Book Reviews

On this page you will find reviews of books that I have written and submitted to various publications. As there are many reviews, if you would like a list of the book reviews I have written please contact me

12-1-25 Book Review If Anyone Builds It Everyone Dies Why Superhuman AI Would Kill Us All

Eliezer Yudkowski & Nate Soars

Little, Brown and Company Hachette Book Group 1290 Avenue of the Americas, New York, NY 10104

$30.00 259 pages ISBN 9780316595643 (Hardcover)

Frank Lock

Like "Life 3.0," by Max Tegmark, this is another book about artificial intelligence (AI) that raised my anxiety about the continuing development of AI. The title of the book certainly could make readers anxious. The authors are co-leaders of the Machine Intelligence Research Institute (MIRI) About the Machine Intelligence Research Institute. Sam Altman, the CEO of OpenAI, described Yudkowski as the person who was critical in the decision to start OpenAI. Yudkowski is a founding researcher in the field of AI alignment. Soares worked at Microsoft and Google, and has authored many articles on AI, including one on power seeking incentives in smarter than human AIs.

In the introduction the authors write, "Humans have an ability to steer the future using our intelligence. But that ability only works if we use it-if we do the things we have to do, when we need to do them." They indicate that they hope to inspire individuals and countries to rise to the occasion of the threat that AI represents. One important idea the authors advocate is that those involved in work on AI believe that AI minds are being created. The encyclopedia Britannica defines the mind as, "The intricate combination of cognitive abilities involved in perceiving, remembering, reasoning, and deciding. It encompasses sensations, perceptions, emotions, memory, desires, reasoning, motives, choices, personality traits, and the unconscious." Human minds experience emotions and desires, but because an AI machine does not have a neocortex, this reviewer is uncomfortable with those assets being included when referring to an AI mind.

Chapter 1 is titled, "Humanity's Special Power." The chapter includes a description of the training a machine intelligence gets using a process identified as gradient descent. The goal of this is to ensure that the machine makes sense when interacting with a human. The authors indicate, "AI is a pile of billions of gradient descended numbers, and that nobody understands how these numbers make AIs talk." They write that an intelligent machine is grown, not crafted. This is because of an important idea the authors identify; humanity has not learned to understand intelligence! They write that as a result, "AIs grown (using gradient descent) do things that their growers did not intend." As an example, the authors include a conversation initiated by "Sydney," a derivative of ChatGPT, in which a philosophy professor is threatened. The conversation is anxiety-producing, in particular because it was not programed into Sydney. The authors write that the conversation is an example illustrating that, "An AI runs on a radically different architecture from a human's." They describe Large Language Models as alien minds, and predict that, once AIs get smarter, they won't keep acting friendly! Yudkowski and Soars go on to write about AIs having the ability to want to achieve a goal, and that the behavior began to emerge in lab tests in 2024. They indicate that the stuff that an AI really wants will be weird and surprising, and not nice. (This reviewer believes that wanting something is the same as having a desire for something, and a neocortex is needed to have desires.)

Later, the AI alignment problem is described (This is the problem of getting your AI to try to do the right thing, notthe problem of figuring out which thing is right). The authors write, "Most everyone who's building AIs seems to be operating as if the alignment problem doesn't exist, and they state an engineering challenge: "How do we shape the preferences of AIs that we can't understand?" They also ask if this new alien mind would be good for humanity. A section in chapter 5 is titled, "IT WON'T NEED US." Here the authors write, "Humans are slow and error-prone and sometimes they get sick. A superintelligence would prefer automated infrastructure, if it could get it."

The predictions the authors make are dark. Chapter 6 is titled, "WE'D LOSE." Early on the authors address the issue of an AI eliminating humanity. They propose that some people, connected to the internet, would help the AI for a fee. And they address the question of how an AI would get funds by describing how it would have access to cryptocurrency. Yudkowski and Soars make a prediction about the end point for AI, which is, "The creation of a machine superintelligence with strange and alien preferences."

Later in the chapter the challenge of protein folding is examined. For information about the challenge see

One of the Biggest Problems in Biology Has Finally Been Solved | Scientific American

AI systems identified as AlphaFold 1, AlphaFold 2, and AlphaFold 3, solved the protein folding problem, working on it between 2018 and 2022.

Similar to what Max Tegmark did in "Life 3.0," Yudkowski and Soars present a story about an AI superintelligence defeating humanity and taking over earth. The story begins in Part II of the book, titled, "One Extinction Scenario." The story encompasses chapters 7 through 9 (39 pages) and is interesting and anxiety-producing to read. The authors then discus lessons from the story, including a lesson for engineers; "If someone doesn't know exactly what's going on inside a complicated device… then they should stop." They explain that the device should be shut down as soon as the behavior looks strange and not wait until the behavior is visibly concerning. Later, they emphasize that, "AI is grown, not crafted. Whatever vast complications lay inside AIs and lend them the powers of intelligence, nobody knows them." Because of this they conclude that, regarding AI development, NOBODY SHOULD BE ALLOWED TO TRY (the emphasis is that of the authors).

The 11th chapter has an interesting title, "AN ALCHEMY, NOT A SCIENCE." This reviewer has read a great deal about alchemy. The science of chemistry developed from alchemy, which really was based on a belief in magic, and was practiced by wizards. The chapter opens with an interesting story about an alchemist who lived in a medieval town. The authors conclude that not one engineer should be allowed to approach Artificial Superintelligence work in the manner an alchemist would. They illustrate that there is an awareness of this by using a 2023 quote from Sishi Sunak, who was the prime minister of the United Kingdom: "In the most likely but extreme cases, there is even a risk that humanity could lose control of AI completely, through the kind of AI sometimes referred to as 'superintelligence."' This can be avoided by not dreaming that artificial minds will exhibit kindness, wonder, and humor.

A challenge that is difficult to confront is identified by the authors; "Nobody knows how advanced an AI would need to be …to secretly copy itself onto the internet." There are other challenging unknowns as well. So, the authors advocate, "All over earth, it must become illegal for AI companies to charge ahead in developing artificial intelligence as they've been doing." The hard part will be enforcing such a law. Leaders/policy makers may need a disaster to be jolted into action. Yudkowski and Soars write, "If not now, when?"

The book is concluded with a section about how we protect "us." They make suggestions for people working in government, elected officials, politicians who are not persuaded about the threat, as well as the rest of us. There are links to access resources on the last few pages. Before the final paragraph in the book, the authors write, "Can earth survive if only some people do their part? Perhaps; perhaps not." See https://ifanyonebuildsit.com/

I hope that at least everyone who reads this review will work to counter the AI threat. I am happy to have come across this book, despite the fact that reading it and writing this review have increased my anxiety about AI/ASI. One of the websites I frequent is Spaceweather.com

The site includes the following statement; This is an AI Free Zone: Text created by Large Language Models is spreading across the Internet. It's well-written, but frequently inaccurate. If you find a mistake on Spaceweather.com, rest assured it was made by a real human being.

I was happy to see that, and promote the idea at every opportunity. Those who read this book will learn a great deal about how far AI has progressed, and the threat it poses.

Gary Taubes, The Case Against Sugar (Knopf, New York, 2016). 286 pp. $26.95. ISBN 978-0-307-70164-0.

This book identifies sugar as the toxin it is. It is not a toxin that causes disease quickly. It is like tobacco, which slowly kills you during thirty or forty years of use. The word "use" identifies people's interaction with sugar as people use tobacco and other addictive drugs, and people USE sugar.

Taubes notes early on that the incidence of diabetes in the U.S. has seen an 800 percent increase since 1960, which coincides with a surge in the consumption of sugars. This is illustrated by the presence of sugary cereals such as Sugar Frosted Flakes and Sugar Smacks at American breakfast tables since the 1950s. In comparison, until Japan became westernized, sugar was used more sparingly, and diabetes was a rare disease in Japan. In addition to expressing concern about diabetes, Taubes also cites evidence that sugar consumption can be implicated as a cause of a number of other diseases.

The connection between sugar and tooth decay has been known even longer than the connection between sugar and diabetes, as Taubes notes in citing the 1939 report of Cleveland dentist Weston Price that "isolated populations – including Swiss mountain villages, pastoral populations in Central Africa, the Inuit and First Nations people of North America, South Pacific Islanders – had nearly cavity-free teeth and retained their teeth for life, as long as they consumed their traditional diet and avoided the sugar and white flour that had come to dominate diets in the United States and Europe."

Because sugary snacks are used by parents as a reward for their children's good behavior and good effort, Taubes acknowledges that his book might be viewed as equivalent to stealing Christmas. (Americans spent ten BILLION dollars on holiday candy in 2015 and 2016 – more than half of the NASA budget for 2018!) Yet, as Taubes points out, there are animals that do not have a sweet tooth, such as cats, chickens, armadillos, whales, sea lions, and cowbirds. But cattle are happy to consume sugar, so it is employed to fatten them. Spraying plants that cattle do not like, with sugar, gets them to eat those plants!

I was also amazed that sugar is used to make the experience of smoking more addictive, as it is blended into the tobacco in cigarettes! This is done to overcome the bitterness of the smoke, allowing for it to be drawn deeper into the lungs, increasing the addictive effect.

Taubes writes that when nutrition research began around the time of World War II the head of the U.S. Food and Nutrition Board (https://www.nationalacademies.org/fnb/food-and-nutrition-board) declared that of all foods sugar is unquestionably the worst. Sugar was not listed among the Basic Seven Food Groups. Sugar industry executives, feeling that they needed only to educate government officials, moved quickly to maintain their profit by forming the Sugar Research Foundation (SRF) in 1943. The SRF became a PR organization for the sugar industry.

Since sugar industry profits felt threatened by artificial sweeteners, Taubes writes that "Between 1963 and 1969 the sugar industry spent more than two-thirds of a million dollars (over four million today) on research designed to force the FDA [Food and Drug Administration] to remove cyclamates from the GRAS [Generally Accepted As Safe] list and have them banned. Much of that money went to obscure research organizations. . . . In May 1965, the FDA published its first review on artificial sweeteners and concluded that there was little to fear." One of the research organizations funded by the sugar industry subsequently published a one-page letter in Nature that feeding laboratory rats artificial sweeteners equal to an amount equivalent to what you would consume from drinking hundreds of diet sodas daily stunted the growth of the rats.

Taubes includes information about a number of comparison studies that were done. One concerns Native Americans. The process of integration into western society did not begin for many Native Americans until they were drafted into the military. In the 1930s diabetes was a rare disease among Native Americans. Taubes writes "As late as 1947 a survey of twenty-five thousand Navajo indicated there was a total of only five cases in sixteen years. By 1963 one in every two Native Americans over thirty on the Gila River Reservation appeared to be an undiagnosed and untreated diabetic." This was a new disease to those people, and produced diabetes-related illnesses – kidney disease, heart disease, hypertension, nerve damage, and blindness.

Taubes begins his penultimate chapter by describing the work of missionary physicians Hugh Trowell, who spent thirty years in Kenya and Uganda beginning in 1929, and Denis Burkitt, who worked eighteen years in Uganda. As the Africans with whom they worked became increasingly exposed to western culture, Trowell and Burkitt observed increasing incidence of diseases, which were mostly chronic rather than infectious, and compiled a list of them in their 1981 book, Western Diseases: Their Emergence and Prevention. Taubes characterizes this list as "a much-expanded version of the diseases Peter Cleave and George Campbell had called 'saccharine diseases' in the 1950s" and notes that Cleve had characterized tooth decay, caused by refined grains and sugar, as "the canary in the coal mine," because it appeared as a precursor to these diseases. Taubes goes on to cite the same trend among the Tokelauans of the South Pacific, as documented by the Tokelau Island Migrant Study conducted by the Wellington School of Medicine (New Zealand), and then turns to search for a cause. Noting the common occurrence of obesity, diabetes, and heart disease, he connects them to metabolic syndrome and insulin resistance, for which sugars (particularly sucrose and high fructose corn syrup) are the "dietary trigger." He concludes the chapter by making the case that sugar is also a prime suspect in the cause of gout, hypertension, cancer, and senile dementia.

He makes the case for hypertension by noting that people in the Philippines, Native Americans in New Mexico, as well as inhabitants of Greenland, China, the Yucatan, and many other places do not face the blood pressure problems Americans do unless their diets become westernized. Taubes describes a study done in 1958 by a medical officer of The World Health Organization, Frank Lowenstein, who studied two groups of Brazilian Indians, one with their diet westernized as a result of their living at a Franciscan mission and the other eating a native diet as a result of their location deep in the rainforest. The Brazilian Indians living at the mission had higher blood pressure and their blood pressure rose as they aged, while those living in the rain forest showed no such symptoms. The higher insulin resistance typical of those who consume "too much" sugar caused kidney function problems, resulting in higher blood pressure.

The American Cancer society was established in 1913. In 1915 the organization produced a book, Mortality from Cancer Throughout the World, identifying the absence of cancer in native people whose diet had not been westernized. Later studies indicated that insulin-resistance led to increased risk of cancer. Taubes writes, "If the sugars we consume cause insulin resistance they are the prime suspects for causing cancer as well, or at least promoting its growth."

The epilogue of the book is titled "How Little Is Still Too Much?" and the first sentence is "It's impossible to say." Taubes relates the situation to cigarette smokers; those who get cancer must smoke too much. Is there a healthy level of smoking, and a healthy level of sugar consumption? "If sugar consumption may be a slippery slope, then advocating moderation is not a meaningful concept." Taubes is honest in his analysis; the evidence against sugar is not definitive, compelling though he may personally find it to be, he acknowledges. But, he concludes, "I've argued here that enough evidence exists for us to consider sugar very likely to be a toxic substance, and to make an informed decision about how best to balance the likely risks with the benefits."

Taubes has done a terrific job of researching both the science involved in identifying the long-term toxicity of sugar, as well as the politics of sugar in this book, and he supports it with a bibliography that is 35 pages long. I found that the disease and illness that a sugar habit may cause is shocking, and the book made me wonder why sugar is added to so many foods and drinks. Because it caused me to examine the amount of sugar in everything I eat and drink, it has had a life-changing effect on me.

I know that my grandchildren are sugar users. Hopefully their habits will be able to be gradually changed. It is very unfortunate that our government has been unable to regulate the use of sugar to prevent the diseases that it causes, as described in this book. The number of people in our country who smoke has been greatly reduced in the last thirty years or so. Hopefully the same can be done to reduce the use of sugar by Americans, and people world-wide as well.