The AI Takeover Has Arrived

...but it looks completely different from what we imagined

Drone surveillance. Image by flysight

Hah! You thought that one really dark day AI would suddenly rise to consciousness, take over the world and turn us into paper clips—or worse: battery cells? Wrong! It won’t do any of that. It doesn’t need to. Before going into details why is that so, however, I have to start with an admission. I seriously underestimated AI and the speed of its development. I always thought that large language models are just that: blathering machines with nothing useful to say—other than hallucinating summaries and images of what real people once wrote, said or made pictures of. The past few months have really changed my worldview, though. I’ve seen demonstrations, presentations, participated in round table discussions on the topic and learned a lot. Things I imagined never to become real or possible have been revealed to me. Even AI experts had to confess they underestimated the speed with which these changes have arrived. The transformation the world is going through is both truly remarkable and frightening at the same time. It’s no exaggeration to say that we have arrived at a juncture in human history, and our future looks more uncertain than ever.

Thank you for reading The Honest Sorcerer. If you value this article or any others please share and consider a subscription, or perhaps buying a virtual coffee. At the same time allow me to express my eternal gratitude to those who already support my work — without you this site could not exist.

The Paper Clip Fallacy

Let’s start our journey discovering what an AI future has in store for us with a popular story among existential risk analysts—because, well yes, there is such a thing. It goes like this:

“When the world ends, it may not be by fire or ice or an evil robot overlord. Our demise may come at the hands of a superintelligence that just wants more paper clips.”

Nick Bostrom

In 2014, Mr Bostrom, a philosopher who founded and directs the Future of Humanity Institute at the University of Oxford, has created a thought experiment, which he called: “the paper-clip maximizer.” It works as follows:

“Imagine a machine programmed with the seemingly harmless, and ethically neutral, goal of getting as many paper clips as possible. First it collects them. Then. realizing that it could get more clips if it were smarter, it tries to improve its own algorithm to maximize computing power and collecting abilities. Unrestrained, its power grows by leaps and bounds, until it will do anything to reach its goal: collect paper clips, yes, but also buy paper clips, steal paper clips, perhaps transform all of earth into a paper-clip factory. “Harmless” goal, bad programming, end of the human race.”

As many other philosophical thought experiments it sounds plausible… Until you examine the core assumptions and realize that nothing of the sort could ever come to life. Yet, we are not safe either… As someone who worked 20 years in the world of manufacturing, supply chains and product engineering, I always felt obliged to disassemble this argument, but somehow never found the time and space to do so. Recent developments, I alluded to above, however, have forced me to do so.

First, you don’t program a machine to get as much paper clips as it can. You just don’t. Not even for fun. Companies, with ample resources to build and operate such a machine want to make profits—not paper clips. (And trust me, you wouldn’t feel better when your room would start to fill up with clips, either.) See, if you inundate the market with paper clips (or whatever) it’s price naturally falls, unless you are willing to invest into a massive warehouse storing those widgets (thereby starving the market). Either way you end up owning a huge investment, with no profits to show for it. A surefire way to bankruptcy.

Second, the world, the biosphere and the human economy in it, is full of feedback loops. Such thought experiments ignore all of that. That’s what makes them so sexy and attractive: they are simple, and contrary to their name, actually don’t require giving much thought to them. Otherwise, they would fall apart. Just as the paper-clip maximizer cited above. Should such an algorithm came into existence—against all odds—it’s activity would send metal and other raw material prices soaring. Shortages would appear, and people would suddenly notice that a crazy AI agent have started to steal, buy up and manufacture a ton of paper clips. What if companies then stopped delivering what the paper clip monster wanted? What if it could no longer do what it does because it runs out of a critical component? Would it start a war? Of course not. Or is it…?

Paper clips… Lots of paper clips. Image via Unsplash

The Real Risk

This little thought experiment, deeply flawed as it is, still points out the real danger in implementing AI: the utter pointlessness of it all. Let me present you with a real thought experiment based on what I’ve learned recently. Every trade in the knowledge business from engineers to marketing specialists and from lawyers to programmers are being asked to “use more AI, because that’s what makes you more productive.” Put more bluntly: “No AI—no job.” If you don’t use AI to produce more, better, faster, then we have no use of you. Teach the AI bots how to do your job, or risk losing your work because you’re not using it. And since the company next door thinks and says the same, they’ll have no use of you either. So here is the thought experiment:

You are asked to automate, speed up, and make your work more productive, which is fine. You are told that reading and replying to e-mails, filling in excel sheets, making power point presentations, coding etc. is no longer what is expected from you. Instead, you are expected to train AI agents to do all that, and become an orchestrator of an army of virtual robots, AI assistants, automated processes etc. It follows, that this activity should free up a considerable amount of your time, hence the increase in productivity, which is also fine. But here comes the conundrum—what should you do with that time exactly? Create more dashboards, scorecards, presentations? I guess not, because beyond a certain point that would take too much of other people’s time to read and consume—or use up more AI compute to process—which would kill the entire productivity idea. OK then, if you have so much time left, why not produce more code, applications, widgets, text, or whatever output you’re paid to generate?

And herein lies the problem. What you are in fact expected to do as a manager of AI agents, is to do more with less. What used to take a bunch of engineers to design, or an entire programming team to make, now just takes one person running a set of agents. This is what productivity increase (without wage increase) actually means. At best you don’t hire anymore, just ask your smartest guy to train an agent. At worst, you begin to ‘streamline’ your team. Because, hey, you cannot expect the world to buy twice as much widgets, products, software code etc. just because your team uses AI… Using it will just make your company, selling all that stuff, more competitive. And if the rivaling company does the same (and you’re told they definitely do the same) then all this great AI adoption will achieve are two things: a reduction in workforce—and therefore consumption—and a reduction of prices. Both of them anathema to capitalism.

The moment you realize you taught AI how to do your job. Image via Unsplash

One of the key reasons companies so eager to implement AI is that they see a steady decline in overall consumption. Well, who could’ve expected that? Not raising the wages in line with inflation (caused by the supply shock of COVID, skyrocketing energy prices, then the coming shock from the lack of West Asian oil), has forced people to spend less. Even China has seen a fall in consumption, forcing them to export more… This is not a situation when people spend more on cars, fridges, home renovation projects etc. —hence the record lows in consumer sentiment. Let’s face it: the economy is no longer working for the 90% of consumers already, and the race to replace each and every one of us with AI has just begun.

‘Okay—one might say—some people might be laid off, but at least prices will fall, thanks to automation and AI. Isn’t that good?’ Well, not quite. If demand gets sufficiently destroyed—either due to high energy and food prices, due to lay-offs or both—prices will fall anyway. As people increasingly become unable to consume as much as they used to, companies will be forced to lower their prices to stay in competition. Implementing AI in this economic environment—while seemingly a great idea to save on costs—will just accelerate the process and ultimately turn into a self-defeating proposition. Firing high-salaried staff, whose spending has kept consumption more or less afloat so far, could all too easily end up in a self-reinforcing feedback loop: a deflationary spiral. A very hard to escape economic situation, where falling prices lead to lower wages, job losses, and delayed consumer spending—just like during the 1930’s.

It’s hard to miss the irony here: while a few months back most investors were afraid that the AI bubble would burst because of bad economics, things might turn out to be the other way around: The economy might crash because of a reckless implementation of AI.

What Comes Next?

This brings us to the ultimate question: what is the endpoint of AI adoption? If the goal of capitalism is to make more profits and to accumulate more and more capital, then AI is a self-defeating proposal. Not because AI companies cannot make a profit and operate at a huge loss, but because AI would destroy the very basis of capitalism itself: consumption. And it really doesn’t take too much: companies doesn’t have to send people away in droves. A couple of percentage point increase in unemployment is more than enough to tip our fragile economies into recession. Combined with the immense squeeze coming from the closure of Hormuz, and the lack of oil, fertilizer, sulfur, aluminum, helium etc., we are definitely looking into a word of immense hurt.

But hey, let’s play this out, shall we?

1. Companies, under immense investor and peer pressure, implement AI at an increasing scale. They ask their employees to automate their own work, then fire two, three, five or even ten percent of their highly paid workforce. Marketing specialists, project managers, programmers, data analysts find themselves without a job and see their chances of finding another one vanishingly low.

2. Consumer sentiment and spending falls further, faster. Consumption in the top 10% of the society begins to fall as well. Only the super-rich spend as they used to.

3. Inflation, just like the economy before, begins to take a K-shape form. The price of essentials keeps rising as a lack of diesel fuel and fertilizer pushes food prices ever higher and as material shortages arrive from the Hormuz closure. Meanwhile the cost of discretionary consumer goods begin to fall as companies face rising inventory levels, plummeting demand and a fierce competition from China.

4. Manufacturing companies now find themselves in a double whammy: rising input costs from scarcity and price pressure from competitors and a weak market. Many find no other alternative than to double down on AI and on automation, then continue with firing their workforce. Others simply go bankrupt. Not every company, everywhere, all at once, but enough of them to keep unemployment high and rising.

5. Consumption of consumer products and cars falls even further, despite another round of discounts on the shelves. The crisis begins to feed on itself.

6. Borrowing slows to a bare minimum as people defer their major purchases. A lot of people also fail to pay their mortgages, auto loans, electric bills etc. in time. The crisis now threatens to trigger a major banking crisis.

7. Governments and central banks around the world do everything in their power to halt the crisis on its tracks: they issue stimulus checks, declare a moratorium on debt payments, lower interest rates to negative territories, bail out failing banks—but ultimately to no avail.

8. The stock and housing market crashes, as markets run out of buyers who still believe “the economy is booming” and it’s worth investing into.

9. The economy shrinks 10-15% to match the new supply-demand balance created by a lack of vital physical inputs on one side, and a lack of purchasing power on the other. Prices, trade flows and production volumes eventually stabilize, but at a permanently lower level. And while growth might eventually return, it will be wholly inadequate to raise economic activity back to its pre-war, pre-AI levels. Historians looking back on 2025 will say: that was the year when we passed peak global economic activity.

10. Unemployment stays stubbornly high, and once wealthy citizens—now permanently out of their well paid jobs—start to grumble. The loss of meaning, a stable job once meant, will only compound matters further. A lot of people will be mightily pissed and will come up with all sorts of “wacky” ideas. And you know what happens next… Whenever in doubt what to do about the economic malaise you navigated your country into, start a war. The bigger, the better.

Nothing said above is set into stone: this is not a prediction, but a warning. Many things can change, and will change for sure. It is still not too late to stop the worst from happening. With our current crop of trapped, mentally challenged, ideologically misguided leaders, however, I see very little chance for things to turn out for the best, though. So, while we were waiting for a self-conscious AI to emerge and to start a war on humanity, the reckless application of its current iteration has put us on track to ruin the economy, and likely facilitated the start of another world war.

As usual, I can but hope I’m wrong in all this.

Until next time,

B

Thank you for reading The Honest Sorcerer. If you value this article or any others please share and consider a subscription, or perhaps buying a virtual coffee. At the same time allow me to express my eternal gratitude to those who already support my work — without you this site could not exist.

P.S: Here is another set of ironies to ponder on—if this whole affair were not absurd enough. AI data centers burn through finite reserves of natural gas and non-renewable water supplies in the desert, both of which we have no idea where to supply from as early as 2030. The math, as Chris Martenson would say isn’t mathing. The only logical consequence is that AI soon will require water and electricity to be diverted from somewhere else… This lack of energy and resources to keep both our high-consumption lifestyles and a still accelerating AI build-out going will create increasing tensions as the system optimizes for survival. No wonder governments around the world pour countless millions into AI based surveillance: not to create prosperity, but to protect themselves from popular backlash and resistance to war. Data centers will thus be increasingly labelled as ‘critical to national security’ and get their energy demands met, consequences be damned.

Besides the US wantonly starting wars around the globe, the drums of war are also being beaten ever louder in Europe, the continent most heavily affected by the series of shocks during the past six years. Germany now openly prepares for war, alongside Baltic and Scandinavian countries, citing 2029 as the most likely year Russia will attack a NATO country. And with the Russians now openly threatening to strike Europe, in retaliation for the facilitation of drone and missile strikes deep inside its territory, a re-run of WWII is rapidly becoming a question of when, not if… I can but hope that we will never get to this point, as the risk of this potential conflict going nuclear is higher than any time in history—especially with hard liners now openly demanding a preemptive nuclear strike on Europe. Sadly, the motivation will only growing stronger to start ‘something’ on both sides as their economies will continue to falter under years of pressure.