Edition #38: There Is No Robot-Proof Skill

A neuroscientist who trained AI on 122 million people explains the one thing it still can't replace — and it isn't a skill

By Mykel Salomon · 2026-09-19

<hr><h3>🔓 Opening Reflection</h3>Almost everyone's AI anxiety lives inside a single question: what can I do that a machine can't?My guest this week says that's the wrong question, and she's spent thirty years earning the right to say so.Dr. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 ember-view" href="https://www.linkedin.com/in/vivienneming/">Vivienne Ming</a> is a theoretical neuroscientist. The Atlantic once called her a "professional mad scientist." She was the chief scientist of one of the first companies ever to use AI in hiring, where she built machine-learning models trained on 122 million professionals to predict human potential. Then she turned that same science on the most human problems imaginable: her own son's diabetes, reuniting refugee children with their families, teaching autistic kids to read a smile. Her new book is called Robot-Proof: When Machines Have All the Answers, Build Better People.When leaders ask her "what's the one magic skill we should teach every kid," here's how it sounds to her: like a person with a trillion dollars asking which single stock to put all of it in and never touch for thirty years. Any advisor on earth would say the same thing, diversify, what are you talking about? There is no one magic skill. And here's the harder truth underneath it: there is no robot-proof skill at all. For any well-posed question, anything with a knowable right answer, AI already beats most of us, at about three cents on the dollar.So if the answer isn't a skill, what is it?Her answer is deceptively simple, and it's the whole reason I wanted this conversation: the answer is you. Not a technique you can be taught. The deeper, stranger, measurable qualities that make you you. Let me show you the science, because it's remarkable.<hr><h3>🔦 Signal of the Week</h3>Vivienne ran an experiment I haven't stopped thinking about.She had people, including bright Berkeley students, forecast real-world questions with genuine, unknowable answers (think: the price of oil six months out), the kind you can score against actual prediction markets. She gave them AI help. And AI working alone crushed the humans, the worst model on its worst question beat the best person.But then she looked at how people worked with the AI, and found three types.The Automators — the majority — just handed the question over and accepted whatever came back. Watch their brains on EEG and you see something chilling: gamma-band activity dropping by roughly 40%. They didn't just outsource the answer. They switched their minds off.The Validators used AI to confirm what they already believed, overruling it only when it disagreed with them. Here's the punchline that should stop every leader deploying "human-in-the-loop" AI: both of these groups performed worse with AI than they did without it. Human-in-the-loop was the worst-performing group of all.Then there was the smallest group — 5 to 10% — that she calls the Cyborgs. They argued with the model. They demanded counterarguments. They asked it to make the strongest case for positions they disagreed with. They treated it as an interlocutor, not an oracle. And they beat everyone — better than the best human, better than the best AI, rivaling the collective wisdom of an entire market of forecasters betting real money. When she read their transcripts, she often couldn't tell where the human ended and the machine began.Now the part that ties this whole newsletter together. What predicted who became a Cyborg? Not IQ. Not technical skill. Not even which AI model they used — the fancy frontier models and the small open-source ones washed out entirely. What predicted it was curiosity, fluid intelligence, intellectual humility, and perspective-taking. The exact same four qualities she found predicted career success across those 122 million people a decade ago — and that, measured in children, predict lifetime earnings and how long you'll live.Human capital, not AI benchmarks, predicts what you can do with AI.<hr><h3>🌍 In the Wild</h3><h3>1. "They know everything and understand nothing"</h3>When Vivienne first used an early large language model, it reminded her of her own PhD students. She'd ask a brilliant, world-expert grad student a hard question and get a fluent, confident, almost-right answer, one that quietly answered a slightly different question than the one she asked.Her line for it is perfect: her students "know everything and understand nothing." And so, she says, do the machines. Their knowledge and fluency are astonishing. Their understanding is zero, and, in her strong view as a computational neuroscientist, making them bigger won't change that. She uses these tools constantly, in her companies and her research. She just never confuses fluency for understanding.The tell she uses is one you can steal today: if the AI's answer would be virtually identical for anyone else asking a similar question, it isn't worth much on its own. Your job is to add the understanding, to take it somewhere no one else in the world would take it. Because, as she puts it, our real superpower isn't the right answer. It's the crazy, out-there answer that probably isn't right — but might be. And if it is, it changes everything.<h3>2. There's no magic skill — but there are measurable human qualities</h3>Back at that AI hiring company, Vivienne found something that should shock every recruiter alive: the skills people list on résumés and LinkedIn predicted almost nothing about who'd actually succeed, once you knew they were already a working professional. Most university degrees didn't predict much either. Everyone who studied Python knows Python.What did predict success were the deeper qualities, what she calls foundation skills, or meta-learning. And she's insistent on the language: she refuses to call them "soft skills," because she's a hard-numbers scientist and these things are measurable. Resilience. Curiosity. Fluid intelligence. Perspective-taking. Intellectual humility. Pull data from something like the UK Biobank and these qualities predict the outcomes people actually want at 65 , health, wealth, friendship, a longer life, far better than whether you can factorize a polynomial.If she had to choose, she said, she'd have every child take a course on curiosity over one more class on algebra. Because it predicts so much more.<h3>3. You can't lecture someone into curiosity — you have to change who they are</h3>Here's the catch. You cannot lecture a person into becoming curious, or resilient, or humble. A lecture never made anyone more curious. These qualities change through experience — through going through things that, over time, make you a genuinely different person. The goal of real education, Vivienne argues, isn't to deliver facts and algorithms. It's to make you a better version of yourself.And it can be done deliberately. She pointed to a controlled study where teachers of 8-to-10-year-olds were trained to do one thing differently: praise interesting questions instead of right answers. Within six weeks, those kids were asking more questions, scoring measurably higher on curiosity, and diving deeper into new material with no prompting at all. Her translation for the boardroom lands hard: how many leaders reward right answers instead of interesting questions? Who are you training the people around you to become?(A related warning she raised, which regular readers will recognize: fresh research shows the diversity of language and thought among people using AI is shrinking — everyone converging on the same words, the same metaphors, the same ideas. The middle of the curve, at the scale of a civilization. And one more field note for anyone who leads people: never fully trust who someone says they are — that's who they want you to think they are. Watch what they choose when it costs them something. That's who they actually are.)<h3>4. The literal cyborgs — and the ethics racing toward us</h3>Vivienne doesn't just mean "cyborg" as a metaphor. Her academic home is neuroprosthetics — machines wired directly into the nervous system. She built an AI that learned to hear inside the constraints of a cochlear implant, a model and a brain learning together, which she'll cheerfully claim made her the builder of the world's first real cyborg. There are AI-driven artificial pancreases now that, in some ways, outperform a biological one. And a recent Nature paper described a man with profound paralysis who, through an implanted brain-computer interface, can speak, type, and move a cursor — at home, on his own, transforming his life.Then she got provocative. There's a famous idea in psychology — "the magic number seven, plus or minus two" — the number of things you can hold in working memory at once. Working-memory span quietly predicts income, wealth, even how long you live. So she asks: what if we could build a neuroprosthetic that made the magic number twenty? The ethics are enormous (the rich would get it first). But flip it toward Alzheimer's, stroke, or a kid with a traumatic brain injury — someone knocked from a seven down to a five — and the question becomes gentler: could we give people back the life they already had? Her current company is working on exactly that. Why do you think Elon Musk has Neuralink, she asked, and Bryan Johnson has Kernel? This is not science fiction anymore.<h3>5. The most human thing she ever built</h3>I asked Vivienne, after thirty years, what's the most human thing she's ever built. She named two.The first was that AI for her son's diabetes — the first of its kind. And here's the twist: after she built it, she gave it away. No patents, no licenses. She went to every major device maker and taught them how to build it — and then taught their biggest competitors too. So on a hard day with her son, she gets to know that because this happened to her family, millions of people might one day be alive.The second was an AI to reunite orphaned refugee children with their extended families — a project she's fiercely proud of, though she says that chapter of her book ends on a darker, harder twist that taught her what actually matters to her. "No one thinks about AI in this context," she said. Everyone asks how to make businesses more efficient. She spent three decades asking a different question: how do we use AI to make ourselves more human?<hr><h3>💬 The Big Question</h3>Stop asking "what can I do that a machine can't?" There's no satisfying answer, and chasing it will only make you smaller.Ask instead: "Who am I willing to become?"And if you lead people, or raise them: are you rewarding right answers — or interesting questions? Are you building automators, or cyborgs?<hr><h3>🧠 A Small Exercise</h3>Five ways to practice being a Cyborg this week — at work, at home, anywhere.<ol><li>When AI hands you a good-enough answer, keep going. Ask for a better one. That's curiosity, and it's trainable.</li><li>When it tells you you're wrong, don't collapse and obey — and don't dismiss it. Ask: "What am I missing here?" That's intellectual humility.</li><li>Argue with it. Ask it to make the strongest possible case against what you believe. Then see what survives.</li><li>Praise one interesting question this week — your kid's, a teammate's, your own — the way you'd normally praise a right answer.</li><li>Notice one thing you did when it cost you something. Not what you'd post. What you actually chose. That's who you are.</li></ol><hr><h3>🎙️ This Week on The Human Protocol Podcast</h3>Robot-Proof: When Machines Have All the Answers, Build Better PeopleGuest: Dr. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 ember-view" href="https://www.linkedin.com/in/vivienneming/">Vivienne Ming</a> — theoretical neuroscientist, founder of The Human Trust and Socos Labs, former chief scientist at Gild, and author of Robot-ProofVivienne has spent thirty years doing the thing almost no one in AI talks about: using it to help actual human beings — treating diabetes, reuniting refugee families, helping autistic children learn emotion. She brings the rare combination of hard science and deep humanity.In this episode, you'll learn:<ul><li>Why "what can I do that a machine can't?" is the wrong question — and what to ask instead</li><li>The Automators, Validators, and Cyborgs — and why "human-in-the-loop" can be worse than no AI at all</li><li>The four measurable qualities that predict who thrives with AI (and who atrophies)</li><li>Why today's AI "knows everything and understands nothing"</li><li>Why you can't lecture someone into curiosity — and what actually builds it</li><li>Neuroprosthetics, working memory, and the coming ethics of cognitive enhancement</li><li>The most human thing she ever built — and why she gave it away</li></ul>🎧 Available now on <a target="_self" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 oeKsuVZibZVkrJDbiVLhrdCYZvAakDtemxMao " href="https://open.spotify.com/episode/6WWJdS7aXMeCccAriCKWXy?si=lOiIEkplR52BEz_znn1X_Q">Spotify</a>, <a target="_self" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 oeKsuVZibZVkrJDbiVLhrdCYZvAakDtemxMao " href="https://podcasts.apple.com/us/podcast/the-only-people-who-actually-beat-ai-dr-vivienne-ming/id1843402023?i=1000790022105">Apple Podcasts</a>, and <a target="_self" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 oeKsuVZibZVkrJDbiVLhrdCYZvAakDtemxMao " href="https://youtu.be/T-MqMgmVwso?si=bwVHJCq9DuYNLZZm">YouTube</a>.<img class="max-w-full h-auto rounded-md my-4" src="https://media.licdn.com/dms/image/v2/D4E12AQEznUdha00RKQ/article-inline_image-shrink_1000_1488/B4EaC5WYnzH8AM-/0/1789816048550?e=1791417600&v=beta&t=7ImQe13qsxyL3nMvyHFFlKQ9dSs-41q9XR9QBU8Ip1o" alt="Article content"><hr><h3>🧭 Thought to End the Week</h3>The entire public conversation about AI is stuck on one word: efficiency. How do we do the same things faster, cheaper, with fewer people?Vivienne spent thirty years on a different question, and it's the one this whole newsletter is built around: how do we use these machines to make ourselves more human?When machines have all the answers, the task was never to find the single skill they can't copy. There isn't one, and there's a strange freedom in admitting it. The task is to become the kind of person who brings something no machine can — the curiosity to keep looking, the humility to ask what you're missing, the resilience to sit in not-knowing, the wild idea that probably won't work but just might change everything.Those aren't soft skills. They're measurable, they're trainable, and they're the most durable investment you will ever make — in your kids, your teams, and yourself.When machines have all the answers, build better people.Start with the one you're becoming.The code may change. And it will.But our humanity is the constant.See you next week.— Mykel<hr><h3>📚 Sources</h3><ol><li>Vivienne Ming, Robot-Proof: When Machines Have All the Answers, Build Better People (Wiley, 2026) — the core thesis, the 122-million-person hiring research, and foundation/meta-learning skills.</li><li>Ming, forthcoming: "Human Capital, Not AI Benchmarks, Predicts Hybrid Intelligence in Forecasting" — the Automator/Validator/Cyborg experiment; ~40% drop in EEG gamma-band activity for Automators; the ~5–10% Cyborgs outperforming both humans and AI; curiosity, fluid intelligence, intellectual humility, and perspective-taking as predictors. Summaries via Fortune and Worth. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://fortune.com/2026/05/16/how-to-really-succeed-with-ai-cyborg-productivity-resistance-backlash-human-skills/">https://fortune.com/2026/05/16/how-to-really-succeed-with-ai-cyborg-productivity-resistance-backlash-human-skills/</a></li><li>Research on curiosity in classrooms (e.g., "Question asking practice fosters curiosity in young children," 2024) — praising questions rather than right answers measurably increases children's curiosity and deeper engagement. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://www.researchsquare.com/article/rs-4000469/v1">https://www.researchsquare.com/article/rs-4000469/v1</a></li><li>Doshi & Hauser (2024), Science Advances; and follow-on work — generative AI raises individual output but reduces the collective diversity of human language and thought. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://www.science.org/doi/10.1126/sciadv.adn5290">https://www.science.org/doi/10.1126/sciadv.adn5290</a></li><li>Nature (2025) — an intracranial brain-computer interface enabling a person with paralysis to generate speech, type, and control a cursor at home. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://www.nature.com/articles/d41586-025-02808-z">https://www.nature.com/articles/d41586-025-02808-z</a></li><li>Alimujiang et al. (2019), JAMA Network Open — a stronger sense of life purpose is associated with significantly lower all-cause mortality (relevant to Ming's forthcoming work on purpose). <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2734064">https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2734064</a></li><li>Miller (1956), "The Magic Number Seven, Plus or Minus Two" — the classic working-memory capacity finding Ming references. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://psychclassics.yorku.ca/Miller/">https://psychclassics.yorku.ca/Miller/</a></li></ol>