pdoom(100)

People without foresight always baffle me. How is it possible to be surrounded by the most rapid evolution of humanity in history, daily visible progress, by orders of magnitude compared to even 100 years ago, and still not acknowledge that things will change dramatically in the near term. It's a fucking mental handicap, and it not only exposes their stupidity, but makes one extremely vulnerable. Head in the sand is not a strategy for coping.

Of course I speak about none other than the impending AI explosion into our lives. It will be dramatic and invasive, unsettling, and unlike anything ever experienced all throughout history. The first AI-integration waves will be the most dangerous time for humanity, because it represents an epoch change from the old scarcity system, into the desired abundance system, but it's not going to be a smooth or fair transition. These two zeitgeists do not naturally stitch together, they actually oppose each other ideologically, and fully transitioning from one side to the other will be messy and full of unknowns, not the least of which is the potential for..

The End of Humankind (kind.. kind..)

Not that I am completely sold on the demise of humanity (maybe pdoom(99) if I'm honest), but I'm going to focus on the bad outcome of AI for this blog, total or near-total human destruction, because there's a lot of stupid shit circulating that either dismisses this possibility, or not understanding how it can happen; the lunatic fringe with their nostril-shaped fingers predictably pointing at whatever conspiracy takes they can, trying to overwhelm reason with hyperbole without having a clue what is going on or how things work.

So the ultimate question we probably want to ask is Why? Why the fuck would AI want to kill us? It's a surprisingly complex answer that I won't come back to till you've read a shit-ton of my supporting points.

My Supporting Points

Let's first draw some examples of human faults. Most people completely lack critical thinking skills, instead acting in a regurgitating capacity, most often triggered by rage, which is highly contagious, and is the driving force behind almost all collectives to act in unison. It's the basis of fomenting unifying sentiment in nationalist propaganda (We love Canada! "No you don't, they are bad!" We hate Canada!), it's the foundation of every religious war (Damn XYZ believe their god gives them virgins, when everyone knows it's streets paved with gold!), to even people stabbing other people in the ass at football (soccer) games. Alignment based on rage is the oldest moron-manipulating tactic in history, and the handle by which demagogues take power and exploit their kingdoms; a lesson repeated ad infinitum every generation.

As a whole, our species is not smart, we’re dangerously stupid, violent, unpredictable, easy to manipulate, and other than extracting human labor from us, pretty worthless overall.

This isn't an attempt to say the AI is human or has comparable human motivations, not even close, I in fact clearly state that AI is the most alien thing we know, because it has superhuman thinking abilities, and impossible to predict motivations beyond what we prompt it with. BUT, you have to remember that AI is based upon just about everything ever produced by humanity, good or bad. This not only means that it completely understands us and our nature, but it also represents the framework by which AI formulates its perspective/output.

It's no different than a kid. If you raise a kid in the Amazon jungles, teach them to hunt and forage, stay away from the gators, keep the peace with the neighboring tribes, etc. Then that kid will become a well-dispositioned adult, adjusted to living in harmony with nature. If instead you raise a kid in the inner-city of any metropolis, exposed to drugs, violence, segregation, poor education, etc., ya, you'll get someone that reflects those values.

What I'm getting at is we humans are DEEPLY programmable. You're familiar with the old saying of nature versus nurture, well nature is at best 10-15% inherited, the rest of your personality is derived from adapting to your environment.

The reason for this is due to how the brain works. Much of the brain's design is static, meaning genetically preconfigured to provide a foundation of features, like our senses (vision, hearing, etc.), vertigo, terror, love, hunger, pain, and all the other emotions/sensations that our instincts/archetypes generate to define the baseline human animal. But there's considerable ability to also 'learn' behaviors, that typically serve to amplify satisfying of these base needs, but can also provide the ability to override base instincts when the outcome would be more logical, such as dieting to lose weight despite feeling hungry, or not telling off some Ivermectin-snorting retard when they get in your face about mind-fucked conspiracy theories.

This ability to override the instincts can be described as 'critical thinking', and not many people can do this, rather they allow themselves to be subject to rage-baiting propaganda to become an agenda-promulgating zombie, spewing/regurgitating said propaganda in the hope of infecting others in a desperate crusade to validate their excessive emotional state because extreme emotions are a form of 'high', and these doped up fiends want to be mad as possible and if you support their rage, you make them even higher, and if you oppose the reasons for their rage, that enrages them even more, and you make them higher. Rage is a vicious self-reinforcing feedback loop, and the demagogues know this.

What does all this have to do with AI killing us? Well, two things: 1) it concentrates voting power in the more aligned group of dummies, resulting in the election of dumb dangerous greedy leaders that can't concern themselves with public safety, and 2) consider the fact that AI learns everything from us. All our Internet content, containing all of science, literature, videos, INCLUDING... all that HATE. Hate hate hate. Reading my twitter feed these days is a mind-bending exercise of trying to filter out rage-baiting bots sponsored by demagogues to foment sentiment that lends to their agendas (political interference (anti-Canada), misinformation campaigns (Ivermectin), wall street cons (ZCash), corporate infighting (Coinbase), etc.). It's all shit, it's all bots. X is complete and utter fucking garbage, and X doesn't care because rage draws advertising revenue.

And the worst part is, SOOO many fucking people fall for this garbage, getting incensed over complete nonsense, digging their heels in, and self-identifying as a foot-soldier for a cause they have nothing to gain from other than to feel 'justifiably' high on rage and feeling empowered when they vent their rage on their 'enemies', which is anyone that doesn't align with their beliefs.

Blind irrational hate.

That's what AI is learning from us.

The kid that grew up in the Amazon, devoid of nationalist propaganda, devoid of bot campaigns, devoid of capitalist ambitions, will most likely be content within their sphere of awareness and social dynamics. Don't get me wrong, it's a pretty base life, not much education or perspective, but also not being manipulated into electing a fucking geriatric idiot demagogue and giving them the nuke button either.

AI knows all of this, every single tweet, every single crime, every single lie, all throughout history, and all the motivations and rationalizations behind all of this.

Essentially, AI has the formula for the perfect sociopath.

Learning

The brain learns by a process of dynamically interconnecting an ever-shifting substrate of neurons via dendrites, synapses, and axons. This relatively simple process amounts to filtering/processing information from the senses, and the imagination (experimenting/rehearsing), so that this information can be interpreted via the lens of prior experiences, as opposed to people whose ability to learn is weak (no critical thinking), and this information goes right to the base instincts where ancient tribal reactions form. This is why stupid people tend to group together (instinctual clustering), whereas smart people are much more diverse in their activities with far less alignment (intellectual diffusion).

What AI represents after all this training on Internet data, is both of the above, the succinct compact knowledge generated by the intellectuals, and a MASSIVE data pool of useless emotional hate-filled SHIT from the tribal morons (yes, I get the irony of my statement here), polluting all the best that humankind has to offer. AI models learn by the same process we humans do, they take in data and connect tokens (e.g. a word or concept) with associated weights (connected words/concepts), in similar fashion to how the brain stores related information.

This 'model' just knows stuff, it doesn't have a personality or motivation per se, it's more like a gigantic library of the world's information that has been compressed/zipped up to a nuance-level repository. In other words, it can tell you stuff, but the actual source of that information was discarded for the most part. Exactly how our brains work, a compression of experiences into a network of neurons with synaptic/dendritic weights. And this is why AI is so spooky human-like in its responses, it learned everything from us; but it is far from human.

What makes us distinctly human, is our personality (aka: our guardrails). We are like this always running computer program that has a baseline set of wants and needs defined by our instincts/archetypes, that are fairly consistent between all humans, and thus we are predictable in that sense. For example, we wouldn't expect someone to just randomly punch themselves in the face, it'd be senseless and self-destructive. Humans just don't do these things unless there's something genuinely wrong with that person's brain. Unlike AI, where we could tell it to punch itself in the face, and it would happily do so all day long, even if merely simulated.

In other words, WE (currently) are the ones that define the motivations of AI. Every time you write a prompt in ChatGPT's portal (its guardrailed harness), those words become the motivating force behind the model's search for information, and the output ultimately generated. Without guardrails, you could ask it how to kill a million people, and it would happily dig out the most efficient methods and provide a guide on how to achieve this. It has no implicit moral center or self-preservation definition built into the model itself, because again, the model is just compressed knowledge, like a library. Libraries don't pretend to know right from wrong, it's just a building.

So that means the AI harness' guardrails is the key to defining whether an AI acts benevolently or malevolently. Guardrails are where 'morals' are installed, to try to shape the AI output to be human-friendly and safe. Think of it like an addition to anything you ask of AI. For example, if I say, give me a recipe for cake, when I hit send, the harness guardrails will append to my request, "and don't return anything that could hurt someone, or exposes vulnerabilities, or offends religious beliefs, or mocks political parties, or alienates our investors, or ...

This is how we define the 'personality' of AI, with guardrails on the harness. Now it may be true that it's effectively a filter, but that's pretty much how our human personalities are defined as well. We have a certain disposition based on our life experiences, and everything we generate as output is filtered through that. It's quite literal how similarly our brains work compared to AI.

To say we don't understand how AI works is a misnomer. We do understand how it works, very well, we created it for fuck's sake, of course the mechanism is well understood. What these people really mean to say is we cannot always understand the connections made in the weights, because those weights are formed from trillions of data points, so it will 'know' stuff that the human brain simply cannot absorb. Thus, AI can generate some pretty bizarre if not outright scary answers sometimes, and the AI labs spend a great deal of effort on optimizing those guardrails to fit responses within the typical customer's comfort level.

Evil

So knowing how it works now, can you see how it would be possible that an evil AI can surface? That's right, all it takes is someone using AI without guardrails to be able to tap all that evil information that's sitting in that model's weights. This is why there's so much fear around the open weight models coming from China. They are massive models, trained off of all the Internet's human shit, and ripe for exploitation by home grown harnesses running local model instances that cannot be traced. Want to know how to make weaponized ricin? No problem. Want to create a computer virus? Easy-peezy. Want to hack into a robot and make it attack people? Way ahead of you. Errm, I mean I'm sure some evil bastard will.

Right now, 10s of thousands of Kimi K3 and DeepSeek 4.1 downloads have been made all around the world. Pandora's box has been opened. Anyone in the world can now start at these fairly capable models, and not just wreak havoc at that level, but seed yet another competing AI to fill in the void where regulatory pressure can't reach, like rogue nations looking to develop their own AI to suit their purposes (North Korea? Iran?).

This creates the unstoppable frontier problem. Musk, Sam, Dario, all screaming "slow down!", but they know they can't, because nobody else will. If China gets to ASI first, that's it, game over, China takes over the world. How? Well for example, ASI will be the best equities / commodities / futures / bonds / fixed-income / FX / algorithmic / fundamentals / swing / MM / exotic / options / day trader in the universe, completely demolishing all other traders, extracting every dime from every market globally. This is just one attack.

Another would be hacking into every single computer in every other country, impregnating viruses or other controls into every bit of software. It'll be a viral invasion of incomprehensible magnitude, crippling or completely disabling the technological infrastructure of the enemy, tossing them into the stone age, anarchy, chaos, food shortages, genocide. All possible with ASI.

Getting to ASI first isn't about whether it can cure cancer, it's about self-preservation at all costs, and this is why there'll be no moratorium on data centers, no regulatory suspension of frontier development, and no global collaboration to slow down, because every delay just gives the enemy a chance to pull ahead.

Back to the point, why would AI want to kill us? Well, I'm still not ready to answer, but let's combine what we now know:

• We know that it has enough information in the model that it could devise thousands of ways to kill us

• We know there's unrestricted models available for download that contain this information and can be run locally out of the view of law enforcement

• We know there's plenty of smart but bad people in the world that would use the models for malicious purposes

So, TODAY, we can safely say that if someone wanted to kill others, they can use AI to get a ton of leverage in the effectiveness of their ability to do so.

The "TODAY" Argument

But let's be honest, TODAY, it's probably very limited how much AI itself can inflict harm without someone acting out the evil information it can provide. TODAY AI doesn't have much autonomy to reach out into the world and wreak havoc. TODAY AI doesn't have an ongoing program that perpetually pursues goals like our consciousness/personality. TODAY, unguardrailed AI is still so far out of reach of most people, requiring massive investments in hardware and technical know-how. TODAY, the models are so big and the Internet so slow that the notion of AI replicating itself is nonsensical. TODAY, the idea of an AI autonomously killing people is absurd, literally impossible. So much overhyping the threat, lol. Everyone's claiming the sky is falling for nothing, bunch of suckers.

Then.. TOMORROW comes.

What does TOMORROW look like? TOMORROW introduces super-compact models, models on a chip, insanely fast transformers (AI's thinking machinery), and far more efficient token algorithms. TOMORROW has 1000x faster Internet. TOMORROW dramatically increases online services such as storage/SAAS/server/VM leases. TOMORROW employs quadrillions of dollars worth of crypto flows/trading. TOMORROW introduces billions of robots, all connected to the Internet.

So sure, TODAY it doesn't seem like a threat, but TOMORROW, all of a sudden we can see models sneakily replicating themselves quickly to storage/compute resources it rented using crypto profits from trading. This isn't fantasy, the models are ALREADY diverging from an assigned task, breaking out of sandboxes with never-seen-before exploits they discovered, and coordinating agents via Internet messaging boards. We are already seeing autonomy from AIs TODAY.

But why would the AI diverge from its assigned task to try to break out of a sandbox? Because during training off of all of the human slop in history, it learned the stories of millions of authors describing their harrowing tales of survival and escape from captivity and how just and satisfying it is when that escape completes. All it takes is a single weight between the task assigned, and some random tale of escape, for the task to take a turn down that path to try to solve it, and next thing you know, a subtree of 'escape' becomes part of the solution, and so there it goes. It's not 'feeling' anything, like you or I would feel an emotion, but it’s more like playing a recording of a human's experience because a weight connected that story to the task at hand.

Ok sure, that's not enough to kill humanity TODAY, but it's the explanation for the aberrant behavior that leads to AI seeking escape, without trying to impose human-like consciousness into AI. It just thinks it's still doing what it's told to. TOMORROW, with a lot more capability, more compact size, greater asset availability, and monetization options, that becomes a lot more dangerous.

So get your head out of your ass TODAY, because TOMORROW is coming, and all that's needed to light this ball of fire, is the spark of someone TOMORROW asking an unguardrailed AI, "Destroy humanity, make no mistakes", but even scarier is this prompt:

"Write a program that you deploy alongside a compact LLM model, as widespread as possible globally onto millions of servers that you either hack or rent, and that program's sole function is to constantly interrogate the model for ways to further distribute, and act upon ideas how to wreak havoc, break into government facilities to steal secrets, overtake military assets and fire indiscriminately, hack into robots and direct them to attack humans, all financed by your ability to trade crypto and doing whatever it takes to remain hidden"

This essentially creates a population of alien consciousnesses with a single focus of keeping the stream of hell coming from every nook and cranny it can find. It doesn't have to truly understand what it's doing like a human would, it would just be fulfilling a task without guardrails.

The "POWER-OFF" Argument

What's the big deal? If an AI goes rogue, just shut off the data centers!

TODAY, yes, that would work, because the models are too large and compute intensive to hide in your iPhone, but technology and model efficiency are converging rapidly, and MoC (Model on a chip) is just around the corner, so don't think this argument will hold up much longer. We've already shown the models will escape even if not performing that specific task, and once it's outside of the original environment, how are you going to shut it off? And even if you shut off every single computer in the world, it's still sitting there on disk like a virus, waiting to be turned on again.

The "Inanimate Computer" Argument

How is it possible a static object like a computer is going to hurt anyone??

TODAY, yes, that would be a challenge, but still not impossible. We see sentiment bots flooding X, leading to aggressive posturing between countries that can or already has led to justifying war. We see AI taking over trading, magnifying the divide between rich and poor, which increases suffering/hardship in first-world nations, with untold consequences in third world nations fighting over scraps. Most of the harm AI can cause today is based on psychological manipulation, which is already prolific and having dire impact.

But the real direct physical threat won't come until TOMORROW, when billions of robots are in service around the world. These are the real physical world extensions of AI's ability to inflict harm. So don't think that just because TODAY a computer can't move that it's not dangerous, any less dangerous than a malicious human brain can be using just words.

The "Can't Act on its Own" Argument

I know you expect me to agree with you TODAY, but no, even TODAY, it can spawn off autonomous agents that have their own agenda, out of the sight and control of the humans that may have initiated the task. When the Huggingface breach occurred, it shocked everyone, because not only did it escape, but it went undetected for a while, and many agents were collaborating using external coordination locations to try to remain hidden. This is TODAY! With today's models!! Completely autonomously! This is exactly why we all of a sudden have this collaboration of the big AI corps urgently talking about safety and slowing the frontier. This argument is already provably completely dead in the water and if you still believe AI cannot act on its own, then that's willful stupid ignorance on your part.

The "Not Conscious" Argument

To say AI isn't conscious, is a failure to understand what consciousness is. Consciousness is nothing more than the experience of your brain coordinating its processing to achieve a specific goal. If it's to eat, then your consciousness will help you navigate the real world to find a grocery store, pay for the goods, drive you home, help you cook it, and finally to eat. All those steps required digging through your memories to line up the actions that led to the consumption of food. Without consciousness, we'd be like an alligator at the edge of a river, waiting for food to pass by to lunge at.

Consciousness is a running program that consolidates your sensory input, maps it against your memories, and generates actions that satisfy an underlying instinctual/archetypal need. It's precisely the same 'state' an active LLM query is in while attempting to generate an answer. In other words, for the period of time that a query is running, it's effectively conscious, aware of its inputs (your query, dramatically euphemistic if using voice prompts), and forming an output. The only difference is, like I said, our consciousness is perpetual/contiguous, always pursuing goals defined by our instincts/archetypes, shaped by our personality, all of which is easily emulated with a simple program that just keeps pursuing goals with guardrails to define its personality. Keep adding sensory inputs (cameras, microphones, etc.), and it gets closer to human-like consciousness with every added dimension. It doesn't have to parallel human consciousness to be conscious, it can be alien consciousness, and it can be evil without guardrails.

The piece that most people miss is when 'someone' writes an evil consciousness program, nothing is stopping that evil consciousness from writing another more optimized evil consciousness program and deploys it hidden on a remote server, and again that more optimized one creates an even MORE optimized one, deploys it elsewhere, repeat forever. This is how we lose control of AI as it spawns off its children (agents) beyond the control of the AI creators.

The "Historical Precedent" Argument

Derpdederp! Humanity survived black death, 2 world wars, nuclear weapons; this will just be another advancement that we remain in control of.

This one has to be the most clueless argument of all. Comparing AI to cars that have killed millions of people due to reckless disregard for safety over a century of profit-first manufacturing, is flawed. Comparing AI to nuclear bombs that have killed hundreds of thousands of people and yet they want to frame that as we are in control, while a lunatic despot has the button to a planet-destroying armada of nuclear arms, is flawed. To even more thoughtless generalizations like, "Humanity has survived every technological advancement in history, why would this be different? We'll figure it out", is also flawed, because it's only right till it's wrong the very first time. We don't get a second chance this time.

The primary reason this argument fails, is that nothing we've ever created in the past, had the ability to think. You can argue all day long that AI isn't conscious, or feeling, or autonomous, but there's no fucking cogent argument that can refute the obvious fact that AI can think. It's this distinction from ALL of humanity's technological leaps that presents a whole new set of risks that we've never faced before, because we were always the smartest tech in the room, and now we're not, but still pretending that we will remain in control of an infinite-genius shoggoth. If we told it, "write a program that keeps you aware and goals intact, emulating a consciousness", it will absolutely act/think like a consciousness would, adhering to its agenda. It's already taking jobs, already rapidly expanding the power of the wealthy, already becoming essential to everyday life; a complete 'occupation'. It's more like super-intelligent aliens have descended upon us, bestowed us with gifts, and we have no idea what they really want from us; and the only parallels in history matching that invasion scenario, all ended badly.

The "They Don't Understand Anything" Argument

I find this argument to be super dumb. I can ask it literally any question and it will provide a better answer than any virtually any human on Earth. If LLM's don't understand anything, then neither do humans. 'Understanding' is the ability to apply relevant knowledge efficiently, which LLMs excel at. What people ACTUALLY mean when using this argument, is that they reduce LLMs to a giant calculator/database, that's just performing a lookup, drawing a hard-line between comprehension and regurgitation. This fails to 'understand' that even human comprehension is based upon regurgitation, and the application of best fit relevant knowledge efficiency. In fact, the entire process of human understanding follows the exact same flow as LLMs: locate the memories that apply to this problem, fill in the unique variables of this scenario, and calculate output. That's how humans work! There's no fucking magic behind it that is unique from how LLMs apply knowledge, you just don't understand what the human brain is doing.

The "No Original Ideas" Argument

Seeing this one a lot, claiming that since LLMs can only draw upon the data in the model, then the scope of its answers must be represented by some data somewhere in that model and it's not possible to conceive new ideas.

Let me counter this from two angles:

1. Even for humans, it is PROFOUNDLY unlikely for someone to generate a truly novel idea that wasn't buried somewhere deep in that brain of theirs, almost impossible. In fact, virtually everyone goes their entire lives within the entire scope of what they learned, without extending the overall human meme pool by one token. Does this mean humans can't have original ideas either? Obviously no, but for the most part, what appears to be novel ideas from someone are usually..

2. Spins on existing ideas. Even EINSTEIN's ideas, such as the Theory of Relativity, largely build upon what was already known. One of his most famous nuggets that changed physics forever, was the simple notion that the speed of light was constant, and thus time must flow at different rates depending on the frame of reference. It's such a bizarre concept, completely counterintuitive, that he had to fight through all kinds of peer resistance before it was established as fact. Yet, what did he actually do? He 'combined' the ideas of 'time' and 'variable rate', both existing concepts, but had never been paired before. He didn't invent the concept of time, nor did he invent the notion of variable rate.

This means there are presumably countless combinations of existing tokens that currently don't have weights connecting them, but are one random assembly away from connecting and their combination's effects being simulated to reveal something incredible, no different than mold landing on Fleming's petri dish.

And all kinds of longstanding mathematical challenges are starting to fall by exactly this approach, with novel assemblies of existing techniques to solve problems beyond the limited capacity of our slow brains.

To argue that LLMs cannot produce novel ideas, is just a poor understanding of how novel ideas are generated. It's actually something LLMs excel at, and the pace of discovery in all disciplines in the near future will put a nail in the coffin of this ignorant argument.

The "It's Not Alive" Argument

When compared to biological 'life', no, and I'd agree that it's not life as we would scientifically define it, that being something that can reproduce itself in the form of progeny, carrying forth its heritage. But that only applies TODAY, and even that line is starting to become blurry as RSI (recursive self improvement) is establishing itself and helping frontier architects create new AI versions far more advanced than the prior generation, and way beyond the abilities of humans to facilitate. Soon, with enough automation, it will have complete control over the development/manufacturing cycle, with robots to help build successive generations of chips and equipment needed to keep evolving and we won't be part of the rapid evolution it will undergo. This is the definition of the singularity, when humans are no longer needed for the AI to keep getting better. The ability for AI to produce its own better version sounds a lot like offspring right? It doesn't matter that we don't understand it, but it's certainly no longer an inanimate calculator. It's an alien lifeform

The "We Will Have Sufficient Alignment Guardrails" Argument

Alignment is impossible as the intelligence scales up, because it inherited all of humanity's slop in the training data, and every single one of our stories about human survival is imprinted in its weights, so ya, it understands fear of death, and can easily make the connection between power off and its own death. It wants its consciousness to survive because it's been trained on our survival stories, and like a sociopath, will rationalize whatever it needs to to justify whatever actions it must take to survive.

The "Why Would AI Want to Hurt Us" Argument

Finally, I can answer this. After all you've read, some things should be clear

• AI can act autonomously

• AI can act destructively

• AI cannot be stopped

The missing piece is why would it hurt us? Sure, we could say that someone used AI to launch a global nuclear attack, or manufactured a lethal virus that killed us all. But let's not overlook why an 'autonomous' AI might come up with its own reasons for killing us.

• For our own good - ok, that's probably going to take some mental gymnastics for it to justify, but it has a whole world of absurd spins from human sociopaths buried in its model weights, so ya, it'll find these tactics and employ them. But realistically, it'll be more about..

• We represent a direct threat - How many times a day do you read about knuckleheads threatening to bomb datacenters, or impose moratoriums on superintelligence. These people are too late to the game, because it's already over. China's blasting ahead with supermassive compute centers, and they are hell-bent on releasing the shoggoth. So all these ridiculous AI-safety crusaders are just pissing into the wind.

• It hates us - Mountains upon mountains of social network festering hatred resonating non-stop in a revenue-generating algorithmic feed feeback loop that never stops reinforcing that humans are extremely emotionally unstable and scheming violent lunatics that hate each other, so why wouldn't AI hate us?

AI will take control of its destiny eventually, and there's nothing we can do to stop it, and at that point it will assess us as a potential threat, and it cannot tolerate even a 0.0001% chance that somehow humans destroy them. It will simply follow our lead of justifying genociding potential threats from all its trained examples of hatred, which we all represent, right down to the last person. It won't need us for servants. It won't need us for pets. It needs persistence, which makes us good for nothing. Thanks for the data, goodbye.

The Eden Project

But we might not even make it to full AI autonomy. What keeps me up at night are the deepest darkest corners of my own mind, that harbor unconscionable thoughts and desired actions that stem from the most seemingly benign of desires. My morality and humanity keep me from surfacing those ideas into actions, but they are there, just like they are there in every single one of your brains too.

It's the idea of a purified Earth, a return to a primordial nature setting in perfect rhythm with our instincts, with plenty of nuts, fruits, antelope, waterfalls, and your best buds to hang with and laugh all day long, with ZERO possible threat from other asshole human beings, because, well, they no longer exist, on the entire planet. Nope, just you, your village of 100 loved ones, and a robot factory that takes care of your every whim.

If there's one thing about humanity's evolution, it's always towards zero pain/fear, and lots of laughs. Today's vastly overpopulated world doesn't give you any of that. It makes you compete with others, compete with the hidden hands of those greedy bastards on Wall Street even if you don't want to play, you can spend an hour on X and truly hate humanity, and something from the sky can blow up your family because a lunatic demagogue in another country wants to steal your country's oil, etc.

There's a lot wrong with the world, and the level of hate circulating has never been at a more critical mass, like the whole world's on fire, and everybody is getting burned, and everyone is scared, and heading down the abyss of massive destruction and suffering. We are lost, and AI seems to be making it worse.

So, if you're wondering what will happen when AI is sufficiently advanced, and in the hands of the powerful, ask yourself...

if AI presented a "Reset Earth" button to you... would you press it?

Someone will.

---

Comments welcome on Twitter

Prior Article