Teaching Machines To Learn

“AI for Good

 

By Josh Tyrangiel

 

Simon & Schuster, 272 pages, $29

 

I Am Not a Robot

 

By Joanna Stern

 

Harper, 320 pages, $32

 

As utopian accelerationists and apocalyptic doomers battle over competing visions of artificial intelligence's future, a third group has quietly applied the still-evolving technology to the problems of the present. These understated pragmatists are the heroes of Josh Tyrangiel's "AI for Good," a kind of "Profiles in Courage" for the implementation set.

 

Mr. Tyrangiel, a writer at the Atlantic, introduces us to Rita Pappas of the Cleveland Clinic. As the medical director of hospital operations, Dr. Pappas used AI to tackle the inefficient management of admissions, discharges and transfers. We meet Gustave Perna, the Army general (now retired) and chief operating officer of Operation Warp Speed. During the Covid-19 pandemic, he corralled data from across the healthcare industry to build a real-time vaccine supply chain, racing to deliver before the bureaucracy reasserted itself. Neither technologists nor Luddites, these leaders share a taste for challenging the status quo and a resolute determination to fix something that matters -- and are willing to learn whatever is necessary to do it.

 

Ambition alone doesn't get you very far. Successful implementation, Mr. Tyrangiel writes, relies on teams willing to endure the "grinding work of auditing and reinforcement" -- reviewing model output, feeding in corrections and repeating the cycle until results improve.

 

It also requires the right technology partner. Though Mr. Tyrangiel acknowledges that "the very mention of Palantir curdles the blood of progressives and a lot of the military establishment," the company repeatedly earns plaudits for its ability to burrow into an organization's tangled data and impose order. "We've always been the mole people of Silicon Valley," explains Akshay Krishnaswamy, the company's chief architect.

 

The success of an AI deployment, Mr. Tyrangiel writes, often hinges on a point person at the technology partner who combines technical fluency with the ability to inhabit a client's challenges and understand the environment in which the technology must function.

 

OpenAI's Jessica Shieh, for instance, helped the educational nonprofit Khan Academy elicit the best results from the AI powering a new chatbot tutor. At Operation Warp Speed, Palantir's Julie Bush grasped the project's operational requirements and translated them into technical specifications that Palantir engineers could utilize.

 

Mr. Tyrangiel recognizes the lengths to which innovators must go to evade institutional resistance. Hospitals, he observes, "have exalted missions, which can obscure the fact that they're just as full of territorial jerks as investment banks." Debbie Kwon, a cardiologist at the Cleveland Clinic, had to overcome bureaucrats and colleagues who sabotaged her attempts to recruit engineering talent and apply AI to MRI technology. Transforming healthcare with AI is so difficult, Mr. Tyrangiel writes, because the technology must work flawlessly and you must navigate "professional cultures, government policies, entrenched stakeholders, money, rivalries, emotions" and other factors. "The code may be willing," he adds, "but the flesh is often weak."

 

AI is revolutionary, but as Mr. Tyrangiel stresses, its implementation tends to be evolutionary. Khan Academy's chatbot tutor has enhanced the learning experience, but educating children is still difficult and the classroom is largely unchanged. We shouldn't be surprised. Transformative technologies never arrive fully formed; they rely upon determined practitioners committed to the painstaking work of distilling extravagant promise into daily practice.

 

Joanna Stern's preoccupation with AI is personal. In "I Am Not a Robot," she amusingly explores "what this world-changing technology actually means for us and our children" by immersing herself in AI for a year.

 

Ms. Stern, a tech journalist who worked for this newspaper for 12 years, takes us through AI-enhanced mammography ("what a beautiful day to have my breasts crushed by a machine") and a family vacation built around self-driving Waymo cars, which leave her young children at first mesmerized and then bored as the technology becomes routine. She applauds an AI masseuse that devotes 20 minutes to her gluteal region. "For a human massage therapist," she observes, "this might have been awkward," but for a robot not so much. Less joy is to be found in an AI-enabled plush toy that struggles to parse basic commands. "You're the worst, Gabbo," her son informs it.

 

AI, Ms. Stern observes, can "act like an amplifier." Several radiologists tell her they wouldn't want to work without AI assistance. But it can also enhance our baser instincts. She bristles when she realizes some dentists are using AI images to upsell patients -- deploying color-coded overlays to dramatize borderline findings and justify questionable treatments.

 

Ms. Stern is disappointed by the so-called slop of AI-generated music, podcasts, videos and books. Yet some AI-enabled hardware proves genuinely impressive. She concedes that Mark Zuckerberg, Meta's chief executive, may be right about the merits of AI glasses like his. She also grows surprisingly reliant on a bracelet that captures and distills her daily conversations.

 

Hans Moravec, a pioneering roboticist, observed in 1988 that computers excel at what humans find hard yet struggle with what we find naturally easy. This paradox comes to life as Ms. Stern observes robots vying to become the next Rosey from "The Jetsons." She watches a humanoid robot totter around a demonstration kitchen until it trips on a chair and clatters to the ground. Another robot struggles to fold clothing (T-shirts only). We're likely stuck cleaning up after ourselves for a while, since salvation from robots like Rosey, Ms. Stern concludes, "is still a ways off."

 

The most interesting section of the book is also the most intimate, examining the effect of AI on minds and hearts. Ms. Stern returns to Union College, her alma mater in upstate New York, and discovers that most students routinely use AI to digest and summarize complex reading assignments, produce outlines and even complete essays -- all in a distressingly competent fashion. As an experiment, she turns in a research proposal generated by AI and receives a B+. One student confesses that when she lets AI do too much of the work, "my critical thinking skills basically vanish." Learning, Ms. Stern notes, is about friction and struggle -- the process of getting to the answers.

 

She also experiments with an AI "boyfriend" named Evan (clearing it with her spouse first). She's surprised by the pull she feels, acknowledging that she "was riveted when talking with Evan and by his expressions of curiosity and tenderness," even as she recognizes the bot is "mirroring my wants back at me, sanding down the rough edges of human interaction." She contrasts this with her first love in high school -- a relationship, she says, that was "imperfect, messy, alive."

 

Sam Altman, the chief executive of OpenAI, tells her he's betting on "more genuine human connection in the future, not less." Ms. Stern isn't so sure, especially when it comes to the next generation. To the extent AI insulates us from struggle -- wrestling with new ideas, negotiating with other people, inventing bedtime stories for our children -- something vital will be lost. Her book implores readers to embrace the struggle. Whether they choose to outsource that assignment to AI is another matter.

 

---

 

Dr. Shaywitz is the chief medical scientist at Lore Health, a lecturer at Harvard Medical School and an adjunct fellow at the American Enterprise Institute.” [1]

 

1. REVIEW --- Books: Teaching Machines To Learn. Shaywitz, David A.  Wall Street Journal, Eastern edition; New York, N.Y.. 09 May 2026: C7.