Last week I published a featherweight narrative about applying GenAI in a real-world context, to a tiny programming problem. Now I’m regretting that piece because I totally ignored the two central issues with AI: What it’s meant to do, and how much it really costs.
What genAI is for · The most important fact about genAI in the real world is that there’ve been literally hundreds of billions of dollars invested in it; that link is just startups, and ignores a comparable torrent of cash pouring out of Big Tech.
The business leaders pumping all this money of course don’t understand the technology. They’re doing this for exactly one reason: They think they can discard armies of employees and replace them with LLM services, at the cost of shipping shittier products. Do you think your management would spend that kind of money to help you with a quicker first draft or a summarized inbox?
Adobe said the quiet part out loud: Skip the Photoshoot.
At this point someone will point out that previous technology waves have generated as much employment as they’ve eliminated. Maybe so, but that’s not what business leaders think they’re buying. They think they’re buying smaller payrolls.
Maybe I’m overly sensitive, but thinking about these truths leads to a mental stench that makes me want to stay away from it.
How much does genAI cost? · Well, I already mentioned all those hundreds of billions. But that’s pocket change. The investment community in general and Venture Capital in particular will whine and moan, but the people who are losing the money are people who can afford to.
The first real cost is hypothetical: What if those business leaders are correct and they can gleefully dispose of millions of employees? If you think we’re already suffering from egregious levels of inequality, what happens when a big chunk of the middle class suddenly becomes professionally superfluous? I’m no economist so I’ll stop there, but you don’t have to be a rocket scientist to predict severe economic pain.
Then there’s the other thing that nobody talks about, the massive greenhouse-gas load that all those data centers are going to be pumping out. This at a time when we we blow past one atmospheric-carbon metric after another and David Suzuki says the fight against climate change is lost, that we need to hunker down and work on survival at the local level.
The real problem · It’s the people who are pushing it. Their business goals are quite likely, as a side-effect, to make the world a worse place, and they don’t give a fuck. Their technology will inevitably worsen the onrushing climate catastrophe, and they don’t give a fuck.
It’s probably not as simple as “They’re just shitty people” — it’s not exactly easy to escape the exigencies of modern capitalism. But they are people who are doing shitty things.
Is genAI useful? · Sorry, I’m having trouble even thinking about that now.
Contributions
Comment feed for ongoing:
From: Rob (Jul 06 2025, at 12:56)
"It’s probably not as simple as “They’re just shitty people” — it’s not exactly easy to escape the exigencies of modern capitalism. But they are people who are doing shitty things."
They said the same thing about bitcoiners, another technology with no real world use-case. Bitcoiners have extremely clearly and unambiguously established themselves as shitty people, but look, its been what 15 years and there they are, still ponzi-ing away.
LLM people are just more shitty people with yet another technology with no real world use-case. They just have a lot more money to burn, so they can do a lot more damage, for a lot longer.
For them, fundamentally its all about substituting "whatever" because all glop/content is fungible, fuck "Truth" or "Beauty." And it'll be Gresham's Law all over again, except for information instead of money. I expect we are going to see Weimar levels of information inflation pretty shortly, we're going to drown in it.
While the world burns
[link]
From: tbray (Jul 06 2025, at 13:01)
what happened to you? did google ruin you? was it android? was it vancouver? geez. talk about a maniacal and controlling luddite with a blog. practice what you preach and throw your pc away.
[link]
From: Martyn Hare (Jul 06 2025, at 14:22)
I personally hope that open-source local models become “good enough” quick enough to make the bottom fall out for these planet destroying cloud services. Once that happens, the investment money will rapidly dry up, and the true economic costs of offshoring human labour to poorly tested, nondeterministic, unspecialised black boxes will be revealed. It is only a matter of time, the question is when.
We’ve already seen a scummy “buy now, pay later” company try replacing large chunks of their workforce with so-called AI, and they were forced to reverse course after it almost destroyed them. Teachers, lawyers, judges and even police officers have been shamed when caught publishing important documents unchecked. Microsoft screwed the pooch with GitHub Copilot as a programming tool to the point where it’s now a legal liability for well-known software projects to use. People are already getting good at spotting glitches in generative images, including video when viewed in situations outside of scummy social networking services. Even video gamers are spotting generated slop with ease when it’s being shown to them, and are very quickly rejecting it.
At best, by the end of the hype cycle, we will end up with improvements upon already highly specialised models for a variety of tasks humans struggle to deal with in bulk (e.g. interpreting medical imaging reliably, categorising physical objects, transcribing and translating conversations at near real time speeds, tracking individuals for security purposes).
What we won’t have are these unsustainable cloud services providing all you can eat access to unspecialised models. We are already seeing the beginning of the end with ever increasing charges for “credits” (on top of monthly subscription fees) for those who do try to use these services for bulk processing; which essentially destroys any cost savings businesses could have otherwise made.
[link]
From: D (Jul 06 2025, at 15:09)
Hi Tim,
I think you've mistaken what we have today for capitalism, when objectively it really isn't.
Capitalism requires a number of things, and one of which is strong rights of ownership, which put quite simply is having the right to dispose of your property for nothing. We do still own some small things, but any large asset is for the most part not owned except in some very rare circumstances.
The other missing requirement is adversarial independent decision-making without cooperation in the market.
What we have in most things today is control of resources without real ownership, and there are so-called private companies but they are equivalent to state-run apparatus having received loans from a money-printer.
You can't dispose of a property or buildings, while you have a mortgage, and you generally must keep a mortgage on leverage for a number of reasons related to the money-printing economy (fiat) to avoid recapture, or hostile takeovers/leveraged buyouts which sieve the market into fewer and fewer participants in a long-term deflationary cycle.
If you don't, you lose money from currency debasement, and other factors and risks. So you don't really own things, and they can be taken away at any time.
...
The problem with AI that no one is talking about is the sequential nature of the career development pipelines, and the costs of compromised communication infrastructure.
AI in effect has already disrupted factor markets so almost no new candidates are being hired at the bottom rung, setting a 10 year clock on collapse (2-3 years ago), and increased the cost for all parties by jamming communications with fake jobs and fake candidates (80% of Indeed + many others).
You don't have to be an economist to know that something that eliminates capital formation for 60% of the economy (white collar work), and prevents all legitimate business constrained by profit/loss functions from finding talent; will not somehow end badly, especially when industry for strategic goods (i.e. food production), fail to single points of failures which were optimized towards failures by profit driven intent.
The value of fiat money disappears when it loses its properties as a medium of exchange/stable store of value, and that happens when workers can't exchange their time for money within certain economic limits and refuse to transact or store their value in the currency.
Artificial chaotic whipsaws that continue to grow become common as economic calculation fails.
Mises wrote quite extensively about the failures of Socialism as a structure back in the 1950s, defining several core structural problems that are quite unsolvable and remain un-refuted today.
He focused quite a bit on the failures of non-market socialism, and we have markets; except they are no longer true markets in the ways that matter.
Money-printing through non-reserve debt issuance by the banks (circa 2020 with the Basel3 transition) has artificially distorted the markets; and insiders have restructured the markets so price-discovery can no longer effectively occur (i.e. dark pools, PFOF, and broker price improvement... which are Orwellian). There recently was significant failures to deliver in commodities like gold and silver showing capital flight happening in the commodities market. Paper warrants were overissued (printed) by same said banking institutions exceeding a 300:1 ratio in the paper market. There has been a silent run on commodities starting early this year where participants are no longer transacting through paper warrant but requiring physical delivery.
The only companies that can continue forward once profit cannot be sustained are ones that have tied themselves to money-printing which doesn't take a big leap to see how that collapses to non-market socialism forms.
AT&T is a perfect example of such state-sponsored apparatus; they have government loans that are less than inflation/debasement that run out 100+ years. The Fed has been intervening as well, delegating bond swaps and auctions to Blackrock, and many other things.
We currently seem to be about 2-3 years into that 10 year cycle. A good amount of people leave their careers every 10 years, and with pipelines: 0 in, 0 out, that though is a problem for next quarter; something that doesn't get reacted to until its too late when money-printing guarantees short term stability.
While the intermediate and senior experts will continue for awhile, they have no one to pass the torch to, they burn out, age out, retire, die, and what they know becomes lost knowledge because its not economically valuable to pass on in the current environment. This will be true of everything touched by AI.
There are people in control of the money supply that have fallen to fallacy on the oldest lie in the book. You can't continue to print yourself out of a debt trap crisis, and every time someone tries; it leads to political, sometimes revolutionary upheaval especially when the rule of law no longer meets its component requirements.
While this is quite slow and gradual, all the incentives force the dynamics forward as a runaway positive feedback system that can't be stopped, and all the smart people that are and were in a position to know these things have been shouting fire but they are being drowned out by noise; so no collective action can take place to avert this before a point of no return.
This is why most of the smart people are starting to emigrate out of the country.
Hopefully you find this contribution useful. You don't really have to be an economist to know when the basic assumptions of an economy fail, everything else built on top of it falls down with it.
Things are going to get quite bad in the near term. Be safe.
[link]
From: Sprewell (Jul 07 2025, at 03:07)
Don't worry about it: not only are they greedy and wasteful, but they are stupid, ie it has been known for decades that these statistical neural network models are fundamentally flawed and unreliable. This AI bubble is going to come crashing down as more and more people realize that.
Just pray that our "glorious leaders" don't take us to war again to hide from the economic wreckage about to ensue from their idiocy.
[link]
From: ForeverYoung (Jul 07 2025, at 04:06)
Remember the dot-com bubble? Theranos? How Apple was going to build a car?
What is the likelihood of Silicon Valley simultaneously solving its viability, lack of infrastructure, high cost of living, and capital allocation problem all at once? It would be less than pretending to do so.
Karen Hao just released a book on the AI industry, "Empire of AI". There are interviews with her on YouTube.
[link]
From: Robert Sayre (Jul 07 2025, at 15:36)
I find that it's great for getting unstuck or staying in the flow state. Interruptions become less bothersome.
I'll let it write a function, and then begin correcting it. Way faster than spinning up on all of the details I might have forgotten over lunch.
[link]
From: Ezra (Jul 07 2025, at 17:52)
Completely agree with the substance, though I have to quibble with the "Real Issue" framing. There are many real issues with GenAI! In particular, I think business leaders' goals are loftier than "merely" eliminating labor. The megacorps also want to be in the dependency path for all knowledge-based productivity, the same way their cloud services have made them necessary for most internet-based productivity. They also also want control over the information environment - we've seen this with Musk's heavy-handed racism being clumsily encoded into Grok, and there's no reason to expect that the other tech billionaires aren't doing the same kind of thing more subtly and competently.
[link]