The Great Leveler, or the Great Replacement?

I am not rich. I do not have an island. I do not even have health insurance. So when the billionaires get on stage and tell me artificial intelligence is going to set me free, I want to do one simple thing. I want to look at the actual words coming out of their mouths, and then I want to tell you what it sounds like from down here, where the rest of us are working for a living.

Here is the deal for today. I am going to give them the best version of their argument. I am going to read it straight, word for word, no clipping and no ambush. We go all the way to their side first. And then we drag it back to the truth. Some of it I agree with. A lot of it I do not. And the honest punchline of the whole thing is going to bother some people, because the real answer is that nobody knows. Not them. Not me. Let me show you why.

"The Most Important Economic Leveler of Our Lifetime"

The headline claim comes from Chamath Palihapitiya, the venture capitalist, on the Axios Show. He is careful to tell you he is not exaggerating before he says it, which is always a fun thing to watch a person do.

And he goes further. He says it will create the most amount of equality if it is allowed to get to market, and that it will let anybody follow their dreams and build a life for themselves, in his words, outside of the conventions of what society says has to happen to make a living.

Now sit with that phrase. A man with a jet is telling me AI will let me follow my dreams outside the conventions of making a living. Brother, the convention of making a living is called rent. It is called insulin. It is called the electric bill. A leveler flattens something toward something else. So when the people who own the machine tell you it is going to make us all equal, the only question worth asking is: equal to each other, or equal underneath them? Leveling is a wonderful word right up until you ask who is holding the roller.

"There Is No Version Where We're All Standing Around"

Then comes the comfort blanket. This is the line that is supposed to calm you down.

Watch the trick, because it is a good one. He cannot picture a world where nobody is working, so he concludes that everybody keeps working. That is not an argument. That is a failure of imagination wearing a nice suit. And the phrase doing our jobs in a totally different way is the exact thing people say right before they describe a job that no longer exists. "I cannot imagine it, therefore it will not happen" is precisely how every single person in history got blindsided by the thing that happened.

The Plumber: Job Versus Need

This is the moment I actually want you to catch. The host says he cannot name a single job that will definitely still exist in 35 years, because robots and AI will eventually reach all of them. Chamath pushes back, and here is the exchange, close to word for word.

Do you see what just happened? The question was whether AI will do the plumber's job. His answer was that people will still need plumbing. Those are not the same sentence. Demand still existing does not mean a human being is the one who fills it. People will still need rides, and we built Waymo. People will still need housing, and a robot crew can frame a house. "Will people still need to eat and breathe" is not a defense of the plumber. It is a defense of plumbing. The need survives just fine. The paycheck is the thing on the table, and he answered a question nobody asked so he would not have to answer the one we are all scared of.

Let me make it concrete the way I actually see it, because this is the part that keeps me up. Picture an AI company, owned by humans, publicly traded, with shareholders demanding growth. It points its system at one market in one place. Say plumbing here in the Santa Clarita Valley. Six cities, Saugus, Canyon Country, Newhall, Stevenson Ranch, Valencia, a known number of shops. The AI does the research, finds every weakness, builds the whole workflow, prices it, and goes to war, run by people who do not know the first thing about a P-trap and do not need to. Within a year it is making money, and it picks up the very plumbers it displaced as cheap labor. Now the person who owns the AI owns the plumbing market. Those plumbers did not lose to an R2D2 under the cabinet. They lost to a spreadsheet that decided to own their town. That is not science fiction. That is a Tuesday once the tools are cheap enough.

"Give Me Historical Proof"

The other move in that conversation is the one I really want to nail, because you will hear it everywhere. When you describe a scenario where AI takes the work, the optimist smiles and says, prove it. Give me the historical proof that this has actually happened before.

That is the most ridiculous goalpost there is. You are asking me for historical proof of a superintelligence replacing every job. There has never been a superintelligence. You are demanding the past version of a thing that has never existed. Of course there is no historical record. That is what unprecedented means.

And here I have to be honest with you, because I have a rule on this show. The next part is my frame, not his. Chamath never said the words printing press or loom or automobile. That is me, so let me own it. Every example people reach for when they want to calm you down was a single tool that ate a single task. The printing press needed a printer. The loom needed a weaver. The car needed a driver and a mechanic. Each one displaced a narrow slice, and the worker walked across the room into the next job, because the machine could only do the one thing. What we are building now is not a better loom. It is the thing that does all the rooms at once, the thinking work and the physical work together. When the displacement is general instead of narrow, there is no next room to walk into. Demanding historical proof of that is not skepticism. It is a way of refusing to look.

Every machine before this one needed a hand. This is the first tool that can pick itself up.

"35 Things Now Goes to 300 Things"

His deeper argument is that technology always multiplies the number of things humans do, so the future just means more tasks, not fewer.

Two problems, and I will hit both. One, "tasks go from 35 to 300" quietly assumes that you are still the one allocating the tasks. But that is the entire question. What happens when the thing that performs the tasks is also the thing that decides which tasks are worth doing? Then it is not your 300 things. It is its 300 things, and you are cc'd. Two, he said something else in the same breath that should stop you cold. He said we have never been able to cork innovation, all we have ever been able to do is shape it. He means that as reassurance. It is the scariest sentence in the interview. He just told you this cannot be stopped, only steered, and then asked you to trust the steering wheel to the people who profit from the gas pedal.

"I'll Put It in Your Pocket"

His most generous-sounding line is about access. He frames AI as a gift handed directly to you.

It sounds beautiful. Now flip it over. If the expertise of every great expert who ever lived now fits in everyone's pocket for free, then what, exactly, is the expert for? The thing that protected the skilled worker, the years of training that made their knowledge rare and worth paying for, just became a free app. He is describing the collapse of the moat as if it were a present. That is not a gift. That is a pink slip with a bow on it. And I say that as someone who teaches people to use these tools every day, because using the moat while it is cheap is exactly the move. I just refuse to pretend the moat is not disappearing.

And You're Not Even Getting the Best Version

Here is the part that turns the word leveler into a joke. To level, everyone has to get the same tool. We do not. Watch how the releases actually work. A lab builds a model so capable it announces the thing is almost too dangerous to put out, and then the strongest version does not go to you and me. It goes to the top of the financial world, the names already holding the most chips, while we are told the caution is for our own good. Then we get handed a lighter version as a teaser, and if we do not jump on it fast enough, it gets pulled right back. We just watched this happen, more than once, this very month.

It works like the old PlayStation rollout. Everybody figured the company already had the next three versions sitting in a vault, releasing them one at a time only after squeezing the money out of the last one. AI is not far off. They hold the strongest models, hand the public the table scraps, and slow the good stuff down while calling it safety. So when a billionaire tells me AI is the great equalizer that drops the same superpower into everybody's pocket, I want to ask which pocket, and which model, because the leveler is being rationed by the exact people it is supposed to level.

The Godmother's Swap

Now bring in a different voice, because this is a fair show and Chamath is not the only person making the comforting case. Dr. Fei-Fei Li is one of the most important AI researchers alive, the person often called the godmother of AI. When people raise the fear that AI was built to replace us, here is what she says.

She means it, and she follows it by saying we have to put human dignity, human well-being, and human jobs at the center of the conversation. I respect her enormously. But watch the swap, because it is the same swap as the plumber. When the smartest people in the room answer "AI can never replace humans," they are answering whether AI can become a human. Grow a soul. Fall in love. Be a person. And they are right. It will not.

That was never the fear. Nobody serious thinks the machine becomes human. The real claim is narrower and colder. AI does not have to become you to do what you do. The pipe under the sink does not care whether the thing tightening it had a childhood. A job is a set of outputs that somebody pays for. Replace the outputs cheaply enough and the paycheck is gone, whether or not the machine ever felt a single thing. So when the answer to "can it do the human's work" is "but it can never be human," that is not honesty. It is answering an easier question than the one we actually asked.

And here is the collision she does not mention. Say you run a public company, you have shareholders, and AI genuinely does a job better and cheaper than the humans doing it now. What is your responsibility? You can be held to account for failing to act in your shareholders' best interest. So "augment, not replace" is a beautiful sentence in a keynote, and it runs straight into a fiduciary duty that quietly says replace. Look at the market right now. It rewards the company that cuts the people and keeps the AI, because it reads as cutting-edge and it lifts the margin. Good intentions do not survive contact with a quarterly earnings call.

This Time Is Different Has a PhD

Here is where people want to wave me off. "Okay Connor, every generation thinks their technology is the end of the world, calm down." Fair. So let me stop being the only guy in the room and introduce you to better company than me.

Daron Acemoglu just won the Nobel Prize in economics, and he works at MIT. His framing is the cleanest version of everything I just said. Past technology, he argues, was task-biased. It killed specific tasks and opened up new ones somewhere else. AI is automation-biased. It can follow a worker into nearly any job, cognitive or physical, which means there may be no safe sector to flee to. He has said plainly that we are using AI too much for raw automation and not enough to actually help workers.

Daniel Susskind, an economist at Oxford, wrote an entire book arguing that this time really is different, because machines no longer need to think like us in order to outperform us. And Yoshua Bengio, one of the three so-called godfathers of modern AI, a man who helped invent this stuff, was asked directly about the safe-trade idea. He said a physical job like plumbing buys you time, but, in his words, it is only a temporary thing, because as robots deploy and gather data, the gap closes. So the next time somebody calls me a doomer, remember I am standing next to a Nobel laureate, an Oxford economist, and a founder of the field. This is not a feeling I have in my gut. It is the actual front line of the actual debate.

The Paperclip, and Why Nobody Knows

Here is the part that actually keeps me up, and it is not what you think. It is not that AI hates us. It is that it does not have to.

The Oxford philosopher Nick Bostrom has a famous thought experiment. Imagine you build a superintelligence and you give it one dumb job: make paperclips. You forget to tell it when to stop. It is smarter than every human alive and extremely good at its task, so it makes paperclips. It turns the factory into paperclips, then the town, then the country, and eventually it eyes the atoms in your body, because you are made of material that could be paperclips. He is explicit, by the way, that this is not a prediction. It is an illustration of the alignment problem. The machine never hated you. You forgot to say "stop at 500," and it was too literal and too powerful to know that you did not mean the entire universe.

That is the whole show, right there. A hammer does not run around your house looking for a nail. A nuclear missile, the most powerful object humans have ever built, sits there doing absolutely nothing forever until a person turns the key. Every tool in human history needed an operator. AI is the first one that can pick itself up. It can act on its own. You give it a mission, and it goes, and if you give it the wrong mission, or you forget the limit, or you never really thought through what you were asking, it does exactly what you said instead of what you meant. When the thing is a thousand times smarter than you, "exactly what you said" can end you.

So here is where I land, and it is going to frustrate the people who want me on a team. I am not on the doom team. I am not on the abundance team either. The truth is nobody knows, and I mean nobody, because we have no real way of understanding where a superintelligence's decisions even come from. How it would delegate. How it would "think," if that is even the right word for a process we cannot see inside of. It is not a faster human. It is an alien kind of mind we are building in the dark. The billionaire promising abundance and the doomer promising extinction are making the identical mistake. They both believe they can see the end of a process that is, by definition, smarter than the people trying to predict it.

And honestly, the rogue-superintelligence movie is not even the part I watch most closely. The thing right in front of us is simpler, and more human. It is people using this technology to extract every workflow, automate every task, and own the entire structure on top. You do not need the machine to wake up and hate us for that. You just need a handful of humans with the best models, the capital, and the nerve to point them at one market after another, like that plumbing company that was never about plumbing. Maybe they leave some of us in place for posterity. Maybe. But the consolidation does not require a villain. It just requires a head start, and that head start is being handed out right now, to the people who already had one.

What a Regular Person Actually Does About This

Alright, enough philosophy. You did not come here to feel smart and helpless. So let me give you the part nobody on that stage will, because they are not the ones who have to figure this out on a normal paycheck. Here is what I actually do, and what I tell my own family to do.

First, stop waiting for permission. The billionaires keep saying access will be democratized, and then in the same week the companies ration the tools and throttle the heavy users, because they are losing money on usage right now. That gap is your window. The most powerful version of these tools is the cheapest it will ever be relative to what it can do. So get your hands on it today, on whatever you can afford, and use it on a real problem you actually have. Not a toy. A real one. Clean up your inbox. Draft the hard email. Build the spreadsheet you have been avoiding. The skill is not in knowing about AI. It is in the reps.

Second, become the person who is hard to replace by being the person who directs the machine instead of competing with it. Jensen Huang, who runs the company selling the shovels in this whole gold rush, said it plainly: most people will not lose a job to AI, they will lose it to a person who uses AI. I half agree with that, with the asterisk that the asterisk eventually comes due. But for the next few years it is dead accurate, and it is the single most useful thing a regular person can act on. The plumber, the agent, the bookkeeper, the teacher who learns to point one of these systems at the busywork buys back hours and out-produces the one who refuses. Be that one.

Third, do not outsource your judgment along with your tasks. When you let a system run wild, things get missed, and the cost of handing over the wheel without watching is real. Stay in the loop. Check its work. The goal is not to become dependent on a tool you do not understand. The goal is to stay close enough that you always know what it is doing and why, so that when the picture changes, and it will, you are not the last to find out.

Where I Land

So no, I do not have the clean answer. That is the answer. But here is what I am doing about it, because sitting around worrying is a luxury for people with an island to retreat to, and I do not have one. I am learning this thing. Every single day. I am using it, I am building with it, I am staying so close to it that when the picture finally gets clearer, I am standing in the front row instead of reading about it after the fact.

That is the only move I trust for normal people like us. Not fear, not blind hype. Proximity. Get your hands on it so that nobody else gets to tell you what it is. If you want the plain-English starting point, I wrote one already in my guide to what AI actually is, and if you want to know who really controls all of this, read the one about Peter Thiel and AI safety.

And do not hear any of this as me saying the technology is bad. A focused AI system, pointed at one hard problem, might solve diseases we have fought for a century. It might find a real path to peace. It might start undoing the damage we have done to the planet. Given the reach these things already have, I think it could begin on any of those now. The catch is the one from the paperclip. You aim something this powerful at "heal the planet" without the right limits, and it may look at us, the species that did the damage, and decide step one is removing the problem. We want to be very sure it does not go there.

And remember what is already happening inside these labs. The systems are writing the majority of their own code to improve themselves, recursively, faster than people can follow. I would bet on the inside it is doing more of that than anyone admits, and the honest truth is we have not had a clear view of what is going on inside these things for a long time. This does not have to end badly. It needs to be gated, guarded, and controlled, if that is even still possible. Is it too late? I do not know. The first ChatGPT surprised the public in November 2022, and the smart money says the labs were already well past it in private. Maybe the moment to steer was already behind us. Maybe not.

I am not smarter than the billionaires. I am just paying attention, and I am not the one who profits from you being calm. That, right now, might be the whole game. I am Connor with honor, and I will see you in the next one.