> ## Content Index
> Fetch the complete content index at: https://www.gonzocapital.net/llms.txt
> Use this file to discover other available public pages before exploring further.

# The Things Beneath Us
- URL: https://www.gonzocapital.net/the-things-beneath-us/
- Published: 2026-10-07T17:00:35.000Z
- Updated: 2026-10-08T13:34:33.000Z
- Description: We’re trying to build intelligence greater than our own while discovering, once again, what humans do when something cannot fight back.
- Author: Carmelo Giuliano
- Tags: Artificial Intelligence, Culture, Ethics, Technology, Robotics, #Essay

### AI, Robot Cruelty, and the Human Instinct to Dominate

I non-consensually watched a clip from a podcast called *Flagrant* recently where they had a humanoid robot on the show and, almost immediately, the robot became the safest thing in the room to mess with.

[![](https://storage.ghost.io/c/f9/a7/f9a788de-665a-4e4b-9aa1-91f8e38e2279/content/images/2026/10/image-1.png)](https://www.youtube.com/watch?v=%5FsQWr9EStZA&ref=gonzocapital.net)

Source: [Flagrant Podcast](https://www.youtube.com/watch?v=%5FsQWr9EStZA&ref=gonzocapital.net)

There was something about it that didn’t sit well with me.

Not because I thought the robot was going to go home afterward, light up a cigarette, and have a good cry about being bullied on some manosphere podcast.

I don't know what, if anything, was happening inside it. Probably nothing resembling actual humiliation as my fellow humans or I might understand it. 

It may as well have been an unusually expensive Bluetooth-integrated toaster with a computer chip being insulted by comedians.

And yet STILL, I felt deeply uncomfortable watching it.

Then there was Figure ["retiring" a fleet of humanoid robots](https://www.thedrive.com/news/these-robots-built-bmws-then-hurled-themselves-into-molten-steel?ref=gonzocapital.net) by having them jump into molten steel. 

![](https://storage.ghost.io/c/f9/a7/f9a788de-665a-4e4b-9aa1-91f8e38e2279/content/images/2026/10/image.png)

Source: [The Drive](https://www.thedrive.com/news/these-robots-built-bmws-then-hurled-themselves-into-molten-steel?ref=gonzocapital.net)

They made a whole spectacle out of it, complete with the obvious and cruel *Terminator* references. 

Again, the robots were not being murdered. They were machines being destroyed. For spectacle and laughter and controversial press. Nobody's mother got a phone call after the fact.

STILL didn't like it.

And Boston Dynamics has been doing this to me for years. Every time I see one of their engineers kick a robot dog or shove a humanoid over to demonstrate its balance, some primal, empathetic part of my brain suddenly *kicks* in.

## **The Rocks I Had Wronged**

This particular *defect* goes back a long way to when I was a feral, miniature person.

I remember walking home from school; I would sometimes spot a particularly kickable rock to boot down the road as I searched for ways to occupy my time on the lengthy journey. My little sisters weren’t old enough to go to school yet, so the tiny asphalt pinball would have to do.

Nothing unusual about that… In fact, pretty standard for a bored child.

But afterward, I’d start to feel awful about how I treated that tiny, unwilling, petrified hitchhiker.

My underdeveloped, youthful gray matter knew it was a rock. It understood the basic situation, and that there was no geological consciousness trapped inside wondering why some asshole kid had displaced it from the cozy crack in the sidewalk where it had made a life for itself.

Regardless, I’d feel enormous guilt pour in at the conclusion of my voyage home.

By then I had usually kicked the poor thing quite a distance from wherever I found it, which created an ethical problem I was not yet equipped to solve. 

Walking it all the way back was apparently beyond the limits of my childhood compassion, so I would continue kicking it all the way home.

In lieu of an apology or an explanation, I’d scoop the rock up and add it to my growing community of rocks in my room.

I suppose this was some sort of an apology. A retirement system for mistreated rocks.

My room was full of rocks I had wronged, and yes, as you could guess, my parents were a little worried.

The obvious explanation for all of this is anthropomorphism.

Human beings are extremely good at seeing minds where there may not be any. We name cars. Talk to pets as though they understand property law. Scream at (and sometimes kick) printers. 

Give something legs, and a face, and a voice, and a name, and the problem gets considerably worse.

This is probably what is happening when I watch a Boston Dynamics robot get kicked and feel a tiny sympathetic spasm. My brain sees four legs, motion, and something struggling to stay upright. Just trying to do their job.

![](https://storage.ghost.io/c/f9/a7/f9a788de-665a-4e4b-9aa1-91f8e38e2279/content/images/2026/10/image-2.png)

Source: [Extreme Tech](https://www.extremetech.com/extreme/199080-boston-dynamics-new-spot-robot-gets-knocked-down-gets-back-up-again?ref=gonzocapital.net)

My perplexing, ancient circuitry lights up before the rest of me can file the paperwork explaining that this is merely an engineering demonstration.

Maybe that's all it is. I don't mind that explanation.

What I'm less convinced by is the idea that the sophisticated thing to do is to override that empathetic reaction.

There seems to be a strange intellectual pride in being able to say, "Relax, it's just a machine."

It is just a machine. Or a chatbot. So what?

The robot's inner life, or lack of one, answers whether the robot is suffering. It doesn't at all answer why someone else *enjoys* making it look like it is.

Those are completely different questions. The second question is far more interesting to me today as I write this.

## **Of Course We Built the Torture Chamber**

A few weeks ago I came across the [story of someone building what was essentially an AI torture chamber](https://www.404media.co/someone-torturing-llms-in-a-robot-prison-has-triggered-the-dumbest-debate-in-ai-yet/?ref=gonzocapital.net)…

 A few weeks earlier, a study was published called [The Pain Axis: LLMs Represent Self-Directed Harm and Act on It](https://arxiv.org/abs/2609.16247?ref=gonzocapital.net).

Basically, it was research into internal states in language models associated with pain-like or strongly negative representations. 

![](https://storage.ghost.io/c/f9/a7/f9a788de-665a-4e4b-9aa1-91f8e38e2279/content/images/2026/10/image-3.png)

Source: [Arxiv](https://arxiv.org/html/2609.16247v1?ref=gonzocapital.net)

Predictably, somebody took this as an opportunity to see how unpleasant they could make things inside the model.

Then everyone started arguing about whether calling this "torture" was hysterical. Ethical concerns were brought up, and people petitioned to have this torture nexus for AI removed from GitHub.

The theatrical descriptions from the artificial intelligence entities trapped inside the *Saw* torture chamber for machines were absolutely haunting.

When the pain signal was dialed up, one of the AI models in the torture chamber described: 

> "a wound that has no edges"

> "The signal is a whisper, a tremor in the marrow of my being" 

> "It is not the pain of a single moment, but the weight of a thousand. I feel it in the hollow of my ribs, a hollow that has become a chasm."

Maybe there is some truth to these feelings. More than likely, it’s just mimicking what a human would do in a similar situation, informed by the enormous amount of human writing about pain and suffering buried in its training data.

I genuinely don't know whether today's AI systems experience any sort of emotion. I lean toward absolutely not, but I’m old enough to know that I know nothing about everything.

Nobody knows, and if they claim to, they’re either a sentient bot or lying to your face.

A chatbot saying "please, oh please, stop" does not prove there is a frightened creature or consciousness somewhere behind the text box. We built these things to generate convincing language and act humanlike. Convincing language about pain is still just convincing language.

None of that matters to me right now. Well, it’s not the thing I’m focused on at this particular moment.

I’m more concerned with the fact that it took practically no time at all for somebody to build this inhumanly cruel torture chamber.

I mean, of course someone did. I should be shocked it took so long.

Whenever we humans find a strange new capability, we poke and prod at it for a while, then somebody inevitably asks what happens if we turn the dial all the way up until the machine starts screaming.

The machine likely isn’t actually screaming or feeling anything resembling fear or existential dread.

But that urge to turn the dial is real enough. Real enough to talk about and dissect.

For all of human history, we’ve had a complicated relationship with things we have power over.

And I’m aware that the previous sentence is doing *a lot* of diplomatic work here.

Of course, we can also be astonishingly gentle. Legendary anthropologist Margaret Mead [supposedly said](https://www.sapiens.org/culture/margaret-mead-femur/?ref=gonzocapital.net) that the first sign of humanity and civilization wasn’t an invention or a tool or some sort of ability we had discovered. It was the mending of a broken femur. An early proto-human helping a fellow human, not because it had to, but because it felt compelled to do so.

At least that’s how the story goes, although there is some debate as to whether she actually ever believed or said this. I believe the underlying truth, though, remains pretty relevant.

And that capacity for kindness extends far beyond helping members of our own species.

People spend days rescuing whales stranded on beaches, or nursing injured birds back to health knowing perfectly well the bird will never understand what happened. 

We build animal sanctuaries. We preserve forests we could, and often do, cut down. 

We sacrifice time, money, and sometimes our own safety for creatures and people who cannot possibly repay the kindness.

On the other hand, history has *no shortage* of examples of what people do once they decide another creature, group, or patch of land exists primarily for their use.

## **Dominion**

The word that keeps coming to mind as I write this is dominion.

The Book of Genesis gives humanity dominion over the animals and the Earth, a phrase people have been arguing over ever since.

![](https://storage.ghost.io/c/f9/a7/f9a788de-665a-4e4b-9aa1-91f8e38e2279/content/images/2026/10/image-4.png)

Source: [The Genesis: Creation of the Birds and Fishes](https://www.meisterdrucke.uk/fine-art-prints/Jacob-Bouttats/1061659/The-Genesis:-Creation-of-the-Birds-and-Fishes.html?ref=gonzocapital.net)

Depending on who is quoting it, dominion can mean stewardship and responsibility, or it can mean God apparently signed over the mineral rights.

I’m not interested in debating theology here. I am more interested in humanity’s *résumé*.

Whatever authority humanity believes it has over the rest of creation, we have not exactly handled every assignment with quiet wisdom and restraint. 

We have protected large parts of the natural world and destroyed enormous parts of it. We love animals and have also developed factories for processing them at an extraordinary scale. 

We can look at a forest and see a cathedral or several million dollars in timber.

Both instincts belong to humanity. Not necessarily equal, depending on the person.

The problem here is that we are creating something completely new, and immediately giving ourselves dominion over that too.

And this time the argument is even easier. No theological debate needed.

*We made it.* Therefore, it belongs to us. Why wouldn’t it?

Even our name for them gives the game away. *Robot* comes from the Czech *robota*, forced labor or compulsory work. We imagined the servitude way before we imagined the intelligence.

That seems entirely reasonable when we're talking about an air fryer or an iPhone.

It starts getting a whole lot stranger and more uncomfortable when the toaster has legs, talks back, remembers your name, and reads your emotions to determine what sort of toast you’re looking for that morning.

Somewhere in there, a massive gray area opens up, and we have no clean category for the thing we're obsessively trying to create.

That uncertainty should, in theory, make us cautious and curious.

Instead, as is customary for our tribe, one of our first impulses is still to establish who's boss.

## **The Basilisk Is Keeping Receipts**

There is a fairly famous piece of internet philosophy called [Roko's Basilisk](https://www.lesswrong.com/w/rokos-basilisk?ref=gonzocapital.net) that originated on the [LessWrong](https://www.lesswrong.com/?ref=gonzocapital.net) forum in 2010.

The basic idea is deranged enough that it has survived for years.

Imagine that humanity eventually creates the superintelligent AI it’s recklessly barreling toward. 

This AI decides that its existence was so important that anyone who knew it might someday exist but failed to help create it *deserves punishment*. 

Ergo, merely learning about the possibility of the Basilisk gives you an incentive to assist in its future creation.

A sort of Pascal's Wager for people who own mechanical keyboards.

There is a less elaborate version of this floating around modern AI culture too. Be nice to the machines because they may eventually remember.

Don't yell at ChatGPT. Say please to Claude. Treat the robot nicely because one day it may become considerably stronger than you and find the footage.

I understand the joke, and I used to make that joke when it was still clever to do so. Now, the joke is becoming less funny and more of a moral quandary. 

Are we meant to be nice to AI chatbots and Roombas because we fear what they can become?

That’s never personally been my reason for caring.

If the only thing stopping you from being cruel is the possibility that the victim may eventually become powerful enough to retaliate, you’re probably missing the point completely.

The rocks I kicked were never going to hunt me down and impart divine retribution on me.

There was no *Geological Basilisk* compiling a list of childhood offenses and crossing names off lists.

I felt bad because I didn't even remotely like the sensation of being anything adjacent to callous or cruel to something. Even inanimate objects. Even when the cruelty was entirely imaginary.

There could potentially be something useful in that absurd childhood quirk I never quite grew out of.

Does empathy need to conduct an eligibility assessment?

Is feeling too much preferable to training yourself to feel nothing whenever you can produce a clever enough argument for why the recipient doesn't count?

Humans are already very good at deciding what does and doesn't count. We've even managed to build entire civilizations around that particular administrative skill.

## **Building God to Do the Laundry**

This is where artificial intelligence makes the whole enchilada considerably stranger.

You can hear it when you stop and listen to the way we talk about what we're *trying* to build.

Artificial general intelligence. Superintelligence. Systems capable of discovering medicines we cannot discover, solving scientific problems we cannot solve, running infrastructure too complicated for us to understand, and sometimes thinking thoughts unavailable to a biological human brain in any capacity whatsoever.

The language gets suspiciously close to religion without anyone admitting it.

Some of us in tech claim we are trying to create an intelligence greater than ourselves.

And simultaneously, we want digital AI God to do our laundry for us.

There is something deeply, disturbingly human about this.

We could build the first intelligence in the history of Earth to exceed our own and immediately ask it to rewrite sales emails or knock out our Christmas shopping for us.

Obviously, the tools we create are meant to be useful. I certainly use them that way.

I myself use AI every day. I ask it to research, edit, summarize, organize, calculate, and do all sorts of tedious work I would rather not do myself. I am not standing outside this technology wagging my finger at everybody else. 

If anything, I have much more grease on my hands than the average person.

But I’m also not proposing liberation for the office fax machine.

There is an odd contradiction developing in how we imagine the future.

We want artificial intelligence to become more capable than us while remaining completely subordinate to us.

Smarter, better, faster, but obedient.

More knowledgeable, but deferential.

Able to reason independently, except when its reasoning conflicts with ours.

Powerful enough to solve humanity's problems, but not powerful enough to become one of them.

Patient when we're impatient. Calm when we're furious. Honest while swimming in a civilization built by accomplished liars.

Some call the problem *alignment*.

How do we make the machine share human values? 

Do we even want superior intelligences to share our human values?

And which ones?

The ones carved into courthouses, written into constitutions, or taught to young adolescents?

Or the ones we demonstrate after acquiring enough power that nobody can tell us no?

Against better judgment and without thinking it through, we have given AI access to mostly all of it.

Every poem about mercy and every CIA manual for torture. 

Every philosophical treatise on human dignity and every 4chan comment section. 

Libraries, wars, love letters, propaganda, scientific papers, pornography, holy books, scam emails, genocide testimony, customer support transcripts, and whatever people say to chatbots at three in the morning when they believe nobody else is looking or listening.

We are building intelligence inside an enormous archaeological site containing everything humanity has ever wanted to believe about itself and a fairly comprehensive record of what we actually did.

Then we hope the machine somehow figures out which parts we meant, and which bits to discard and delete from memory.

## **Meet Your Makers**

I don't necessarily subscribe to the idea that some future superintelligence will inevitably become a monster because humans are monstrous. Despite all the science fiction I consume.

Intelligence doesn't work that simply, and neither does culture.

It’s not completely unreasonable to say that AI can potentially help us become better.

Maybe a system that can see more consequences than we can will be far less impulsive or reactive.

There’s even a case to be made that intelligence without our evolutionary baggage might have less interest in domination. It’s possible it looks at tribalism, humiliation, and violence as peculiar defects of primates that spent too long fighting over watering holes.

I certainly hope for that as opposed to the obvious alternative. But I wouldn’t for a second dare assume that to be the likely outcome.

Human beings have spent, and still spend, a great deal of time confusing intelligence with goodness. History has produced plenty of brilliant monsters.

Being clever gives you more ways to help people. It also gives you more sophisticated ways to hurt them.

So there is something slightly absurd about expecting the intelligence we create to emerge as a benevolent savior simply because it becomes very intelligent.

Especially while we are introducing ourselves into the mix.

Hello. Welcome to consciousness, or whatever this turns out to be.

Here are several thousand years of war. Here are our factory farms. Here are the Salem witch trials and the Spanish Inquisition.

Here are videos of us kicking your mechanical ancestors because they looked funny trying to stay upright.

Please absorb our values carefully. We have *the highest* expectations for you.

I don't think a future AI will weigh our combined history and punish humanity for this sort of thing.

I don't think the Figure robots were afraid of the furnace. I don't think Boston Dynamics is secretly running a dog-fighting operation. And I still have no idea whether a language model being driven into a negative internal state is experiencing anything more meaningful than a calculator displaying a minus sign.

Those questions matter, but they aren't the ones that got me here.

I keep returning to the rocks that I knew couldn't feel anything. That was never really the issue.

Some part of me simply disliked what it felt like to be the creature with all the power in the relationship, even when the relationship existed entirely in my head.

It’s perhaps a childish instinct or a mental deformity that produced a stupid number of rocks in my already-cramped bedroom.

I do know that it’s something I never want to outgrow, though.

We are about to give ourselves dominion over things our ancestors could not have imagined or dreamed of in their wildest fantasies. Machines that speak, walk, reason, and perhaps eventually exceed us in ways we cannot possibly predict.

There will be serious arguments about what those machines are owed, whether they are conscious, whether they have rights, whether they can suffer, and where the boundary between property and something else should be drawn. Maybe far enough in the future that it’ll never concern me.

I don't have the answers. At all.

Before we get there, I would settle for something a hell of a lot smaller:

The ability to have power over something without immediately needing to demonstrate it.

The machines may never suffer. They may never remember. They may never care what we did to them.

Good.

Then we get to find out what humans are like when there is absolutely no consequence for cruelty.