officer at NOLA Children's Hospital posted both articles on his Twitter feed. Compare the comments he received.

Data that unvaccinated individuals have worse outcomes article:

https://twitter.com/MarkWKlineMD1/status/1618651408185573388

Data that vaccines are safe and effective for 5-11-year-olds article:

https://twitter.com/MarkWKlineMD1/status/1618026754177630208

A group of Canadian epidemiologists explored pre-print (not-yet-peer-reviewed) articles. The concern has been the trade-off between rapid access to data (pre-print) vs. quality of the data and conclusions before peer review. They found that pre-print article data is available faster and that studies with larger sample sizes were more likely to get accepted in a peer-reviewed medical journal. Moreover, although the peer-reviewed published articles differed from their pre-print status in some ways, "the main conclusion remained consistent for most studies." It is a great reminder to be thoughtful and appropriately contextualize non-peer-reviewed data.

https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2800857

Medical Trends and Technology

ChatGPT was made public on 11/30/22. I am fascinated by the speed with which it has become "a thing."

Here is a Seeking Alpha article on the use of A.I. in healthcare. From decision support to pharma research, companies have employed A.I. tools for some time. https://seekingalpha.com/news/3929184-the-growing-use-of-artificial-intelligence-in-medical-devices-diagnostics-drug-development

Check out the list of FDA-approved AI/ML medical devices.

https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-aiml-enabled-medical-devices

ChatGPT passed the U.S. Medical Licensing Exam (USMLE) (3 parts typically taken during and right after medical school).

Pre-print article

https://www.medrxiv.org/content/10.1101/2022.12.19.22283643v2.full

British tabloid hysteria

https://www.dailymail.co.uk/news/article-11666429/ChatGPT-pass-United-States-Medical-Licensing-Exam-Bar-Exam.html

FYI - ChatGPT also passed the Bar Exam and various business school exams:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4314839

and

https://mackinstitute.wharton.upenn.edu/wp-content/uploads/2023/01/Christian-Terwiesch-Chat-GTP.pdf

My kids' high school is now implementing policies and procedures to help minimize using ChatGPT in writing assignments, including using A.I. writing detectors. A Wharton professor has taken a different tactic. https://www.npr.org/2023/01/26/1151499213/chatgpt-ai-education-cheating-classroom-wharton-school

Here is a review of GPTZero - one of the detectors scoring written materials in terms of text randomness (more = human) and the distribution of that randomness.

https://goldpenguin.org/blog/gptzero-review/

In its current [incarnation? Version? Instance?] ChatGPT is "falsifying" references. (For example, when asking ChatGPT to write me a paper and cite five sources, it fabricates those sources). Chess Grandmaster and economist David Smerdon explains why this results from a probabilistic algorithm, not programmed evil. Either way, best to understand your tools.

https://twitter.com/dsmerdon/status/1618816703923912704

And in a vision of one possible dystopian future, our A.I. will have numerous artificial bodies to inhabit and interact with real space (beyond Boston Dynamics). Remember, there is no science fiction - just science fact and science not yet invented.

Engineers have developed hybrid bio-electronics with living muscle cells attached to mechanical and electronic devices. Watch the video.

https://www.theregister.com/2023/01/21/mouse_muscle_cyborg_light/

Or the A.I. can be implanted in phase-shifting materials - metals that can turn from solid to liquid and back to solid again. Seriously. Like Terminator T-1000. Watch the video.

https://www.iflscience.com/shape-shifting-robot-escapes-miniature-prison-cell-by-liquifying-before-re-solidifying-67249

Infographics

I feel too old for zombie apocalypse movies (it feels too real), but I watched two episodes of The Last of Us with my younger son this weekend. Thanks to the last 15 years of books, movies, and T.V., we should be well prepared no matter the underlying pathogenesis of zombification(?). Here is a collection of 19 zombie infographics across multiple storylines. If only the characters had access to all this valuable information.

https://rockcontent.com/blog/zombie-apocalypse-19-infographics-to-help-you-survive/

I like the Venn diagrams and the last one - "The Science of Surviving the Zombie Apocalypse."

Things I learned this week

I scored 100% on the N.Y. Times "Can you interpret ape hand signals?" quiz (2/2!). I am (once again) not surprised that non-human creatures have complex communication. I am amazed at how easy it is to interpret ape hand gestures. (See previous emails about dogs using speech buttons and interpreting whale songs.)

https://www.nytimes.com/interactive/2023/01/24/science/ape-gestures-quiz.html

I learned the word for my typical 2 pm fatigue - acedia. The New Yorker published a surprisingly engaging book review of The Wandering Mind - a book about the strategy and tactics of monks aiming to stay focused. Even before social media and the internet, paying prolonged attention was hard.

https://www.newyorker.com/magazine/2023/01/30/what-monks-can-teach-us-about-paying-attention-wandering-mind-jamie-kreiner

More about acedia - which has complex connotations like umami is a complex flavor.

https://theconversation.com/acedia-the-lost-name-for-the-emotion-were-all-feeling-right-now-144058

Topics from the newsletter through the "eyes" of our A.I. Overlords!

(What is this section? - https://openai.com/dall-e-2/)

"robot zombie monkeys holding medical journals and using sign language to speak with dogs degas-style"

https://labs.openai.com/s/uOMvnrvgL0YsZt1RGUlxDcMl

Clean hands and sharp minds,

Adam