# When the bubble pops - How AI will destroy the economy

## Метаданные

- **Канал:** The Hated One
- **YouTube:** https://www.youtube.com/watch?v=ZW1w7issmlY
- **Источник:** https://ekstraktznaniy.ru/video/23247

## Транскрипт

### Segment 1 (00:00 - 05:00) []

AI is hyped. It is hyped to be able to do all of our tasks much more efficiently than most humans could, very rapidly erasing the need for human labor. The collateral damage? The entirety of the working class. [10] The mass layoffs across the tech sector serve to prove this point. The labor cuts are no longer counted in numbers, but in percentages. In January, Microsoft reduced its headcount by 1%, in May it was 2. 5% more and throughout this year they will continue laying off additional 4% – totaling about 18,000 employees. [0] Meta is looking to cut its staff of 72,000 by 5%. Intel has let go 15% of their employees, and some smaller companies like Fiverr reduced their workforce by 30%. [1] If this is truly a systemic trend, then the mass increase in unemployment is irreversible. AI will wipe out the working class as a whole. If that happens, tens of millions of people will be left without income. Income they’d need to buy goods, services and assets. The prices of assets will fall because the retail won’t have the money to spend on them. Profits will fall, leading to the stock market crash, pulverizing incomes and net worth of millions. The economy as we’ve known it for decades will be destroyed. This argument isn’t coming from overzealous regulators. It’s coming directly from AI makers themselves. [2] CNN is quoting Dario Amodei, CEO of Anthropic, “one of the world's most powerful creators of artificial intelligence”. [3] 20% unemployment would be Great Depression levels of bad. Extreme poverty, starvation and famine, social unrest… And Amodei isn’t the only one thinking this. That’s OpenAI’s CEO Sam Altman, the superstar behind ChatGPT. [4] Elon Musk is also banking on AI and automation to replace workers in his own factories. As is Amazon, which is speedrunning that trajectory. When Jimmy Fallon asked Bill Gates if we’ll still need humans, he replied: “not for most things”. [5] And it seems like they are selling it as a good thing. That this will somehow “unlock” new potential for all 8 billion people on the planet. [6] But these AI makers obviously have their own skin in the game. They’ve committed trillions into this, so it’s in their best interest to convince everyone that this is the next big thing that’s going to be good for everybody. [7, 8] What should you believe? Is it this all just a marketing hype to boost the stock prices? In this analysis, I won’t be selling you any product whatsoever. There is no sponsorship here. I need you to appreciate that this is truly financially unbiased content. Which means what I am telling you here is what I believe to be true and it is not tainted by sponsor interest. But YouTube is punishing me for refusing to kowtow to the advertiser pressure. Case in point – here’s my last month’s ad revenue. I can’t do this without your support. So if you appreciate my principles and my work, join my Patreon to fight corporate censorship. Unlock unlimited access to my podcast where I dive deeper into these topics in hundreds of long-form episodes. Join the fight against big-money media. This is the most recent report on the impact of artificial intelligence on the job market. And it’s terrifying. It says 97 million US jobs could be replaced by AI and automation in just the next ten years. [9] That means 89% of fast food and counter workers, 83% of customer service reps, 81% of laborers and stock movers, 80% of administrative works, and so on. [9] 97 million workers is just about the entire working class of the United States, which this report says won’t exist in a decade from now. This would be beyond catastrophic. If this report is true, it’s very bad. It was published by the US Congress HELP Committee and that is a big deal. But this report has a big problem – it is not being taken seriously. Why? The HELP Committee used serious data to compile their report – like investor transcripts, corporate financial filings and US economic data. But then they asked ChatGPT to analyze

### Segment 2 (05:00 - 10:00) [5:00]

job descriptions for the whole US economy and predict which tasks could be performed by AI and automation. But ChatGPT told the Committee about all the jobs that are going to be lost to AI. See the problem? If ChatGPT had a PhD, it would never equate tasks with jobs. Because jobs are comprised of many tasks. If an AI can automate some tasks, it doesn’t automatically mean you’ll lose your job. And this report proves why you shouldn’t delegate some tasks to AI to automate your job. Because real academic studies do not support ChatGPT’s conclusions. This is a study by Yale where they conclude that the fact that people are getting fired left and right has nothing to do with AI nor automation but it has probably more to do with cost cutting and boosting quarterly earnings. This study also says that the current trends of changes in the workforce predate the introduction of AI, which was from the first release of ChatGPT in late 2022. [11] But I am not going to pretend jobs aren’t being replaced by AI to at least some extent. So to what extent is this really happening? And is your job really at risk of being automated by AI? This study from Stanford is trying to answer those questions. It looks at the impacts of generative AI on employment. They take young workers aged 22 – 25 and they only find a 6% overall decline in their employment. [12] That’s not nothing, but it’s not the whole picture. Because for older workers, their employment increased by 9% and it was specifically in those jobs most exposed to AI. So clearly not all jobs are equally exposed to being replaced by AI. So what makes the difference? [12] The difference is between jobs that AI automates and jobs that it augments. Because for jobs that AI augments, you know the task vs jobs distinction, those jobs actually see an increase in employment. The difference between augmentation and automation is where delegating tasks either substitutes human labor or complements it. [12] So automated tasks are directly executed by an AI with minimal human involvement. Whereas AI augmented tasks require active human collaboration. As an illustrative example, an automated task would be something like: “Format this technical documentation in Markdown”. While augmented task would be: “Let’s draft a marketing strategy for our new product... Good start, but can we add some concrete metrics? ”. [12] So which jobs are most exposed to automation and which ones are the least? By various metrics, most manual labor will still be performed by humans. Anything from cooks through repair workers, including drivers, nurses or housekeeping. Jobs that will be most exposed to automation will be customer service reps, admin workers, programmers, and chief executives! Going back to the HELP Committee report by ChatGPT, what did the most powerful chatbot get wrong? [9] 89% of fast food and counter workers? No. 83% of customer service representatives? Maybe, but probably not 83%. 81% of laborers and material movers? Nope! Not by any chance! 80% of secretaries and admin assistants? Yeah, they are going to have it rough. 76% of bookkeeping, auditing and accounting workers? Yeah, possible. 76% of stockers and order fillers? May be exposed to automation but not by AI. 66% of office clerks seems unlikely. Augmentation for sure, but not necessarily full automation of office work. 65% of teaching assistants, please no! 64% of auditors and accounts – I don’t know why this is a separate category. But… We are talking about AI that can’t do basic math right. If you are filing your tax returns with a chatbot, the IRS would like to know your location. [9] And where are all these percentages even coming from? Nobody knows. ChatGPT was told to come up with numbers so it came up with numbers. Does it cite a benchmark methodologies like the Stanford study did? Nope! Does it feel like ChatGPT wants to pedal whatever its user wants to hear? Looks like it. But a report like this is adding to the hype. Because it sensational headlines that predict a doomsday scenario, but somehow the only action taken to prevent it is that the stock prices of AI companies go up. Is this a coincidence? The myth of all-powerful AI that will replace us within a decade is the biggest driver behind the hype. But the truth is AI is not ready to replace humans while keeping the same output and success rate. Look at this example: this September, Salesforce slashed 4,000 customer support roles and replaced them with generative AI agents [13] – AIs that can perform various tasks on your device automatically without having you do them yourself. So – glorified chatbots with the ability to hack your phone. The irony is Salesforce’s own study found that these agentic LLM AI agents perform tasks below their very own benchmarks. [14] Their single-turn

### Segment 3 (10:00 - 15:00) [10:00]

performance had only 58% success rate, while completing anything more complex than that collapsed to 35% success rate. [14] 60% of the time it works every time Analysts now predict that over 40% of these agentic AI projects will be canceled by 2027. So sounds like a galactic brain business strategy by Salesforce. [15] Another study benchmarked a bunch of these LLM agents and they found the best of them all could only complete 30% of tasks autonomously. Anything that required longer-horizon focus, AI would completely fall apart. [16] So wherever humans area being replaced by AI, the results are worse than human labor. So companies that are replacing their workers with AI must be comfortable with the fact their product is going to be worse and probably even cost more. The training and development of ChatGPT-3 cost $50 million. GPT-4 cost $500 million. GPT-5 $5 billion. [17] And its release had to be delayed because it wasn’t significantly better than the previous version. [17] Do you want to take a wild guess at how much GPT-6 will cost? If you guessed $500 billion, you are wrong! Sam Altman is already securing $1 trillion in investments. [7] How much better do you think GPT-6 will be compared to GPT-5 or even GPT-4? Just to give you a perspective of what this money could’ve been doing instead… every 10 months, the big tech is spending money equivalent of an Apollo space mission. [18] That’s right, we could’ve gone to the Moon about once a year for the money the big tech spends on bots that tell you to add glue to thicken your pizza dough. And going back to the $1 trillion OpenAI is securing to finance their operation. Why $1 trillion? Well, because OpenAI needs 20 gigawatts of computing capacity to train their models. [7] And one gigawatt sells for about $50 billion on a discount. How much is 20 gigawatts? It’s about a full output of 20 nuclear reactors. [7] So… a lot. It sounds like these costs just keep ballooning over and over… it’s like they are bubbling up…. It’s almost as if… hey there is a bubble To say AI is a bubble is an understatement. It doesn’t really make you grasp how much of a problem this gold-rush fever really is. To compare to the two previous biggest market bubbles, the AI hype is 17 times the size of the dotcom bubble of 2000. [17] And it is four times bigger than the 2008 bubble that let to the global financial collapse. [17] The 2008 crisis was the biggest recession since the Great Depression. You know, the big one, with 30% unemployment, mass starvation, unrest and eventually… a world war. So if you were complaining: “We have no great war, no great depression”… welp, now might be your time! But it’s even worse than this. According to a Harvard economist AI investments are responsible for 92% of US GDP growth in the first half of 2025, but they only account for 4% of the US GDP. Which means that without the AI hype, economic growth of the country would’ve been 0. 1%. [19 – 21] All these jobs that the big tech backed by the government promised to create while building these giant water guzzling data centers… they are standing on the shoulders of the giant historic bubble. Even the AI tech lords realize it. At the center of this bubble are “circular” deals between AI companies. Nvidia lends money to OpenAI, and OpenAI uses that money to buy chips from Nvidia. [22] Nvidia also gives money to data center and cloud computing companies, who are themselves Nvidia’s customers – they buy chips from Nvidia to build their data centers. And what do they do with their data centers? They lend their computing power and data center space to OpenAI, who is again Nvidia’s customer. [18, 22] The circle is complete. Nvidia gives money to their customers to buy products from Nvidia. In the AI industry, everybody is Nvidia customer and Nvidia is everybody’s investor. No surprise then that Nvidia’s stock market price has soared by 1000% since ChatGPT was first released publicly in late 2022. [22] It’s the world’s first $4 trillion company and all they do is sell shovels in the gold rush. [22] If someone can’t afford their shovels, they’ll give them the money to buy their own shovels. This is bad. In S&P 500, the index of the world’s richest companies listed in US stock exchanges, the AI industry has consumed all of it. It’s responsible for 75% of all gains, 80% of all profits, and 90% of all capital expenditure. [8] In other words, if the AI industry goes bust, the whole economy goes bust. So when the music stops, when the bubble bursts, when the mania

### Segment 4 (15:00 - 17:00) [15:00]

hits a brick wall, what will it look like? Well, if history is any good teacher, then three things will happen simultaneously 1. Tens of millions of workers and retail investors will lose their jobs and savings. The value of their incomes and net worth will be pulverized into a fraction of what it was before. 2. Companies that engaged in the circular financing schemes will receive massive bailouts from the tax money of the regular people and from the central banks inflating the currency to protect the “too big to fail” corporate sector. 3. They will all blame immigrants and poor people for this. And so the tech bros and the stinky fear mongering regulators will both be proven right. AI will cost tens if not hundreds of millions of jobs and AI will erase the need for human labor. Just in different ways they expected. What’s the solution for this? Another solution might be Butlerian Jihad. An uprising against the machines, but before they enslave us rather than after. In the adaptations of Dune, Butlerian Jihad is depicted as resistance against “thinking machines”. But in the books, it was more about who controlled those machines that was the problem. The core issue is ownership, not necessarily automation itself. If automated workflows are collectively or democratically owned by the working class, then they can all benefit from the productivity more equitably. That requires profound systemic changes that no serious political debate is willing to even remotely entertain under the current climate. But it could prevent the biggest harms of the current industrial transformation. Naturally, this isn’t anything the AI tech lords would do of their own accord. They are betting that they will all be bailed out from their own failures and that minorities will be blamed for it. The best way to immunize yourself against automation is to be able work jobs with complex tasks with long focus or complicated manual labor. Automation is a legitimate threat to job safety, but it it’s not coming in the form of chatbots and LLMs. Automating workflows in factories has been done for decades and it’s not going to stop anytime soon. But again, this is much less of an issue if you have a stake in that automation. So if you made it till the end, you are a legend! Spread the word, join my Patreon and don’t be enslaved by the machines or those who own them!