AI for Good: How Real People Are Using Artificial Intelligence to Fix Things That Matter

by Josh Tyrangiel

In contrast to the wave of noisy polemics around AI, AI For Good explores how, in practice, it can actually improve our lives and tells the stories of everyday citizens at the forefront of this new “AI entrepreneurship.”

AI is often framed as a force of radical transformation, either catapulting us into a utopian future or dragging us toward existential ruin. But this book tells a different story. It’s not about high-profile tech CEOs who want to use AI to “break shit,” but about a bunch of smart pragmatists using AI to make the world better.

Josh Tyrangiel’s journey into AI began with a late-night YouTube video featuring General Gustave Perna, the retired four-star general who orchestrated the distribution of Covid vaccines during Operation Warp Speed. Perna’s success—and the end of the pandemic—depended on AI’s practical ability to synthesize and standardize vast amounts of logistical data. AI wasn’t the hero of the story—it was the tool that helped real people get things done.

This book follows those people, who make up a kind of AI counterculture. It explores AI’s quiet revolution in government services, medicine, education, and human connection—places where it’s being used to amplify human judgment rather than replace it. It tells the stories of teachers, doctors, and bureaucrats who often stumbled into AI as a means to solve specific, tangible problems, often with no prior software expertise.

While the loudest voices in AI debate doomsday scenarios and trillion-dollar market opportunities, this book focuses on those working in the messy, incremental, but deeply impactful space of AI practice. However, there is one big caveat—success is not guaranteed. Change is hard. Institutions move slowly. But even in failure there are lessons for everyone who’s interested in using AI—carefully, thoughtfully—to build a better world today.

Genres: Artificial Intelligence, Nonfiction, Technology, Business, Science, Audiobook, Economics

272 pages, Hardcover

Published May 12, 2026

Reviews

[Michelle Skelton]

I went into "AI for Good" curious, but also a little skeptical.

I do use AI regularly but I'm also adamant on teaching students to be cautious about over-relying on it, I’m very aware of the concerns: privacy, data storage, hallucinations, corporate control, environmental impact, and all the harms we already know about, never mind the ones we probably don’t yet.

What I appreciated about this book is that it does not try to convince the reader that AI is harmless or magical.

It also is not a deep dive into every ethical concern surrounding AI. Instead, it is a readable, people-centered look at how AI is already being used by educators, doctors, researchers, parents, and others who are trying to solve real problems.

The education chapters were the most personally relevant to me, but I was also fascinated by the medical examples, especially around sepsis, and the final chapter on communication and nonverbal individuals. The most compelling parts of the book are not really about what AI can do. They are about the people asking whether AI can help them teach, heal, connect, or improve someone’s life.

This is a quick, accessible read, and probably a useful primer for people who want to think about AI without getting buried in technical language.

It is optimistic, maybe even sunny and breezy, but the author does not tell us to leave our umbrella at home. That balance worked for me.

[Ashley]

sometimes you have to dance with the devil

[Val]

This book reads like a Peter Thiel psy-op. It accomplishes its goal of describing some exceptionally positive uses of AI, from driven and brilliant minds solving complex problems. Some of these innovators had multiple chapters dedicated to them. When getting to the chapter on the critique of Palantir? Four pages. Three and a half were dedicated to Alex Karp, with more direct quotes from Karp than any of his critics. Peter Thiel got a paragraph that concluded with “The enmity here is mutual, but also kind of trivial.” No mention of Thiel’s controversial perspective on the survival of the human race. No mention of Palantir’s controversial surveillance projects. No mention of any of the lawsuits. At best, the author is naive. At worst, he lacks journalistic integrity. Either way I’m over it, and only finished the book so I have content to point to at the next company book club.

[Joseph J Studer]

How can you not love a book that compares ChatGPT 3.5 to a Roomba?! It vacuums great, and then head-butts a wall for ten minutes. Spot on and hilarious.

So well written. The witty commentary keeps the book moving smoothly.

Learning how Palantir brought logistical efficiency to the Covid epidemic — in record-breaking time— makes me loathe Peter Thiel just a little bit less. Thank you for illustrating how Elon Musk’s DOGE initiative was a theatrical dumpster fire. I despise him just a little more -- despite his brilliance. And maybe the IRS isn’t a total disaster. Those 5 degree pivots using AI are making the IRS almost seem innovative. Almost.

I did feel that the book fell short in a couple areas, but not enough to warrant anything less than 5 stars. If this book is about the GOOD of AI, then the environmental and resource impacts should have been addressed -- hopefully for the GOOD. Might quantum computing, for example, reduce the resource power needed to drive AI? Will new technologies reduce the storage and water needed for AI?

Also missing, which I feel is lacking from many promises of AI, is essentially -- Does AI simply "speed up" things we are already doing? The Sepsis example with Epic illustrates this well. Epic already provides tools to alert users of a potential for sepsis, without the use of AI. The Bayesian tool just does it... a little better? The non-verbal autism AI project is able to crunch numbers faster, but the story never ended with some big break-through in terms of understanding non-verbal autism. And the garbage pick-up story and the airbag story didn't leave me feeling like AI was a technological break-through that revolutionized garbage pick-up efficiency or in figuring out how to calculate the volume of baking soda needed to simulate an airbag.

Like so much of AI, I feel like the big payoff so far is just theoretical, hypothetical or... "right around the corner". The world does not feel much different now than it did five years ago. But I'm hopeful.

[Bernie Gourley]

There're a lot of books out about artificial intelligence (AI) these days. There are how-to manuals. There are books about the making of billionaires on the back of AI-centric business models. There are books that consider how AI will destroy the human economy (and possibly humanity as we know it.) Josh Tyrangiel's book seeks to carve out a niche by taking a positive view towards AI but focusing not so much on how it produces more billionaire tech executives, but rather on how it can help fix persistent social problems in education, healthcare, governance, and human communication and connectedness. Of course, this isn't a completely separate topic from business use of AI (e.g. healthcare is one of the biggest businesses in America [which is no doubt emblematic of America's unrelentingly shitty healthcare;]) however, these are areas that each feature their own unique challenges, problems for which the lessons of the business sector, broadly, are often of limited value.

I found this book to be illuminating. It introduced several fascinating characters from various domains. Among the most intriguing discussions were those with a short-lived DOGE employee and one with a Hoosier high school principal. It was also interesting to learn about the evolution of AI language translators.

If you are interested in how AI is being applied beyond hardcore business uses like supply chain optimization and computer programming, you may want to give this book a look.

[Spencer Rugen]

3.5/5. This was better than I expected but still not terribly exciting. I was pleasantly surprised that it followed like a Freakonomics vibes in the sense that the book covered various ambitious AI ventures in parts. So we did get to sit with the characters for a while. It covered school systems leveraging AI with Sal Khan to give kids with varying abilities the same attention; Cleveland Clinic using AI to expedite cardiac MRIs in one hospital and flag for potential sepsis cases in another; MIT engineers using it to study non verbal communication from kids with autism. All cool stuff but didn’t feel super earnest at times. A lot of it also went over my head but not mad I read it!

[Jay Mawicke]

Didn’t hate this book but its thesis is pretty weak overall, wasn’t really sure the book was trying to make any overarching point. It’s a little comical how many random people this short book introduces you to, i didn’t really find any deep resonation with anyone besides the mother of Felix. That part alongside the parts about AI for government efficiency were interesting, but I felt like the book was trying to be overtly “human” ie focusing on human characters not AI itself as a sort of spite towards AI. I read this book to learn more about the direct application of AI today, I got a little but the specifics are still really vague and all these anecdotes across the board didn’t help me too much in building that framework in my head. Fine read if you’re looking to learn ways people are applying AI to real life, but nothing particularly eye opening or epiphany inducing.

[Jay]

May 17, 2026

“AI For Good” is a well-written and engaging book that delves into the positive potential of artificial intelligence. The book is filled with intriguing stories and examples that demonstrate how AI can be harnessed for good, presenting a hopeful and optimistic outlook without being overly idealistic.

What I appreciated most is that it serves as a strong counterweight to many of the doom and gloom books about AI replacing people and destroying careers. Instead, it does a good job illustrating how AI can amplify human capability and creativity when used as a partner rather than a replacement. The examples throughout the book reinforce the idea that the most powerful outcomes will likely come from people and AI working together, each contributing different strengths.

That said, many of the stories stay at a fairly high level. At times it felt more like reading a collection of thoughtful long-form articles than a deeply immersive exploration of the topics. I often found myself wanting the author to go deeper into the experiences, decisions, and implications behind the stories. Still, a very worthwhile and timely read.

[Jessica]

As the debate surrounding AI usage rages on, this book offers a few stories of those using AI to solve problems in our society. From differentiation in education settings to sepsis prevention and helping nonverbal children, it was interesting to see just how beneficial AI can be in the real world.

[Evelyn Petschek]

Interesting read. Tangible and interesting real examples of how AI can be effectively used.

[Jon Barr]

A few stories of the positive projects being undertaken with AI. I especially enjoyed his section about education.

[Cyndi T]

Some good examples of positive AI use

[Tom]

A collection of AI case studies focused on education, healthcare, and government. Cases included Khan Academy and the Cleveland Clinic.

“It wasn't long after that I stumbled across General Perna and Operation Warp Speed. Soon there were promising threads elsewhere-in other government agencies, in education, health care, human connection-where people were tinkering with Al to make the things that matter in the world work better. To make our lives longer, and more meaningful. Not in some distant future, but today. It's stories from this AI counterculture that make up this book.
Like the doomers and accelerationists, these people also tend to have things in common. Most are quiet, practical, with little vanity. Many had no previous software expertise. They'd run into a problem that defied conventional solutions, and were stubborn or desperate enough—or just cared with enough irrational force-to keep going, even if it meant having to learn more about technology than they'd ever wanted to.”

“The result of all that computing is that ChatGPT appears to know and understand things, That's the magie and the trick: ChaiGPT doesn't think, or have opinions, or a moral compass. On its own, it has no ability to assess whether what it's saying is correct or useful, which is why it can explain the laws of thermodynamics one moment, and confidently invent fake laws of thermodynamies the next. The model optimizes for fluency, not truth, It does not fact-check itself because it does not understand facts.”

“There are many things we cannot control, so it's important that we regiment processes that create uniformity, transparency, and ac-countability," Mihaljevic says. "It's roughly the same approach with AI. In order to innovate, you have to standardize the things that are indisputably the best practice. That allows you to focus on all the stuff that takes you by surprise.”

“Coronary angiography is the gold standard for detecting coronary artery disease, but it requires an expensive medical team to make a small incision and send contrast dyes into the patient's body.
It can take weeks to schedule and hours to recover from. In between there's a bunch of other tests and scans that are a medical version of the contractor problem: Your heart exam can be fast, good, or cheap.
Pick two.”

“Chandra had underestimated the challenge of change management in health care. The second was that identifying opportunities for Al was less important than identifying people…. “Trying to find a way to make health care cheaper, better, safer, more sustainable-anything and everything that technology can do to fix that or help with that is imperative. I've got to find people who believe that. It's not easy. If you were to ask me to name people in the organi-zation? Less than ten. I don't know all eighty thousand people, but in my interactions there are maybe half a dozen go-to people who I think have that fight in them. Everybody else is wonderful. It's not my job to judge. They do what they do."
I asked what these less-than-ten people had in common. Chandra said they would "put their neck" on the line. "I look to see: Do I have a partner who will live and die with this problem, or is this en-tertainment? I have to have somebody who can look me in the eye and say, 'Yes, I will kill myself to drive this."

“The only adjustment to his technique was that he needed to narrate his actions so the app could listen in and make notes. In the hands of my own doctor-who barely speaks and almost certainly doesn't know my name-the app would have been useless. Boose, with his Midwestern warmth and singsong cheer, was waltzing with it.
When the fake exam was over, Boose hit a button to generate the notes and handed me his phone. The tests he'd ordered were spelled out, and the patient instructions were clearly sequenced. If a patient or their caregiver needed materials in Spanish or a bunch of other lan-guages, no problem. Boose reread the notes for accuracy-nothing is initiated without a doctor's review-and mimed pressing send. "Then the tests get ordered and the notes go into the EHR."

“The scribe pilot was voluntary-all 250 highly educated, fully grown adults who participated put their hands up to do so. And yet: Some never bothered to turn the app on. Others tried it and gave up when they realized they'd have to verbalize their patient exams for the scribes to be useful. "You give an app to fifty people and I hear back from thirty," says Boose. "Twenty ghost me.
Then I try to chase them down, saying, 'Hey, you kind of said you wanted to try this?' And they're like, 'Oh no, it's not for me right now.' So there's a lot of that going on."

“With Ambience, 32 percent of doctors said they increased their face time with patients, improving the care experience. The Clinic also saw a 7 percent bump in same-day chart closures, which leads to faster billing. Chandra's team is now weighing how to roll out scribes in hospital settings.
"We're seeing a few minutes shaved off of each visit. We're seeing people closing their charts," says Boose. Ambience will remain volun-tary, but four thousand providers are using it, and some doctors are taking advantage of the time they used to spend writing notes to add extra patients. "We had a physician who was getting ready to retire say, 'I think I can work for another year or two with this.' And we've had others say this is life-changing. I would even say it's life-changing for me."
I told Boose it seemed like such a small thing, and he laughed.
"So are our margins."

“The team of twenty-five was initially scared that Hospital 360 would take their jobs. Pappas told them, "We're not going to get rid of you, because we need your eyes on what's happening. You're the final check." But as Cleveland Clinic expands and potentially acquires more hospitals, it doesn't anticipate adding anyone new to the team.
It made a bet on AT finding efficiencies, and the scoreboard didn't lie.
Daily transfer volume was already up 10 percent, and new patients were being admitted much earlier in the day. Minutes had been cut from each part of a patient's surgical journey. Emergency room wait times had been reduced by ninety minutes.”

What Jessica Shieh once said about the model behind ChatGPT-you'll be disappointed for a long time until you're not-applies far beyond one product.

"If you tell someone that even imperfect Al has contributed-even in a small way-to reducing sepsis mortality by 40 percent, you would have to call that success." Bayesian had standardized a basic level of detection across a massive, complicated system, prompting Mihaljevic to add a new maxim to his core beliefs: AI does not need to be perfect to be useful.”

“In metaphysics, ontology is the study of being. In Al, it's come to mean the untangling of messes and the creation of a functional information ecosystem. Once Palantir standardizes an organizations data and defines the relationships betteen the pipes, it can build an application or interface on top of it. This combination-cleaned and integrated data, useful app-is what allows everyone from middle managers to four-star generals to have an AT copilot, to see themselves with the God view. "It's the Iron Man suit for the person who's using it," says Krishnaswamy.”

"The act of writing data pipelines is usually pretty simple," says Aaron Jaffe, who was the lead product engineer on Operation Warp Speed. "The challenging thing is not always the engineering work. I's collaborating with the team and working with the customer.
Like, what are the actual ideas we're trying to show with the data? And how do I get what I'm building to advance what the mission's trying to accomplish?" Like Debbie Kwon at Cleveland Clinic, experience had taught him that "culture, alignment, people-that tends to be the hard part."

“Palantir's CEO, Alex Karp, told me,
"It's not a completely meaningless distinction. People hate hype, so we should try to be precise about how we describe or define things." He also finds the whole "Is it or isn't it AI?" debate to be ex-hausting. "The best definition I've got is that AI is software that works.
And our stuff works."”

“We take for granted that computers are good at math, but this is not actually true. A calculator doesn't know what a square root is-it just follows the instructions in its code for producing one. The same is true of ChatGPT, except its instructions aren't as simple, or even mathematical. It's a language model, trained to predict the next word or symbol in a sentence based on text patterns it's seen. So when you ask a language model to solve a math problem, it's not crunching numbers-it's trying to guess what a "correct-looking" answer to that kind of problem should be based on previous examples. That's why it can sound so fluent while being so wrong.”

“Students are not Khanmigo's only users. Khanmigo launched with just a few teacher tools, and now has close to thirty. It writes rubrics-scoring guides that break down an assignment so students understand what the grading criteria are (and leads to better student work). It unpacks curriculum standards into learning objectives.”

“"Maybe that's the lesson?" Walls said. "I don't have a lot of government experience, but maybe you have to make big AI policy things seem less big. Don't try to do too much or beat the hard parts of the system into working the way you want. Start in softer places where nobody's paying attention." I couldn't tell if he was describing a policy version of agile development or making a case that the secret to happiness is low expectations. "Then if you deliver the right outcome, everybody will be so happy they won't even care if you try it again."”

“The term human-in-the-loop evolved from the aerospace and defense industries, where the use of automated flight simulators and war games has been common for decades. Humans were given authority over the systems to help prevent catastrophes and provide cover, so that no general or CEO would have to testify, "Jeez, Senator, we just thought the drone would know that the stadium was full of senior citizens." In artificial intelligence, human-in-the-loop means humans actively participate in training, testing, refining, and operating AI models and programs. It's an approach that creates a continuous feedback loop between people and machines, theoretically making both perform better. The cover is valuable, too.”

“Al tools such as Llama, Claude, and ChatGPT can digest COBOL and ALC and create pseudo-code. It's not a one-for-one translation machine-"I never want anybody to walk away thinking there's a magic bullet that generative AI provides," says Pandya. It's an AI assistant that extracts the logic of the original code and gives Java developers a foundation to build upon. What took months on the IMF project, Al is doing in days.
These same tools also automate documentation, the process by which software engineers are supposed to-but never do-note all their thinking so that future engineers can modify or maintain the code. "When I talk to people outside of work and say we're using Al so our developers can save two hours a week on documentation, they're like, 'So what?' But it matters!" says Pandya. "When we have five hundred or a thousand developers, all of a sudden two extra hours a week turns into some real development progress that we can make at a much faster rate."”

“Conceptually, human-in-the-loop gives the IRS a solid legal framework for AI that also makes sense technologically. Practically, it means that IRS employees deal with Al a lot more than its customers do.”

“The bottleneck was data. Johnson kept staring at Bonini's para-dox, which holds that the more information you stuff into a model, the less useful the model becomes. (A perfectly detailed map would be the same size as the territory it's mapping-which defeats the purpose of having a map.) She had tried using sensors to capture everything she could about Felix's communication, and the process annoyed her son and generated so many signals that the data obliterated its own mean-ing. Simplify too much, though, and she risked draining away the richness she wanted the world to recognize.”

“Meaning operates on its own timeline. Which is one of the reasons I've tried to balance explanations of how Al works and the societal problems it can solve with an understanding of the humans leading the effort. The tech may change faster than the weather, but we model ourselves after other people. Sal Khan, Peggy Buffington, Debbie Kwon, Chris Nguyen, Deacon Maddox, Cliff Walls, and Kristy Johnson all have extraordinary qualities, but it's their ordinary desire to be useful that abides. Each wanted to fix something that matters in the world, and each figured out a specific way that AI could help.
If you can relate, what comes next isn't that complicated. It just requires a bit of effort.”

“First, get to know the technology. You don't need to become an engineer, but you should flirt with AI enough to know what it can actually do versus what companies claim it can do. Spend an afternoon with ChatGPT, Claude, Gemini, or whatever equivalent you like. The free versions are more than adequate to learn where the tech is useful (drafting emails, brainstorming ideas, simplifying concepts, helping with home repairs) and where its value is less clear (anything requiring subtle human judgment, or where being wrong has consequences… Just as important as what you begin to do with Al is what you decide not to do. When a company offers an Al feature that makes you uneasy-because it's generating sloppy content, or replacing a human you'd rather talk to, or feels like a solution in search of a problem-don't use it. This isn't about being a Luddite or a dissident. Its about the fact that user behavior is the most powerful feedback mechanism in tech. Companies will build what we tolerate. If enough people ignore a feature, it dies quietly. If everyone adopts it, it becomes in-
frastructure.
Every AI service is a swap: convenience for something else. Sometimes it's your data. Sometimes it's your attention. None of this makes Al inherently bad, but the terms should be clear. Read the privacy policy once in a while. Better yet: Make an Al summarize its own privacy policy in bullet points an eighth grader can understand. Know what's being taken, and whether what you're getting in return is worth it.
Most importantly, stay close to Team Human. This means supporting Al that amplifies human judgment, not replaces it. A teacher using Al to customize lessons for thirty different students is different from a company using Al to eliminate teachers. A doctor using Al to spot patterns in scans is different from an insurance company using Al to deny claims. Support the former. Be vigilant about the latter The question isn't whether Al is involved in our daily lives-its utility is undeniable, and the regular improvements are worthy of awe. The question is who it's empowering.”

[Joshua Rosati]

Almost a round up, gave some decent insights.

[Tom Armstrong]

The world today seems split between AI Maximalists who view the technology as the solution to all of the world’s problems and AI Doomers who are certain the spread of artificial intelligence will bring about a long-feared dystopia.

In AI for Good, Josh Tyrangiel tackles a middle ground by starting from the other side of the equation. Instead of “what can AI do,” he asks, “how can I solve this problem,” and finds that sometimes AI is the solution. This book isn’t a prescient look into the future but rather a sober look at how AI is solving today’s challenges. It captures the successes, but also the struggles, false starts, backtracking, and messiness of the practical application in high-stakes settings of technology that changes week to week.

Through a series of compelling stories from the military, government, education, healthcare, and research, we’re shown how AI can move the needle on intractable problems, but only if it’s applied by a committed group of humans who deeply understand the problem and know how to navigate the people, politics, and culture that make up our organizations.

Tyrangiel shows how AI helped Operation Warp Speed ensure COVID vaccines were delivered when and where they were needed, how Khan Academy and OpenAI struggled to incorporate chatbots into Khan’s learning platform, and how the Cleveland Clinic is using AI to enhance patient care. In each case, the technology turned out to be the easy part.

In many ways, this is an optimistic book. We see people, most of whom have no special training in technology or machine learning or large language models, find ways to plug AI into a larger ecosystem to help them with their lives’ work. But this is also a sober book. There’s no magic wand any of us can wave, no single prompt we can craft. There’s no shortcut to doing the hard work.

AI for Good is not a book for Maximalists or Doomers. They’ve already made up their minds. It’s a book for the rest of us who show up every day trying to move something forward and wondering whether this particular tool is worth our time. For that reader, and especially for those who lead or work within the large institutions where change is slow and politics are unforgiving, Tyrangiel has written something useful. He doesn't promise transformation. Instead, he offers evidence that progress is possible, and an honest account of what it costs.

Thanks to NetGalley and Simon & Schuster for an advance copy of this book. All views expressed herein are mine and mine alone. Beyond a free review copy of the book, I was not compensated for this review.

[Margo]

This was very good - looking for an overview of real world applications of AI.

Found this to be very Michael Lewis-like. The author has a thesis - that AI can and is being used for really cool, GOOD things - and then finds several examples with lots of real people and anecdotes to back it up. The writing style is very much like Lewis as well, breezy, accessible.

He focuses on four industries:

Education - profiling Kahn Academy and how they partnered with Open AI to build a model that would act as a tutor for students, to help them get to the answer, but not provide the answer. They also had a model to help teachers create lesson plans. He had some interesting examples of how it worked for some students, didn't work for others either because they wanted a human, or they were to unengaged to use it to improve their understanding. Would have liked a 10,000 foot view; Khan isn't the only company using this, all publishers/educational content people are - would have liked a broader perspective.

Healthcare - Focused on the Cleveland Clinic and their CEO who's goal was to use AI only if it improved PATIENT care, not improve procedures, or reduce costs. The example used was the cardiac MRI and how using AI models of this procedure vastly improved heart outcomes. The cardiac MRI is way more challenging than an MRI on another body part - the technician has to be trained to do the imaging while the heart was still, in between breaths. By training AI to analyze scans, filtering out blurry images, they could do the procedure in less time with more accuracy and more importantly making it available to more patients. Using AI to interact with health databases for more accuracy - most compelling example was to use AI to predict patients most likely to have sepsis and to provide treatment before disease became fatal.

Government - using AI for procurement to ensure vaccines were available during Operation Warp Speed during COVID. The disaster of the DOGE experiment.

Neurodiversity - to use AI to create models of alternative speech patterns (autistic individuals) so that "normal" people could better understand these people.

In all, very good, would have just liked some more context for each of these examples, although as soon as he provides context, the situation would be different in this ever changing field.

[Scott Bordelon]

AI for Good is a collection of example applications for AI that are small, practical deployments rather than grand, futuristic visions. Real people (teachers, doctors, civil servants, parents) use AI to fix broken systems. It’s a grounded, case-driven book illustrating how human-centric, incremental AI has been used to improve education, healthcare, and public services.

First of all, this book was extremely well written. Tyrangiel’s style is casual, informative and humorous, which makes the book engaging and somewhat relatable, but after Part 1, the content became boring, repetitive and uninspiring for me. Part 1 had an interesting history lesson about OpenAI and its early use in education before ChatGPT became a household name as an LLM. Then the book proceeded with more examples of how AI has been used in healthcare, government, and enables mission-driven people to scale their compassion rather than replace them. That’s all good I guess, but I’m already aware of lots of applications for how AI can be used today. I was hoping for more forward-looking proposals for how AI might be used in the future, applying some of the state-of-the-art AI research that hasn’t quite come to fruition yet. Reading example after example got pretty dull after a while.

Tyrangiel’s examples had a handful of recurring takeaways:
1. Start small with narrow problems that produce real wins.
2. It’s important to keep humans in the loop for the sake of grounded judgement, empathy, and accountability.
3. Expect cultural resistance. In almost every example, the hardest part wasn’t writing the code; it was getting people to change their habits.
4. The best AI deployments (in this book) weren’t coming from tech giants. They came from mission-driven people who were close to the problems they were trying to solve.
5. People shape the norms, uses, and boundaries of the technology. Its use for public good is a choice.

Tyrangiel is filled with optimism but also pragmatism. He doesn’t treat AI like magic, and he acknowledges the hurdles encountered in his examples that ultimately show the power of AI when wielded by people who care deeply about the systems they’re trying to fix. This is probably a good book for AI skeptics who stubbornly refuse to believe there’s any net benefit from AI. It just wasn’t for me.

[Heidi]

AI for Good is a series of stories that all circle the same idea: AI isn’t going to save us, but it’s already deciding things that matter.

Instead of speculating about the future, Tyrangiel stays in the present. The Operation Warp Speed story strips away the mystique and proves that AI is not magic.

The education chapters show how the Khan Academy rollout was not smooth, which as a parent, didn’t surprise me. Students dislike it (mine included), teachers lean too hard on it, and districts try to reshape it. In Newark, it starts working for the school system because people refuse to give up on it.

Not everything works. The LA chatbot story just… unravels. It shows the cost of relying too hard on the hope that labeling something AI solves the underlying problem.

I’d heard the Eliza story before, but it’s one of those things that sticks. A 1960s chatbot that just mirrored users’ words back to them still created emotional attachment, even when people knew it was a script. The surprise wasn’t the technology. It was us. Now we’re doing the same thing at scale through ChatGPT and Claude.

In the recycling story, the system technically works, but only because it nudges behavior in a very human way. Not by being smarter, but by being persuasive. The line between intelligence and influence is thin.

It keeps coming back to the idea that the right move is to engage, to get your hands on the tools and shape them. That sounds empowering. But it also assumes a level of access and agency that not everyone has. It acknowledges this, but moves on quickly.

It also avoids going too deep into the incentives behind AI, which feels like a choice. This is not a critique of the system. It’s a look at what people are doing inside it.

It’s readable and doesn’t try to impress you. It tries to show you what’s happening. The message is that AI will amplify whatever we point it at. AI is not going to fix broken systems. But it will make them more efficient, whether we like that or not.

[David]

***.5

The author attempts to cut through the hype and evaluate actual productive uses of AI, in fields like education, medicine, and government efficiency. He starts off with a good explanation the difficulty with the term "AI" meaning too many different things, but then proceeds to blur the lines in the rest of the book, not adequately differentiating between traditional machine learning "big data" statistical analysis and the current crop of LLM-based generative AI models.

The first and last chapters are the best, in between he goes into what I consider "journalist mode," focusing less on concepts and trends and more on individual people and the companies they own or work for. While this can help humanize the issues, such as the researcher developing AI-based communications tools to help communicate with their non-vernal autistic child, it tends to shrink the scope and hence the potential impacts that the technology is/will have on everything.

The other problem with this approach is that he's overly reliant on the people he managed to score interviews with to provide information. As a result, we get some one-sided accounts of contentious issues, and large swaths read like Khan Academy and Palantir marketing brochures. As with many tech/business writers, even while criticizing his erratic disposition and extreme politics, he also wildly overestimates Elon Musk's engineering prowess, attributing achievements that he had no part in other than in an executive (ie financial) capacity.

The book does have some merit in looking outside of the latest glitzy press release, emphasizing that the best applications for AI as it currently stands often receive the least attention (and hence funding), while useless gadget-level applications are overly hyped.

[Eric Goldman]

I don't know who the book's intended audience was supposed to be, but I don't think I was in it. Some of the things I didn't get:

* the book never defines the titular "AI." Instead, it non-defines AI as a "constellation of overlapping techniques and capabilities" with "porous" boundaries. By eliding any rigorous definition of AI, the author gives himself permission to discuss divergent technologies in a single book, some of which didn't seem very "AI"-like at all.
* along those lines, many of the stories involve complex systems integrations at scale. These are interesting topics for those who have ever encountered such challenges, but they may have little or nothing to do with "AI."
* the book lost me on page 3. Its curtain-raising story was about Operation Warp Speed, with the grand reveal being a celebration of Palantir for its role. Palantir...really??? Later, the book briefly both-sides acknowledges some criticisms of Palantir. Too late. If your curtain-raiser involves a shadowy and reviled government contractor that may pose significant dangers to our democracy, maybe choose a different curtain-raiser?
* the book's subtitle says the book is about "how real people are using artificial intelligence to fix things that matter." The book intentionally valorizes humans in the story, but that leads to many unenlightening tangents and side remarks.

After reading the book, I came away with no greater understanding of AI or how it can help. At most I learned about a few anecdotes of how technologists are trying to solve some problems, without any overarching themes or insights.

We very much need a book that explains the many ways that AI can in fact improve humanity. This isn't that book. Pity.

[Richard Thompson]

There seems to be a genre of nonfiction writing consisting of anecdotal stories about smart humane diligent people who find ways to make a difficult system work. Think of books about teachers or about scientists with theories outside of the mainstream or stories about aid workers who find ways to bridge cultural divides. These stories are always uplifting, but in the service of storytelling they sometimes omit inconvenient facts.

This book tells feel good stories about people in diverse fields from logistics to teaching to medicine who use AI to deliver great results that help people and that always include a human in the loop in some critical way. It's just the kind of stuff that the Pope advocates in his recent encyclical on AI. The big problem, apart from the fact that stories like this always turn out to be not completely true, is that these are all situations that are essentially non-reproducible. Just give me 10,000 ethical geniuses working on tractable adequately funded socially positive projects, and I'll show you 1000 good outcomes that make good stories. But ethical geniuses don't grow on trees; nor do tractable adequately funded socially positive projects. There will always be a few of these and in the right circumstances AI can be an important part of the solution. So yeah, AI can do some good, so can social media, atomic energy and the military. But they also all involve massive dangers that can easily outweigh the good, so they all need to be watched carefully, regulated and channeled in positive directions.

[Chris]

Tyrangel’s case studies of people using artificial intelligence to solve real problems is an eye-opener. The technical details can be difficult but the author stays focused on the amazing power of AI as a tool in the hands of people who are open to using it for purposes that improve lives. He does not discount ethical and environment concerns but does not delve into them.

One of the more easily understood examples is the effort to distribute COVID vaccines equitably to all 50 states, with speed and efficiency, as soon as approved. Operation Warp Speed would clearly have been less successful without a handful of people using AI to summarize thousands of data points that enabled the General in charge to make a plan and implement it.

Each case study usually has multiple points of failure. Khan Academy’s collaboration with Hobart, Indiana, students to test an educational AI tool hilariously details the amount of trial and error needed to mold and tweak AI to do what you intend it to do.

Schools, hospitals, government agencies — all could benefit from AI but institutions are filled with people resistant to change and skeptical of technology. And this particular tech unfortunately comes from a handful of ethically questionable millionaires whose focus may not be on improving anyone’s lives but their own.

Still, the author advocates getting to know AI and separating the technology from its founders and AI for Good presents a compelling argument.

[Janice]

This book introduces AI in ways that are good, in ways that can help people in their jobs, not to replace people's jobs. It talks about how it can help teachers and students, identify how hospital patients might be close to death under certain situations, help non-verbal autistic and special students be able to talk, etc. I was interested in the person who worked for DOGE since he had not voted in a recent election (which disqualified voters voting against Trump) and was placed to work in the VA section. He discovered no fraud, but did identify a significant way to make the system work more efficiently and less expensively. He was told that that was not his purpose. However, the book did a address how it was important to use AI in a positive way, as when AI could be trained to "provide support and advice" and sound empathetic, but could lead to advice for suicide and life-long counseling with AI. It also did not address other negative aspects.

[Maureen]

This book about the looming specter of artificial intelligence will not help you to understand how AI works. But that's okay! Because the purpose of this book is to show how AI is already doing some real good (not just a solution in search of a problem, skewing financial markets and destroying the environment). So that's nice to learn.

The author tells us about a number of ways that AI is being deployed to make lives easier, to make lives safer, to make lives better. Individualized teaching and learning. Mind-numbing reading of medical scans. Tedious transcription of medical records. Incredibly complicated hospital systems management. Facilitating communication with those once thought non-communicative.

You need to listen to this book or read it so that you can regain some of the awe that technological advances can inspire. AI can fix some things that matter in the world (if it doesn’t blow up the financial markets and fry the environment first!).

[Melissa Kearney]

Excellent contribution to the public discussion around AI and how it is likely to affect our lives and our world. Josh has given us all a gift with this book. He deliberately avoids perpetuating the hysteria on either side of the issue and instead, gives us -- in accessible, engaging, and at times chuckle out loud funny prose -- stories about real people who are using AI to improve their classroom, their hospital, and their government agency. The stories also highlight some key points and lessons, most notably that the promise of AI to do good relies on it being deployed by humans with domain expertise who deeply understand a particular problem and can envision how a machine can help address it. Humans are the heros of his stories, and current AI models are efficiency-enhancing, albeit imperfect, assistants. Well worth reading and considering!

[Jill Morningstar]

The author did a great job of making complex things easy to understand. What I did not expect, yet appreciated, was his ability to explain how difficult it is to bring about change for good when the majority are resistant to any kind of change. People are the worst. Even though this was a one-sided book (the "for good" part was literal), I really appreciated what the author was trying to do. As a bonus, the back of the book has a two-column table with questions to help people decide if the book is for them. Then there was the following question: Did you watch The Matrix and root for Agent Smith? I cracked up laughing. I answered YES to the questions in the "Buy" category and YES to the Agent Smith one on the "Do Not Buy" side. Come on, the way Hugo Weaving says "Mr. Anderson" is one of the highlights of the movie.

[Brenda Kochis]

My original expectation of this book was that it was going to tell me how to use AI. But instead it showed several real-world use cases where AI was incorporated. The use cases were highly complex which gave me an appreciation of what is involved; it is not just a matter of using a prompt.

The cases were Kahn Academy, Cleavland Clinic, the IRS, building something for autistic communication, and the Pentagon. In every case the issues were complex, data was often messy, and it took a champion or two to see the project through. Not to mention it often took time to refine the project as well as ongoing tweeks.

In these cases there was a specific desired outcome, but often a number of barriers to accomplish the objectives. Fastinating reading and a good sense of what it takes to incorporate AI into solutions.

[Carol]

AI has people looking at the best and worst of what the future holds, but this book explains more of how it came to be, why it is the way it is and shows how it has learned how to help in many areas.

The most important thing I saw was how AI helped during the Covid crisis days. I never knew how slow bureaucracy would have been if AI had not been used to collect various diverse batches of info and make it useable. For all of us, it sped up location and distribution of supplies and vaccine to get it into the hands or arms of those who distributed it and those who needed it.

Content is dry at times, but factual. I'm glad I read it.
Thank you NetGalley for an advance reader copy. Honest opinions expressed here are my own and are freely given.

[Jeff Berman]

When was the last time you read non-fiction with passages more exquisitely crafted and evocative than most heralded novels?

“Buffington is in her early sixties, tiny, stylish, with shoulder-length hair the same shade of gold as the cross around her neck. When she parks her black Cadillac directly in front of a school - and enters in her black suit, black heels, and black Coach bag - the effect is like Johnny Cash arriving in Loretta Lynn's body.”

“Birx's pandemic tenure was not smooth. In fairness, I would rather dance the hora with a gun in my mouth than try to speak scientific truths while standing next to Donald Trump.”

This is an antidote to doomerism from Josh Tyrangiel, a skeptic who happens to be an epically talented storyteller.

[Mark Baker]

I have been paying attention to the subject of AI/Robotics for about 40 years. As a current skeptic that it will actually help humanity, I find this book to be very valuable. While it discusses skepticism and why we should be, that isn’t the primary focus. It discusses several specific cases of how it can be used in ways that are beneficial to man. They are enlightening. It also discusses the need of us all to be very aware and what we can do to make it a tool and not just empower a few companies.