Tuesday, May 19, 2026

Thoughts on Wachter's Giant Leap: Great points about AI's early missteps vs current successes in medicine

I enjoyed reading Robert Wachter’s book A Giant Leap very much. The book chronicles the early missteps in using AI in medicine, particularly in disease diagnosis. It also discusses some of the more recent successful developments, such as assistance with billing.

The book explores the future of AI in medicine, blending in-depth character portraits of many leaders in the field with intelligent common sense and a systematic overview of its many facets. One of the things I found most notable was the contrast between the relatively limited progress of AI in medicine and its remarkable advances in other fields, such as self-driving cars. One of the key differences Wachter points out is the availability of massive amounts of training data for self-driving cars and the difficulty of obtaining comparable data in medicine, particularly given the siloed nature of medical information and persistent privacy concerns. 

Wachter also points out that one of the key problems with AI in medicine is that early efforts focused on very difficult problems, such as diagnosis, where the stakes are high, rather than on simpler tasks like transcribing physicians’ notes or assisting with billing, where the stakes are lower and incremental innovation is more feasible. A notable success I particularly got a kick out of was the rapid uptake of the AI scribe for documenting clinical encounters. One reason for this popularity is that the computerization of medicine has created an enormous task of data entry, which Wachter describes as  "documentation burden" for physicians; AI scribes help alleviate this. The book also examines the economic forces shaping AI in medicine. One of the most notable for me was the near-monopoly that Epic holds over electronic health records and the potential lock-in effect this creates for applications built on its platform. 

The book further explores how AI will impact medicine beyond physicians, including its effects on education and the patient experience. In relation to education, I found it fascinating how Wachter describes AI’s influence at New York University, where it is used in reviewing medical school applications. Given that many applications are now partially written by AI, this creates what seems like a strange arms race. 

Another anecdote I found particularly amusing involves Wachter’s discussion of safety in AI-assisted systems in highly regulated fields. While we may worry about the safety of AI in clinical settings, the risks associated with human drivers can be even more immediate. He illustrates this with a hilarious story about an Uber driver who insisted on watching a movie throughout the trip. Is this safe?

Overall, I found this to be a very entertaining read, and I will keep it in mind when thinking about the future of AI in medicine.

A Giant Leap: How AI Is Transforming Healthcare and What That Means for Our Future by Robert Wachter

Tuesday, December 23, 2025

Thoughts on Ellenberg's Shape: Engaging Stories about Hard Stuff, esp. liked the Discussion of Math behind AI

I liked Jordan Ellenberg’s book Shape very much. It gives an easy-to-read understanding of many key (but difficult to grasp) mathematical concepts. For instance, it includes a deep discussion of topology that I found revealing.

The parts I enjoyed most were those related to current AI approaches. Ellenberg describes how one can understand a simple perceptron, a basic neural network, through a familiar decision-making process -- the Electoral College, where electors in each state come together to make a binary decision and the weighted outcomes combine into a national judgment. I wish he had gone further in this discussion, perhaps by exploring how one might compute a loss function in this context and compare it to the popular vote.

I also appreciated his treatment of gerrymandering. He explains the geometric aspects of constructing districts to maximize a political party’s advantage in statewide elections, and he highlights how difficult it is, in geometric terms, to define what a fair or well-shaped district should be. His discussion of simulating districts at random and examining how unlikely a particular enacted map is relative to that baseline was helpful, although I would have preferred a clearer framing in terms of a null process and P-values.

The book also provides a clear description of epidemics, R0, and geometric series. I liked the way it covered historical figures in epidemic research, especially Ronald Ross, the British epidemiologist. In relation to machine learning, I found the explanation of gradient descent, including stochastic gradient descent, intuitive and useful.

Overall, I felt Shape was a good read, offering straightforward explanations of mathematical ideas that can be hard to grasp.

Shape: The Hidden Geometry of Information, Biology, Strategy, Democracy, and Everything Else by Jordan Ellenberg

Saturday, August 02, 2025

Thoughts on Pearl's Book of Why: Lots of Good Intuition on Bayesian Networks, Sometimes with Strong Opinions

The Book of Why by Judea Pearl with great interest. The author is undoubtedly one of the most brilliant thinkers in recent years on causative reasoning and artificial intelligence. The book provides a detailed, somewhat technical, and historical account of causation. I particularly appreciated how it delved into key concepts such as Bayesian networks and confounders, topics that Pearl has explored extensively. 

I also enjoyed his discussion of smoking, the Bradford Hill criteria, and how we needed to move beyond mere correlation and association to make informed decisions. Pearl’s ability to place these discussions in a historical context, highlighting the contributions of figures like Fisher and Sewall Wright, was impressive. I particularly appreciated how Pearl explained the three key graphical elements in Bayesian networks, which he referred to as chains (A leads to B, which leads to C), colliders (A and C both point to B), and forks (B leads to both A and C). He explained how these motifs differ in terms of the independence of the three variables and the extent to which they remain independent when conditioned on B. He provided intuitive examples to help readers easily grasp these concepts and emphasized the fundamental problem of confounding in statistical design, as well as how randomized controlled trials address this issue.  

 However, at times, I found it challenging to distinguish between background information on established concepts and Pearl’s more polemical arguments advocating for what he calls the “causal revolution,” which may be less widely accepted in the statistical and AI communities. This occasionally led to some confusion, but overall, I enjoyed the book immensely and would highly recommend it to any reader.

Sunday, April 20, 2025

Thoughts on Zimmer's Airborne: Great book about an overlooked mode of disease transmission, now prominent because of the pandemic

Air-borne by Carl Zimmer is a well-researched and quite enjoyable book that combines the science of airborne disease transmission, the history of key researchers and their contributions, and the political debates over biodefense and the public health response to COVID-19. It provides a detailed account of airborne disease transmission research, from early pioneers to modern scientists who challenged long-standing misconceptions. The book also critiques how public health institutions, particularly the CDC and WHO, mishandled airborne transmission during the COVID-19 pandemic. Zimmer presents a strong argument that scientific knowledge on airborne transmission was ignored or resisted, leading to missteps in public health guidance. All of these elements come together to make for a compelling and insightful read.

The book begins with the original pioneers of germ theory, focusing on Louis Pasteur and Robert Koch in establishing that microorganisms cause disease. Zimmer walks the reader through some of their most famous experiments, which became a foundation for proving microbial causation of disease. Along the way, Zimmer introduces early debates on germ theory, including Pettigrew’s opposition, where he argued that diseases were caused by gaseous emissions from microbes rather than the microbes themselves. Though ultimately disproven, Pettigrew made lasting contributions, pioneering carbon dioxide measurement and establishing the Pettigrew limit (1,000 ppm of the gas), which is still used today as an indoor air standard. Zimmer also discusses the early development of masks, first for epidemic control in China (~1910) and later during the 1918 flu pandemic, even though their effectiveness was not well understood at the time. While Zimmer provides a great history of these key figures and their contributions, he also takes the reader through the evolving science of airborne disease transmission.

The 1930s and 1940s were a key period for airborne transmission research, though much of it was later forgotten and rediscovered. Among the early pioneers were William and Mildred Wells, who conducted studies at the Germantown School and Loch Raven Hospital, including early work on UV light disinfection. Richard Riley expanded on their research, developing mathematical models for airborne disease spread. Harold Egerton at MIT, who produced early sneeze photographs, provided further evidence of airborne transmission. Another key figure in airborne research was Peter Tsai, whose ideas on electrostatic filtration led to the development of the N95 respirator by 3M. Zimmer highlights how these early discoveries were largely ignored for decades, delaying progress in recognizing airborne transmission.

Despite these advancements, airborne transmission remained controversial. Charles V. Chapin, a major public health authority, argued that most diseases spread via large droplets rather than aerosols. This dogma became embedded in medical training and public health policy, despite evidence suggesting otherwise. The book details early debates on this topic that shaped how airborne transmission was understood and, in some cases, dismissed. Zimmer details early debates, including experiments in Toronto, where one scientist studied exhalation and sneezing at different heights. The findings were interpreted as supporting the large droplet model, reinforcing skepticism toward airborne spread.

Zimmer also examines the intersection of public health and biodefense. The book describes the history of biowarfare research in the U.S. and the debate between Alexander Langmuir and Theodore Rosebury, which mirrored, to some degree, the nuclear weapons divide between Oppenheimer and Teller. The book discusses how the CDC was originally formed as a quasi-military entity, with a dual role in civilian health monitoring and biological weapons surveillance. Zimmer contrasts the US program with that of the Soviet Union, which experienced multiple accidents, including the Sverdlovsk anthrax release. Zimmer critiques how biodefense policies under Bush and Clinton focused more on bioweapons threats (e.g., Iraq) than pandemic preparedness. Bush’s 2005 Biodefense Strategy, following the 2001 anthrax attacks, aimed to prepare for biological threats, including stockpiling N95 respirators and building a manufacturing facility—a plan that never materialized. The stockpile was depleted during H1N1 (2009) and never replenished, leaving the U.S. unprepared for COVID-19. 

As the book builds toward the COVID-19 pandemic, Zimmer provides a historical overview of pandemics, including the 1918 flu, SARS, and MERS, emphasizing how each should have better informed the COVID-19 response. He details Event 201, a 2019 pandemic simulation that accurately predicted many of the failures that later occurred. Despite these warnings, when COVID-19 emerged, the WHO and CDC were slow to acknowledge airborne transmission, instead focusing on surface cleaning and six-foot distancing, which failed to fully address how the virus spread. Zimmer highlights how airborne transmission experts, including Lindsay Marr (V Tech), Bora Berry (MIT), and Donald Milton (University of Maryland) recognized early that COVID-19 was airborne, yet their findings were ignored. During this period, an informal group of 36 scientists petitioned the WHO to update its guidance, but their efforts were dismissed. 

Zimmer also critiques inconsistent messaging in the pandemic, citing remarks by Donald Trump and Joe Biden that initially acknowledged the airborne spread but were later walked back. A WHO official even admitted COVID-19 was airborne before correcting himself, contributing to public confusion and delayed mitigation efforts. Eventually, the WHO quietly revised its stance without addressing its initial resistance. Zimmer concludes by discussing overlooked solutions, such as ventilation improvements, the Corsi-Rosenthal box, and far UV technology, which offer promising but underutilized mitigation strategies.


Air-Borne: The Hidden History of the Life We Breathe Hardcover – February 25, 2025
by Carl Zimmer (Author)

Wednesday, February 12, 2025

Thoughts on Esposito's Breslin: A Great Story about a Historic Storyteller

Richard Esposito's book on Jimmy Breslin provides an engaging look at both the man and the turbulent times he lived through in New York City during the 1960s, 70s, and 80s. The book captures the city’s shifting landscape, from the strife of the 1960s to the violence of the 70s and 80s, such as the Son of Sam murders, the Crown Heights riots, and the vigilante actions of Bernhard Goetz. Esposito uses these events as a backdrop to Breslin’s world, painting a rich picture of New York during those decades. 

 Breslin is portrayed as a fearless journalist who went far beyond the typical duties of a reporter. He traveled to Vietnam to report on the war and ventured deep into the toughest neighborhoods of Brooklyn. He even found himself in the middle of a riot -- in this particular incident, his cab was overturned, he was stripped of his clothes, beaten, and forced to walk back to a police station. Breslin’s willingness to put himself in such dangerous situations highlights his dedication to telling "the story." 

One of the most compelling aspects of the book is how it shows Breslin as a political figure in his own right. His coverage of major political events was fascinating, particularly his account of Nixon’s impeachment. It highlights how Nixon’s downfall ultimately rested on Tip O’Neill’s considerable power and influence. Breslin’s political engagement wasn’t limited to Washington; he also addressed local matters of law and order, interjecting his personal sensibility to his reporting. Beyond covering politics, Breslin stepped into the arena himself, running for mayor alongside Norman Mailer in a symbolic campaign designed to make a point rather than win. 

 Esposito also explores Breslin's character, highlighting his sharp eye for detail, his ability to capture the heart of any story, and his gruff, unpolished demeanor. Breslin moved easily between different worlds, from working-class Queens to the wealth of Central Park West, reflecting the diversity of New York itself. He brought the same depth to his crime reporting, offering insight into both the events and the minds of those involved. 

Overall, Richard Esposito’s book is an engaging tribute to Jimmy Breslin. Through snippets of his writing and stories from his extraordinary career, the book reveals Breslin as a journalist who didn’t just report on events but became part of the story.

Amazon.com: Jimmy Breslin: The Man Who Told the Truth by Esposito, Richard

twitter: @REspositoNYC

Sunday, August 20, 2023

Some Random Thoughts on Scaling Laws (Moore's, Eroom's, Lockheed's, &c)

I enjoyed this recent news story in The Economist on Lockheed's Law (https://www.economist.com/finance-and-economics/2023/06/22/can-the-west-build-up-its-armed-forces-on-the-cheap). The article describes an almost inverse scaling behavior where defense and military orders are getting more and more inexpensive, mostly because the allocation of the military budget to manpower is progressively decreasing, and it is possible for manufacturers to make munitions more efficiently. Lockheed's Law is extremely important in terms of how the overall market opportunities (using the language of Kolter et al.) will change in the future. One can imagine that war will become less manpower-intensive and more capital-intensive because of this type of scaling. 

It is interesting to compare the scaling that we see in Lockheed's Law to other famous scaling relationships that have changed markets dramatically. The most famous is Moore's Law, which describes the exponential scaling of computer chips and computation in general over the last couple of decades. This scaling powered the computer and electronics industry. There are several related laws, such as Kryder's Law, which describes the exponential scaling in disk drive capacity. These laws made possible new market opportunities. The extreme case is the handheld or wristwatch computer and the fact that people now carry around computers that are much more powerful than the massive supercomputers of the past.

An associated law describes an opposite trend in the pharmaceutical industry: this "law," dubbed Eroom's Law (i.e., Moore's Law in reverse), states that drugs are becoming exponentially more expensive to put on the market. This is a great comparison to Lockheed's Law. The idea is that it is becoming much more expensive to make pharmaceuticals due to the ever-increasing requirements for research development and larger and more complicated clinical trials. A related scaling law describes the ability to sequence genomic material. This indicates a hyper-exponential increase that, at some points, is even faster than Moore's law. The scaling for DNA sequencing enables very large-scale biomedical research and diagnostic testing, but it hasn't made an impact on new drug discoveries because of the inverse scaling described in Eroom's law. Combined, Eroom's Law and the related law for DNA sequencing are reshaping the biomedical and pharmaceutical industry, making drugs comparatively more expensive but creating new opportunities for diagnostics, which are very cheap. This, in turn, is creating market opportunities for companies that are marketing only drug targets and associated research but don't have the capital to develop a drug themselves.

Altogether, different scaling laws illuminate large-scale trends that are dramatically impacting production as well as market opportunities. What was inconceivable 30 years ago is now trivial because of the exponential scaling in Moore's Law. Perhaps Lockheed's Law will lead to a type of warfare in the future that is so capital-intensive that it will be fundamentally different from the type of engagement we saw in the First and Second World Wars, and this, in turn, will reshape the market for defense products. 

References

Can the West build up its armed forces on the cheap?
https://www.economist.com/finance-and-economics/2023/06/22/can-the-west-build-up-its-armed-forces-on-the-cheap

Breaking Eroom’s Law
https://www.nature.com/articles/d41573-020-00059-3

Moore's Law
https://en.wikipedia.org/wiki/Moore%27s_law

DNA Sequencing Costs
https://www.genome.gov/about-genomics/fact-sheets/DNA-Sequencing-Costs-Data

Other Stuff Related to Scaling





Nice scaling graph for the size of transformer models

Saturday, June 25, 2022

Thoughts on 3 of Kandel's books on the Brain (The Disordered Mind, In Search of Memory & Reductionism in Art and Brain Science): Different Views of the Elephant

Here, I summarize my thoughts on three books by Eric Kandel: The Disordered Mind, In Search of Memory, and Reductionism in Art and Brain Science: Bridging the Two Cultures. Eric Kandel is a great American scientist and winner of the Nobel Prize. Interestingly, he started his career as a humanities major at Harvard, and he writes very much in that tradition. 

    The books cover various topics, including psychiatric disorders such as schizophrenia, the molecular basis of memory, and the relationship between the subconscious and art, and incorporate his recollections of his life journey and world history, particularly related to Vienna. While the books focus on different things, they all look at different aspects of the same subject, like taking different views of one elephant.

    When discussing psychiatric diseases, I like how Dr. Kandel described the root causes and history of schizophrenia and autism. These diseases trace much of their early history to Vienna and some famous early brain scientists there, such as Kraepelin and Asperger. Within the topic of memory, I liked Dr. Kandel's reflection on how memory is stored in synapses from inhibitory and excitatory neurons, and, in parallel fashion, these synaptic memories turn into molecular events and gene expression through activating and repressing transcription factors of the CREB family. Kandel also talks about his own memories. It was striking how Vienna was such a center of scholarship in the early 20th century and so quickly fell into tragedy with the advent of the Nazis and has changed dramatically since then.

    Finally, I enjoyed reading about Dr. Kandel's relationship between the subconscious and art. He talked about how many recent artists have tried to move beyond the conscious representation of the figure and harness their subconscious and how abstract art can play into our deep mental processes, such as face recognition.

    Overall, I found these books very interesting reads that give an encompassing picture of both the mind and a great person.



The Disordered Mind
In Search of Memory
Reductionism in Art and Brain Science: Bridging the Two Cultures
by Eric Kandel

Tagged