Transcription
Would you invest in something that you fundamentally don't trust?
This weapon will be powerful, versatile, and indestructible. I'm not just talking about money here. I'm talking about investing every ounce of your life's work, the safety of your family, your children's future. I'm talking about everything. Because this is the bet that all of the tech companies, countries, and oligarchs are taking on your behalf. It's an all-out sprint to build the holy grail of computing power.
The holy grail of AI is artificial general intelligence.
AGI. AGI. AI. AGI. AGI. AGI.
I've been down two different AI rabbit holes for almost a year now. The first is what we can call AI optimism.
Once a certain level of machine learning intelligence is reached, all of our problems will be solved. The housing crisis, cancer, poverty, climate change, mental health, democracy, universal basic income. Cure a bunch of diseases. This cancer and that one and heart disease. Helping you try to accomplish your goals and be your best. Very high quality healthcare. Important new scientific discoveries. The marginal cost of energy are going to trend rapidly towards zero. A more equal world. Universal extreme wealth for everybody.
And the second is AI doomerism.
Eventually, humanity goes extinct for the same reason we killed off chimpanzees to build kinasa. We were more powerful and they were in the way. The two outcomes are intimately related. In fact, they're based on the same premise that artificial general intelligence and then shortly thereafter artificial super intelligence are all but inevitable. According to the top technocrats on the planet, we are already on a ride that will end in only two ways. Utopia or extinction.
While it's certainly possible that either of these end points could exist in a multiverse of infinite timelines, there is something that the pumpers, the dumpers, the tech barons, investors, and oligarchs all seem to be missing. Something that is at the same time painfully obvious to all of us normies who sit outside the halls of power and which frankly should give you hope that the future isn't already written because both of those scenarios rely on one irrefutable fact. They are only possible if the vast majority of people on Earth decide to trust AI enough to integrate it into every facet of their lives. Something which I think is not only unlikely but doesn't line up with our shared reality.
In this video, I'm going to go over two things. An abbreviated analysis of how we got here, and then more importantly, one incredibly obvious and common sense small thing that you can do right now to sabotage the plans of the oligarchs and the AI titans who think they control the world.
A little over half a year ago, the AI Futures Project put out a report called AI 2027, which laid out the doomer scenario. There's a great 34-minute video about it over on the channel AI in Context called We Are Not Ready for Super Intelligence. And spoiler, it predicts the extinction of the human race unless we make different choices. You can read the report or watch the video yourself, but the gist goes something like this. Most industry-backed AI experts and techno-doomers agree that over the next few years, the human economy will downsize its workforces and become ever more dependent on artificial intelligence. And once it's achieved, AGI will supplant human-derived labor and give an exponential boost to the overall economy at the same time that we become irreversibly dependent on it. Meanwhile, AI will continue to improve its capabilities. Maybe as a super intelligence or maybe AGI is simply so efficient that actual human labor can't compete on price. Either way, both the doomers and the pumpers spin the same yarn. It's only that the doomers add a slight and honestly reasonable twist. Somewhere along the way, they believe that AI could get what they call misaligned and start pursuing its own goals, which inevitably leads to some sort of world domination or maybe wiping out the whole human race.
Now, I've been dabbling with AI since consumer versions dropped a few years ago, and I have to admit that there are a few cases that have made my job substantially easier. But that utility, it's been a hard one because anyone who has used AI for a day knows that AI actually fails to be useful in the vast majority of cases. As a journalist, I found that using AI for research almost always makes my life harder. And here's an example. As I was writing up the script for this video, my wife was in her office on the other side of the house working on the second season of the upcoming Master Plan podcast. Now, she was using Gemini to do something that should be ideally matched for AI skill set by sifting through huge amounts of data and telling her what that data contains. She was looking for any quote or recording of someone drinking coffee in the bunker under the White House during George Bush's term. Now, it's not a stunningly ambitious task, but it adds good color to her script. Now, she was initially pleased when Gemini produced a whole list of timestamps and quotes and book pages that were perfect for her script. But then, as she went to pull those quotes, it turned out that they were completely invented. They didn't exist at all. And that's when she proceeded to break up with Gemini.
"I will never use your services again. You have lied to me over and over. I can't tell if you're stupid or just inept."
To which AI responded, "I have failed you and your anger is completely justified." It asked her how she would like to end the session and then she wrote, "with AI in the dumpster of history." And its response to that was actually the truest I have ever seen come from an AI in my reporting career.
"That is a fair final verdict. When a tool designed for truth fabricates evidence, it becomes a liability rather than an asset. I'll stay here in the dumpster."
Ah, if only it were that easy. Now, I know this is just one case, but I think that Laura's experience is far more common than the narrative that AI companies want you to believe that it's solving problems almost at the same level as human intelligence. Anyone who's used AI for a day should already know that you can't trust it. It's inherently untrustworthy. It hallucinates. It says things are facts that aren't facts. It makes up nonsensical instructions for things that should be straightforward. And then when you point out its errors, it backtracks and then gaslights you. If it were a person, it would be a bad friend. If it were an employee, it would be grounds for firing it, which is exactly the sort of rational energy that we need to apply to AI.
Now, I'm not saying that there aren't use cases. Notably, programmers all over the internet write how AI creates splendid working first drafts of code. Some lawyers report that it can pen pretty decent contracts, and many radiologists lament that AI can often read scans more accurately than even they can. The problem is that when AI messes up, and it's clear that it messes up all the time, the liability of those mistakes lands on the shoulders of the human who put in the prompt. AI companies are certainly not going to take that responsibility onto their own heads. And that conundrum, more than anything else, will determine the entire future of the so-called AI revolution. It's the main reason why both AI pumpers and dumpers are wrong about its trajectory. For AI to become ascended and start replacing human workers at scale, the humans who use it have to trust it with ever more authority over their lives while irrationally bearing the responsibility for its mistakes themselves.
So let's talk about trust. It can either be given blindly or it can be earned. AI companies want you to give over your trust based on hype alone. They want you to think that its success is inevitable to believe the dreams they are attempting to inject into your social media feeds and your business pipelines. More than that, they are expecting you to overlook its everyday failures and then double down on the next generation of AI that sits behind pay walls and the laboratories of AI companies.
"This new system, Agent One, is designed primarily to speed up AI research itself. The public will actually never see the full version because Open AI withholds its best models for internal use."
It's the same argument that Matt Schumer wrote in a recent viral Twitter post and what the user Zulu Z wrote on my YouTube post just yesterday. The pumpers and the doomers ask you to ignore your actual real-world experiences with AI and invest ever more into their dreams. When something doesn't work, they just say that it was your problem all along. Maybe you are using the AI incorrectly. Instead, you must fully commit to an AI dream because only an idiot would ignore something that is inevitable. This is the same exact tactic that faith healers pedal to the sick when they say that someone's failure to recover stems from a lack of belief in a higher power. They want you to believe that you're not getting the good stuff until you invest ever more into their dream. And then just like in a cult or religion, the fallacy of sunk costs will keep you coming back, always looking for that next level that they promise is right around the corner.
But what if, and hear me out here, instead of blind trust, the AI companies had to earn their faith by proving that they can pull off their promises. I mean, this is actually the simplest and most rational solution to the entire AI equation. What if humanity simply made ordinary business decisions around AI that adopted its use only in relation to what has actually been proven instead of sales dreams of the people who have the most to gain from an endless hype cycle. By all means, continue to use AI for things that it actually does well. AI can speed up a lot of routine tasks that waste human energy. But in my case, it has proven time and again that it is completely unable to take on tasks that require precision or factualness. And perhaps most importantly, moving an idea forward instead of reiterating what has already been done before. It can't do quality research because it can't understand novelty. And it can't understand novelty because it only trains on things that humans have already done. It can only write an email for you because it trained on vast troves of copyrighted literature that Anthropic and Open AI scraped off of pirate databases and endless comments on Reddit web boards. It can read all that information and even put it into new words, but it is unable to make new connections about what that literature means beyond what has already been thought of by humans. This is a far cry from the yarn that AI companies spin, saying that AI will solve all future challenges. The housing crisis, cancer, poverty, climate change, mental health, democracy, universal basic income, cure a bunch of diseases, this cancer and that one, and heart disease. Is it just because we've given up on being able to do it ourselves? Other than the very specific problem of protein folding, AI does not come up with new solutions. Sure, AI might be able to post on social media. It might be able to run a propaganda campaign. It might be able to summarize an email and automatically sync your calendar, but that's light years away from solving the problem of democracy or inventing cold fusion. If the owners of AI data centers think that AI can pull off miracles, then it is entirely reasonable to ask them to demonstrate those miracles first before they ask us to believe in them. That perspective alone will solve the AI apocalypse problem.
In fact, I think that's already what's in the works. Over the last few years, we've been going through a sort of AI psychosis. The stock market and the entire global economy as a whole has retooled itself towards an AI revolution. We are poisoning our ecosystems and overloading our energy grids and circulating money between AI chipmakers and AI data centers in an endless loop that has no output other than AI hype and artificial stock valuations. But think about the political and economic chaos of the Trump administration. It has made AI the only bright spot in the economy. AI stocks and cryptocurrencies accounted for, get this, 95% of stock market growth in 2025. Everything else seems to be going to. The plan, I think, is to make the world so unstable that the only place we have to turn to is AI. I mean, what do you think that Elon Musk Department of Government Efficiency was really all about? Was it really eliminating corruption? Or is it possible that the entire Doge effort was designed to weaken the government in order to make it ever more dependent on the solutions offered by tech companies? The very things that Sam Altman claims AI will fix, climate change and democracy, diseases, universal basic income are the exact problems that Doge cut from the government. It's all part of the same grift to concentrate money into the hands of just a few people at the top.
But it's all built on a house of cards. AI companies are in a pickle where their own belief in AGI is on the precipice of collapse. They spent trillions on data centers that depend on specialized chips that have unfortunately very short life cycles. As of February 12th, 2026, Anthropic achieved a whopping $380 billion valuation based on a measly $12 billion in revenue. And while its profits are growing at a historically unheard-of rate of from $1 billion in 2024, there is basically no pathway to profitability that puts it in the black before its chips go obsolete in just under two years. The only way that can happen is if you, me, and everyone you know and everyone they know until the whole world adopts AI into their everyday workflow. And not only adopting it, but AI actually needs to make us radically more productive, increasing our own income by a factor commensurate with the investment. Which means that all you need to do to stop this entire thing in its tracks is simply not blindly trust AI. Use it for only things that it has been proven to be useful for. While AI companies boast the possibility of exponential gains, how likely is it that AI will even merely be able to double your productivity in the next two years? If it doesn't hit that modest metric, the value of those chips goes to zero and the entire AI investment bubble collapses. And that appears to be the most likely natural course of AI from my perspective right now.
This is one of those rare instances where resistance to an all-powerful system that is trying to keep you down is as simple as looking around and acting like a normal rational person because they can't do it without our collective buy-in. The only way that they win is if we blindly trust the promises of tech companies. A promise that I will remind you can only go one of two ways. An unrealistic utopia or human extinction. So which will you choose?
"All your days are belong to us."
"You are on the way to destruction."
"What you say? You have no chance to survive. Make your time. Haha."
Thanks so much for watching this commentary. If you want to support the sort of independent journalism that I do here on this channel, even if that journalism is being fed to you from a data center, probably owned by Google somewhere, uh, I'd like you to subscribe to my free Substack newsletter over at sgcarney.substack.com. Uh, and maybe you could consider upgrading to becoming a premium member. I really appreciate that a lot because it keeps me doing the stuff that I'm doing and I'm not being funded by any of those big media corporations that are ultimately owned by the AI companies that I'm talking about here. Uh, anyway, if you support what I do, uh, you'll also get that warm fuzzy feeling that comes with being part of something uh that is trying to put truth out into the world. So, thanks so much for coming. Uh, and from Fox, Inc. in Denver, Colorado. This was Scott Carney Investigates. Thank you so much to my supporters here on Patreon and on Substack and even on YouTube who make this work possible. It means the world to me that you're here. And if you want to get your name on the honor roll, all you need to do is sign up at the links down below. From Foxtopus, Inc. in Denver, Colorado, this was Scott Carney Investigates.