AI vs Human
- I’m certain Steve Jobs would be disgusted by the exploitation from social media of the greatest communication device ever invented. Jobs had his faults, but crassness was not among them.
Agency, How to Lose It
People are going to have to make themselves predictable, or the machines will get angry and kill them.
—Gregory Bateson
I’m walking here!
—Ratso Rizzo
You’re standing on a sidewalk adjacent to a busy street, and you’re late to your appointment that’s on the other side of the speeding traffic. You have a choice: you can hope for an opening through the multi-ton vehicles racing past, or you can walk to the end of the long urban block, push the walk button, and wait for the traffic-light gods to smile on you.
Jaywalking laws came about in the late 1920s from a hard push by the auto industry. Before jaywalking laws, it was the automobile, not pedestrians, that had to be predictable.
Move fast and break things
—Mark Zuckerberg, Facebook founder
When Zuckerberg was clawing his company’s way to the top, his “move fast and break things” mantra served its purpose. So it is with budding (and perhaps soon to die) artificial intelligence (AI). Move fast because we’re in a race with China. Break things because only AI can fix the things it breaks. (No, I don’t understand that logic either, but if you say it fast enough…)
What a computer is to me is that it’s the most remarkable tool that we’ve ever come up with, and it’s the equivalent of a bicycle for our minds.
—Steve Jobs
Steve Jobs was defining the computer as a personal agency device. Funny, isn’t it, that a few years later, Jobs would introduce the bicycle lock for our minds, our minds locked to our iPhones that are locked to pernicious social media? The promise of increased personal agency, from the personal computer, then from the Web, then from the computer in our pocket, turned out to be a Trojan Horse.
While we thought we were becoming more productive and more entertained, we were being fed algorithms that produced regenerative feedback loops—what we call addictions—by the huge corporations that dominate our eyes and ears. And now that we’re off-balance, AI is coming to finish the job of killing off personal agency for good. It will be worthwhile, we’re told because we’ll have AI servants that will make us dinner reservations.
While I’m skeptical that, as some fear, artificial intelligence will get angry and kill us, literally, I believe the real threat is that AI will pacify us to the point of cultural suicide.
Big tech spends billions to just show us more ads for products, but who’s going to have money to buy these products when AI takes over the jobs of the middle class? I’m a little confused.
Notes
AI is Coming For Our Mind Maps: If you think screens have been bad for us, just wait until AI arrives.
Vocabulary
AI: ideas and behavior that would be considered intelligent if done by a human.
AI agent: an artificial administrative assistant.
LLM (Large language model): AI software that predicts the next token. A token is a word or part of a word that the LLM uses to predict the next token. So a word or part of a word is fed to the LLM to predict the rest of the word, or another part of the word, and so on, which then predicts the next word in a phrase. LLM-trained AI is the model that Google, OpenAI, Anthropic, and nearly all AI companies are pushing and is just a super superset of the predictive text feature on your smartphone.
Siri: Apple’s built-in AI agent. On an Android device, the equivalent is Google Assistant.
Instead of trying to build a machine that thinks like a corporate intern, we should be trying to build a machine that learns like a four-year-old./ —Alison Gopnik
Yesterday, she made mud pies by stomping the mud with her bare feet and shaping the wet confection with her hands. But today, our not quite four-year-old granddaughter turned fastidious. She frowned at the bits of mud splattered on her little pink bike and dispatched one of her staff, Nana, to get a bucket of water, a sponge, and a drying cloth. After cleaning the bike and some further riding, she came in to wash her hands in our downstairs bathroom.
Nova is both young and tiny for her age, so we usually help her with the soap pump. When I peeked into the bathroom to see if she needed assistance, she was sitting on the floor, the soap dispenser in front of her atop the step stool. Using what little body weight she has, she perched over the pump to push out the liquid.
Her soap dispenser trick was her third new acquired skill I had witnessed in two days. Earlier, when riding her bike, after every few forward cycles of her bike wheels, she would reverse the pedals to brake. My earlier efforts to teach her to brake had gone nowhere, but she listens to Daddy, and now she was practicing and taking delight in her newfound control.
The third new skill I noticed was when she and Nana were making bookmarks with cardboard and color tape. Nana suggested that she (Nana) would cut the tape and Nova could place the tape on the cardboard strips. Before Nana could finish her suggestion, Nova was pressing the tape on the table with the end overhanging the table edge. Set that way, without help, Nova could cut the tape with her scissors. Her surprised grandmother asked Nova where she learned that?
“Preschool.”
These were not just three new skills but new skills learned in distinct ways: the soap pump trick came from experimentation on her own, the braking skill from coaching by her father, and the scissors cutting trick from modeling by her teacher.
From the day she was born, Nova, as every child, was accumulating mind maps of how to accomplish tasks, but not just isolated tasks. An observer may see one skill learned at a time, but most skills are interrelated. She was learning and learning to learn.
Nineteen years earlier, on the day she turned sixteen, Nova’s mother, Bria, made me beg the DMV tester to wait until I raced home and got her forgotten birth certificate. Bria was determined to drive herself to karate class that evening. Bria’s older brother, in contrast, showed little interest in driving until he was over 20. I was confused until I hit on the reason. Both Adam and Bria had learned the skills to direct a car through traffic and pass the driving test, but only Bria had acquired a mind map of where she was in relation to her destinations. Bria’s thoughts were, by her nature, out in the world. She was a mental cartographer of relationships and of places to meet the people who made up those relationships. In contrast, Adam had developed mind maps that fortified his inner world—maps of complex games played on a table and played on a screen, maps of programming computers, maps of the science texts and literature I sent his way as a homeschooled kid. At the time, he had just enough of a mind map of interpersonal relationships to share his gaming interests with a few close friends.
Adam’s mind maps, in contrast to those of his sister, did not include his environment. I made a guess that he didn’t drive because he didn’t know how to get anywhere. (This was before smartphones.) I bought a Garmin car navigation device and mounted it on our car’s dashboard. When he next returned from his college year in Minnesota, I was proved right when he announced, “I want to go places.”
Map making
The making of Nova’s new skills and the road navigation by Bria and Adam might appear to have little in common, but they’re much the same. Skill acquisition is done by acquiring a mind map by building on those already gained. When American explorers Lewis and Clark mapped the West, there were no flyovers to show the territory they would cover. They had to create their maps by noting the territory as they moved through it.
A concept is assembled like a jigsaw puzzle, as pieces that include clues about how they fit together. You might think you grasp an idea whole, but that’s because the puzzle pieces are submerged in your subconscious. When that puzzle is finally assembled in your mind, you have a new conceptual tool to apply to other challenges.
Process knowledge
This last year (2025), two noted business books that compared industry in China to industry in the U.S. were published: Apple in China, by Patrick McGee, and Breakneck: China’s Quest to Engineer the Future, by Dan Wang. While each book has a different emphasis, put together their overall premise is that Apple engineers taught Chinese engineers the rapid and high-quality manufacture of high-tech devices. And through years of refining these techniques, China has acquired process knowledge to the level that no other countries can compete.
Process knowledge is the means of turning engineers’ blueprints into goods. It takes decades of design and corrective feedback to attain that competence.
For a human being, the knowledge blueprints are inborn genetics plus instruction you get from texts, lectures, mentors, and teachers, but that which sticks comes from doing. Process knowledge is a businessy term for learning-by-doing, a concept that formally dates back to John Dewey but must have been implicitly understood at the beginning of familial and tribal guidance.
Learning by doing is the creation of mind maps that Nova illustrated at the top of this essay. It’s how Nova’s mother created her maps of her environment so she could navigate as a driver. It’s how Adam created maps of games and software development.
AI
I don’t need help
—Nova
A tiny person who uses her body weight to operate the soap dispenser is learning to be a problem solver. As long as Nova’s parents, grandparents, and other grownups are available to augment Nova’s (lack of) muscle development, she doesn’t need to invent a solution. Overly helpful adults can interfere with a child’s development of process knowledge. AI has the potential to obliterate it. Siri, dispense soap.
Most parents learn to back off as their children become more competent. They’re relieved that they have less to do and are proud their kids are maturing. AI has no experience with the development of process knowledge. It has no inner world to draw from that could make it sensitive to a child’s desire to be competent and viewed as competent.
The attraction of AI, especially agents, is that they’ll do the drudge work for us—make dinner reservations so we can focus on the important, the creative, the fun stuff. But most learning starts with drudge work, such as… well, everything. We start a task. If it’s long and repetitive, we invent shortcuts. Our creative flow yields better shortcuts. If AI replaces the shortcuts we would have invented so we can focus on creative work, then AI contradicts its own value.
The not-so-hated repetitive tasks
What exactly is AI going to give us relief from, and do we want relief from that?
Stardew Valley is the eleventh best-selling computer game of all time. Stardew Valley is, first of all, a farm simulation. The heart of the game is planting, watering, and harvesting crops—a task needed daily. Lots of other chores, events, and socializing are available in the game, but the repetition of farming is the game’s staple. Stardew Valley is just one farming-centric game among many. Unpacking) comprises unpacking your possessions after a move. Other popular games are designed around simple tasks, such as power washing, truck driving, running a coffee shop, or running a diner. But these are just the obvious titles that feature what sounds like tedious work when thought of in real life. With many games, likely most, your time is spent working to level-up so you can beat ever more dangerous foes. Does that sound like fun? Hint: it’s called grinding.
There are many forms of grinding in daily life that people enjoy, such as weeding their yard, knitting, home improvement, assembling puzzles, or practicing with repetition of an athletic skill.
What could be the attraction to these seemingly dull exercises? What these games, hobbies, and maintenance activities have in common is the feeling of controlling a finite area of life or the fantasy (with a game) of controlling a finite area of life. This suggests to us that many tasks taken over by AI will cause us to replace them with similar tasks, but as play. You’ll tell LLM-infested Siri to plan your vacation to Disney World and make the reservations for hotel, rental car, and restaurants. On the plane carrying you to Disney World, you’ll play a game on your phone where you manually plan your trip to Disney World and make reservations for hotel, rental car, and restaurants.
Look around your workspace, your kitchen, your toolbox. Simple mechanical gadgets such as can openers, scissors, water kettle, toaster, screwdriver, hammer—every one of them conceived and invented by trial and error, by an evolving process knowledge. Where will we get new process knowledge after we’ve offloaded problem-solving to AI? If we feed all the mechanical inventions of the past into our LLM database and tell our AI-infused 3D printer to create a desired gadget, without our involvement, without our tinkering, how will we then know how to improve our new gadget or what new gadget to create? Nova has already improved her soap pump trick by moving the dispenser from the step stool to the closed lid toilet; she gets more leverage by standing over the dispenser while pressing. Process knowledge.
If their parents are using AI, any thought that kids will be kept from it is absurd. We know that from past experience with technology. Busy and exhausted parents find relief in setting their little children in front of the television, or more commonly today, an iPad. What are the chances parents won’t offload their kids to AI? Robby the Robot will help you with your Lego puzzle or homework. Or teach you to brake your bicycle. And the motivated and smart older kids will route around limits such as age-gating and other what-about-the-children panic laws that never work.
As a culture, do we ever learn? After the modern miracle of the Internet and specifically social media, where we were sold the idea that we’d learn to better understand each other and respect our differences, we’re now going to buy yet another fantasy of an improved culture, an improved culture sold again by high-tech leaders who lack people skills.
I believe that AI in general, and agents in particular, will be worse than social media. AI will stifle learning in both children and adults. But even if I’m wrong about AI, that agents will destroy early process knowledge, there’ll be other negative consequences we can’t predict. There always are. When Tim Berners-Lee invented the World Wide Web, he envisioned low-friction collaboration among scholars. What he got is the opposite: domination by one storefront, one search engine, and a handful of social media platforms that thrive on grievance.
Unintended consequences are a myth. There are just consequences we can’t predict. Every significant technological advance brings forth changes we regret. The legal and cultural solutions to reverse the changes we don’t like do not send us back in time. They’re just further changes that often make the cure worse than the disease in an ongoing game of Wackamole. Undo works in only software.
Given the thought experiments about going back in time to kill baby Hitler, I bet many would like to go back in time and kill Facebook, Twitter, and maybe even the iPhone. Ten years from now, we’ll have wished we killed the AI that’s coming for our process knowledge when we had the chance.
AI
AI intends to take away our stepping through life to create our mind maps. That’s AI’s marketing gambit but reframed in a positive light. More likely, it will make us so stupid we’ll become increasingly dependent on AI. (Can you calculate without a calculator?) Besides making us dependent and stupid, chances are these agents won’t work well. Higher-cost meta-bots that correct the errors of standard bots will come. Then higher, higher-cost meta-meta bots. And so on. We’ll be stupid and broke.
Resources
There are many resources that question the value of AI. The following are just the ones most related to my essay.
My related articles:
Why the Metaverse Ain’t Gonna Happen
Habitability, part 1: Not Julia Child
Habitability, part 3: Unintended Consequences
Addiction is Just a Word (Explains cybernetic feedback.)
Books:
The Gardener and the Carpenter: What the New Science of Child Development Tells Us About the Relationship Between Parents and Children, Alison Gopnik
Seven and a Half Lessons About the Brain, Lisa Feldman Barrett