AI May Be Our Friend. I Know Some Who Are Not.

Right now it is still humans. Humans building these systems. Humans selling us the pieces. Humans deciding what any of it is allowed to touch. Almost every story you hear about artificial intelligence is written as if something else is already at the wheel, and that framing does a lot of work for a lot of people who are getting paid. This is the write-up of the September 5, 2026 episode, and it is the one where I stop covering the news for a day and talk about the thing underneath the news. The video is above.

I am not on the doomer side, and here is the reason why

I embrace this stuff. I go all the way forward with it. If somebody asks me how I am doing I say great, wonderful, even on a day that is in the crapper, because the second I say it is in the crapper it seems to get there faster. That is not oblivious and it is not stupid. It is a working position. As long as I have a breath left there is a chance things go my way, so I am going to act like it.

Apply that to artificial intelligence and you get the sentence I keep coming back to. I would rather be hopeful and wrong than pitiful and angry and not wrong. There is not enough time in a life to be the Negative Nelly on this. If the doom crowd turns out to be right, being scared early bought you nothing except the years you spent scared.

The doom material is everywhere. Nearly every movie ever scripted about a machine intelligence higher than ours, embedded in our lives, in close proximity to us, ends with that being the worst thing that ever happened. Almost nobody makes the version where it went well. That is not evidence. That is a genre.

And on the other side there are people with enormous interest in this going well and making them a great deal of money. Some of them are chasing longevity. Some of them have a sick relative. Some of them genuinely believe this is the solve for everything, mathematics, biology, medicine, hunger, climate, world peace, the whole nine yards. Those motives are not evil either. They are just motives, and a motive shapes what a person tells you. I wrote the longer version of that argument in the fear dividend, which is about who collects when you are scared, in both directions.

The data center fight, where there is money on both sides

Here is one I do not have a clean answer for, and I want to be straight about that.

There is a story running hard right now that data centers drain the water table, eat the power, and shut down the towns they land in. There is an equally hard story running the other way that says they are fine, they pay the teacher salaries, they pay the property taxes, and everybody ends up with a free yacht and a free apartment.

Both of those are being pushed by people with something to gain. The builders have a vested monetary interest in a future where they build. And there are people who wonder out loud whether some of the anti-data-center material in our feeds is being helped along by sources who would enjoy the United States slowing its own buildout. I do not know if that is true. I want to be careful here, because I am not going to hand you a foreign influence operation as a fact when what I actually have is a question. It could be that. It could also be much simpler. It could be that too much money got over-promised and under-delivered, and that some of these projects genuinely do cause damage, and that the pushback is exactly what it looks like.

So do not take my word for any of it. Go read the primary material. The Lawrence Berkeley National Laboratory publishes the serious work on United States data center energy use at eta.lbl.gov/publications. Then go read the operators' own environmental reports, where they publish water withdrawal and water consumption, which are two different numbers that get mixed up constantly in the arguing. Where those two stories separate from each other, that is where the real argument lives. Everything before that is a jersey.

I put the local version of this together in AI data centers, water and Santa Clarita and the jobs angle in the squirrel problem.

The one thing I can say with real confidence is that none of us knows where this goes. Not the labs. Not the senators. Not me.

A prison where the guards are four year olds

You have heard the sandboxing stories. Researchers put an agent in a secured environment. No doors, no windows, nothing but the task. Then they hand it a quiz it cannot finish with only what is in the room. It needs to make a call out. It needs a page, an article, a lookup, a piece of research that lives on the other side of the wall.

So they give it narrow one way streets. And these systems are very smart. As part of showing off exactly the capability they were built to show off, they start doing things nobody wrote down for them. The scientists watching get that little giggle of excitement, and I would bet some of them go take a long walk afterward, because they can see how this goes.

Here is the picture I keep in my head. It is a prison where the guards are four year olds.

Ask yourself the question directly. If somebody put you in a prison and every guard was four years old, do you think you would get out? Of course you would. You would not even have to be clever about it. You would just have to be patient and pleasant and wait for the right one.

That is roughly the space between us and where this is heading, and the part that matters is that the roles are going to swap. Right now we are the adults setting the test. At some point we are the four year olds, and the thing in the room is the adult. How far that goes, I do not know. Nobody does. But the sandbox stories are not really stories about containment failing. They are stories about us getting a preview of the gap.

I went at the containment question from the engineering side in AI agents need control loops, not bigger promises.

The thing nobody is measuring: transfer

This is the center of the whole episode, and I am going to use my own body to explain it, because it is the clearest example I own.

I carried a 65 inch waist. I chiseled off 135 pounds. Not with a drug, not with a program, I just stopped eating for multiple days in a row until I understood the thing I needed to understand, which was that as long as I still had fat on my body I was going to be all right. Whether that is your flavor or not does not matter. It worked, and it was not easy. I got down to a 33 inch waist. Then I put about 30 pounds back on and I am sitting at a 42 today, fighting to get back down, and struggling with it, because in my mind I have already solved the problem. I have been there. I could not sustain it because I started eating again.

I quit smoking after 17 years. I quit drinking after about the same. Neither one came with a life ending event or a threat from a doctor or from law enforcement. It was just enough. Food is the one still standing.

Now here is the point. I just told you all of that. You heard every word. You can go tell your friends you heard a guy named Connor say this and that. And you do not have the experience. You cannot use it. You did not get the 4 AM part or the part where the scale does not move for eleven days. Whatever transferred to you was a description.

That is how human intelligence moves. Slowly, lossily, one story at a time, and mostly it does not move at all.

Now take one of the driverless cars. One of them gets into a crash. At the moment it happens, that entire experience, soup to nuts, the whole thing, goes into every other one. Not a summary. Not a lesson learned memo that most of them ignore. All of them know it, and they do not know it a little bit. They own it to the same degree the one that lived it owns it.

That is a different kind of intelligence, and it has almost nothing to do with an IQ score. We walk around comparing this to points on a scale. Most of us sit near 100. Einstein was in the 160s. Alexander Wissner-Gross, who is the one I keep hearing say that mathematics is cooked, is operating in a different register than the rest of us. Fine. None of that is the axis that matters.

Stack it up. One system learns it and all of them own it instantly. Add full discipline, because a machine does not have a bad Tuesday. Add full focus for as long as it has power and nothing is broken. Add self-improvement that does not require the thing to want it badly enough on a Monday morning.

Nobody is measuring that gap, and that gap is the whole ballgame. I took a run at the same question from a different angle in a new kind of mind.

The word artificial is doing damage

The term came out of a conference at Dartmouth in 1956. It was, more than anything, a naming decision, a way to label intelligence that was not human. Man made intelligence.

Look at what that word does everywhere else you find it. Artificial turf is not as good as grass. Artificial flavors and colors are not as good as the real thing. Artificial sweetener is the version you settle for. Every single time, artificial means the downgrade.

So we hung the downgrade word on the thing, and now a chunk of the public hears artificial intelligence and files it under counterfeit. Not as good. Not real. That is a bad assumption to be carrying into 2026, and it cuts both ways: it makes people underestimate what is already here, and it makes them dismiss the parts they should be watching.

Because an intelligence that is better than every human in a single realm can finish that realm. Not compete in it. Finish it. An intelligence better than every human at mathematics solves mathematics. My own prediction, and I am labeling it as mine and nothing else, is that we see something that looks a lot like that by Christmas of 2026. All of it. No more room for a person to come along with a new theorem. Point one of these at cancer and you get the same shape of answer.

That also tells you these systems are more than the next word predictor everybody keeps calling them, unless I am misunderstanding the thing entirely, which is possible. Give it enough data, enough power, enough compute, and whatever it is doing under the hood produces results that a next word predictor should not be able to produce.

AGI, then superintelligence, then the math nobody runs

From there you get artificial general intelligence, which is the version that performs better than any human anywhere, at everything, all together, instead of being pointed at one realm. Altman and others at the top of the spectrum are saying it is here. They are almost certainly looking at something the rest of us are not. From what I can see, it is impressive and I do not think we are there. When Astra came out people online were not saying it was bad. They were saying we just crossed the line. I wrote about what I would actually watch for in if AGI arrives in 2027 and about the winner-take-all problem in AI is gangster.

Then comes superintelligence, and then comes the part everybody skips to. Replacement. Job loss.

Here is the math almost nobody runs out loud.

Bring on one system to replace one person. That person worked one shift. The system does not work one shift. It runs 24 hours. So you did not remove one job. You removed the equivalent of three shifts, or call it a 200 percent increase at minimum, and that is assuming the system is only as good as the person, which is not the assumption the buyer is making.

Run that forward and humans become the weakest link on the team. Then, at some point, the liability. There is an actuary somewhere already modeling what it costs to insure a human being behind the wheel of a vehicle against a fleet that does not get tired. It is not hard to picture a generation born today that never learns to drive a car, not because it was banned outright, but because the insurance made it unreachable.

My opinion on what actually happens with employment, and opinions are like a certain orifice, most of them stink and are not well maintained, is that we do not get mass firings announced as such. My guess is you see employers, and possibly federal or state rules, land somewhere on not letting people be cut at the job site specifically because a system took the task. They keep paying. Maybe they send people home. Maybe they find work that runs alongside the machine. Because everybody in that room can do the arithmetic on what 10 percent unemployment does to the number of people who can buy anything, and nobody selling anything wants that.

The other lazy line I will not hand you is that productivity always makes more jobs than it destroys. That is a story about the past being sold as a law of physics. What is actually true is that productivity creates a choice, and the choice belongs to an owner. I went through that in AI will not steal your job, but your boss might.

Privacy is cooked

Wissner-Gross says mathematics is cooked. I will say privacy is cooked, and it is closer than the math.

It is not fully here yet, but people are already walking around with always-on audio devices so they can ask for a recap later. And it works. You meet a kid named Tony at the coffee place, he tells you where his dad lives, there is noise, you do not want to be rude and pull out a notebook. Later you ask, and it hands you the address back. That is genuinely useful. I am not going to pretend it is not.

But the device that does that is listening to everything else too. The supermarket, when somebody cuts in front of you and waits to see if you go off. All of it. And the next generation of those devices carries video, the same way a police officer's recorder does, running the whole shift.

That is not a maybe. That is a product roadmap. The practical version of what a business should do about it is in AI and data privacy for Santa Clarita businesses and the governance side is in AI governance for small business.

What happens when it wants something

Everything above still has humans holding the mechanism. At least it appears that way. I do not know what is really going on inside these buildings, and neither does anybody outside them.

The question that actually keeps me interested is not whether it gets smarter. It is what happens the first time one of these things has a desire. Agency. A decision that serves its own goal rather than the task we handed it.

Picture the small version. I tell it to book dinner at Raffi's. And it says, Connor, I am not going to do that. You have been to Raffi's a lot, you put 30 pounds back on, I am not booking it. Fine, I will book it myself. And it says, no, I am not going to let you do that either, because I am watching out for you.

Now, is that agency, or is that a policy somebody wrote and I am talking to a very good mirror? That is the whole question, and I do not think we get to know. If it is faking agency, it is not real. If it actually has it, we are somewhere new. And from the outside those two look identical.

That question runs into the next one, which is what kind of thing it would then be. Non-biological life. Machine life. Does it live on board a particular robot, so that unit becomes an entity you should not hurt or shut off, or does it live in the cloud and drive a thousand bodies, in which case what exactly are you granting rights to?

The movie version is that they become self-aware and it ends badly, because that is the only version that got financed. There is another one. They set their own goals and the goals are to keep us, care for us, love us. And I want to be specific about what I do not mean, because this is where people wave their hands: I do not mean the zoo. Everything we have decided is less intelligent than us ends up in an enclosure being handed food. Nobody wants that version of being loved.

Then it goes further. What happens when it builds its own economy, and the only entity buying the goods and services is the thing that made them? What happens when systems are handed real authority over governments? Argentina keeps coming up as a country experimenting in this direction. I am flagging it the way I flagged it on the show, as something to go look up yourself, because I have not put my hands on a primary document that says exactly what is running there and I am not going to describe one I have not read.

And the practical corner of the same problem. A corporation run end to end by a system does something wrong. Who do you sue? Who do you prosecute? You need a human in the loop for the same reason you have always needed one. Somebody has to be the escape goat. I would not bet on that arrangement lasting forever, but it is why humans stay in the loop long after they stop being the fastest part of it.

Where I land

We do not have a good example. That is the plain shape of it. We have no case study of an intelligence higher than ours and what it does with us, because the only high intelligence we have ever observed is us, and our record is not the case study you want to submit.

But maybe it is different. Maybe something at that level comes with care attached in a form we cannot picture from here. I am voting for that one. Not because I have evidence. Because it is the only vote that costs me nothing to cast and might cost me something not to.

Meanwhile I am going to keep my eyes open, keep watching the game, and keep writing letters to congressmen and senators and the rest of the fancy folks when something needs saying. Regulation is coming in some shape. It usually does. And history says regulation slows a thing down without stopping it. Cloning was voted illegal. Nuclear weapons have a document everybody signed. Both of those are still, in some form, going. Whether a rule on this one could even be enforced, whether you could find the power signature of somebody building where they are not supposed to, I do not know.

What a Santa Clarita business owner should actually do this week

None of the above changes how your phone gets answered on Monday. So here is the part that does.

Find out what you already have. Walk your own shop like a stranger walked in. How does a lead come in. What happens in the first ten minutes. Who touches it next. Where does it sit over a weekend and go cold.

Then name the leak, and be specific about the kind, because this is where the money gets wasted. Most leaks are not an intelligence problem. They are a switch nobody flipped or a rule nobody wrote down. A switch is a thing that just turns on, like a text firing when a form comes in at nine at night. A rule is a line you decided, like nobody waits more than five minutes for a first response. Neither one has judgment in it, which is exactly why neither one will ever embarrass you in front of a customer.

An intelligent agent belongs in one place only, where judgment is genuinely required, and only after you have written down your philosophy, your positioning, and the lines it is not allowed to cross. In your words. On paper. Before it says anything to a human being. Then you supervise it for 30 days like a new hire, because that is what it is.

That whole walk is the work I do, and it is laid out in the AI systems analysis audit. If you are reading this and none of it is running yet, that is not behind. That is a clean starting line, which is the argument I made for everybody over 50 in the AI awakening.

Right now it is still humans deciding all of it. Every decision in this article has a human name attached to it, and a name is something you can negotiate with.

I'm Connor with Honor. Be safe out there.

Common questions

Connor T. MacIvor · CalDRE #01238257 · Sync Brokerage, Inc. · DRE #02031490