With LLMs not only do we see massive increases in overhead costs due to the training process necessary to build a usable model, each request that gets sent has a higher cost. This changes the scaling logic in ways that don’t appear to be getting priced in or planned for in discussions of the glorious AI technocapital future

This is a very important point, I believe. I find it particularly ironic that the "traditional" Internet was fairly efficient in particular because many people were shown more or less the same content, and this fact also made it easier to carry out a certain degree of quality assurance. Now with chatbots, all this is being thrown overboard and extreme inefficiencies are being created, and apparently, the AI hypemongers are largely ignoring that.

It's quite noteworthy how often these shots start out somewhat okay at the first prompt, but then deteriorate markedly over the following seconds.

As a layperson, I would try to explain this as follows: At the beginning, the AI is - to some extent - free to "pick" how the characters and their surroundings would look like (while staying within the constraints of the prompt, of course, even if this doesn't always work out either).

Therefore, the AI can basically "fill in the blanks" from its training data and create something that may look somewhat impressive at first glance.

However, for continuing the shot, the AI is now stuck with these characters and surroundings while having to follow a plot that may not be represented in its training data, especially not for the characters and surroundings it had picked. This is why we frequently see inconsistencies, deviations from the prompt or just plain nonsense.

If I am right about this assumption, it might be very difficult to improve these video generators, I guess (because an unrealistic amount of additional training data would be required).

Edit: According to other people, it may also be related to memory/hardware etc. In that case, my guesses above may not apply. Or maybe it is a mixture of both.

I have been thinking about the true cost of running LLMs (of course, Ed Zitron and others have written about this a lot).

We take it for granted that large parts of the internet are available for free. Sure, a lot of it is plastered with ads, and paywalls are becoming increasingly common, but thanks to economies of scale (and a level of intrinsic motivation/altruism/idealism/vanity), it still used to be viable to provide information online without charging users for every bit of it. Same appears to be true for the tools to discover said information (search engines).

Compare this to the estimated true cost of running AI chatbots, which (according to the numbers I'm familiar with) may be tens or even hundreds of dollars a month for each user. For this price, users would get unreliable slop, and this slop could only be produced from the (mostly free) information that is already available online while disincentivizing creators from producing more of it (because search engine driven traffic is dying down).

I think the math is really abysmal here, and it may take some time to realize how bad it really is. We are used to big numbers from tech companies, but we rarely break them down to individual users.

Somehow reminds me of the astronomical cost of each bitcoin transaction (especially compared to the tiny cost of processing a single payment through established payment systems).

Is it that unimaginable for SV tech that people speak more than one language? And that maybe you fucking ask before shoving a horribly bad machine translation into people’s faces?

This really gets on my nerves too. They probably came up with the idea that they could increase time spent on their platforms and thus revenue by providing more content in their users' native languages (especially non-English). Simply forcing it on everyone, without giving their users a choice, was probably the cheapest way to implement it. Even if this annoys most of their user base, it makes their investors happy, I guess, at least over the short term. If this bubble has shown us anything, it is that investors hardly care whether a feature is desirable from the users' point of view or not.

I'm not sure how much this observation can be generalized, but I've also wondered how much the people who overestimate the usefulness of AI image generators underestimate the chances of licensing decent artwork from real creatives with just a few clicks and at low cost. For example, if I'm looking for an illustration for a PowerPoint presentation, I'll usually find something suitable fairly quickly in Canva's library. That's why I don't understand why so many people believe they absolutely need AI-generated slop for this. Of course, however, Canva is participating in the AI hype now as well. I guess they have to keep their investors happy.

What fascinates me is why coders who use LLMs think they’re more productive.

As @dgerard@awful.systems wrote, LLM usage has been compared to gambling addiction: https://pivot-to-ai.com/2025/06/05/generative-ai-runs-on-gambling-addiction-just-one-more-prompt-bro/

I wonder to what extent this might explain this phenomenon. Many gambling addicts aren't fully aware of their losses, either, I guess.

... and just a few paragraphs further down:

The number of people tested in the study was n=16. That’s a small number. But it’s a lot better than the usual AI coding promotion, where n=1 ’cos it’s just one guy saying “I’m so much faster now, trust me bro. No, I didn’t measure it.”

I wouldn't call that "burying information".

there’s no use case for LLMs or generative AI that stands up to even mild scrutiny, but the people funneling money into this crap don’t seem to have noticed yet

This is why I dislike the narrative that we should resist "AI" with all our power because supposedly, if our employers got us to train the chatbots, they would become super smart and would be able to replace us in no time. In my view, this is simply not true, as the past years have shown. Spreading this narrative (no matter how well-intentioned) will only empower the AI grifters and reinforce employers' beliefs that they could easily lay off people and replace them with slop generators because supposedly the tech can do it all.

There are other very good reasons to fight the slop generators, but this is not one of them, in my view.

I'm old enough to remember the dotcom bubble. Even at my young age back then, I found it easy to spot many of the "bubbly" aspects of it. Yet, as a nerd, I was very impressed by the internet itself and was showing a little bit of youthful obsession about it (while many of my same-aged peers were still hesitant to embrace it, to be honest).

Now with LLMs/generative AI, I simply find myself unable to identify any potential that is even remotely similar to the internet. Of course, it is easy to argue that today, I am simply too old to embrace new tech or whatever. What strikes me, however, is that some of the worst LLM hypemongers I know are people my age (or older) who missed out on the early internet boom and somehow never seemed to be able to get over that fact.

As I mentioned before, some spammers and scammers might actually need the tech to remain competitive in their markets from now on, I guess. And I think they might be the only ones (except for a few addicts) who would either be willing to pay full price or start running their own slop generators locally.

This is pretty much the only reason I could imagine why "AI" (at least in its current form) might be "here to stay".

On the other hand, maybe the public will eventually become so saturated with AI slop that not even criminals will be able to use it to con their victims anymore.

I don't understand. Everybody keeps telling me that LLMs are easily capable of replacing pretty much every software developer on this planet. And now they complain that $71 a day (or even $200 a month) is too much for such amazing tech? /s

In my experience, copy that "sells" must evoke the impression of being unique in some way, while also conforming to certain established standards. After all, if the copy reads like something you could read anywhere else, how could the product be any different from all the competing products? Why should you pay any attention to it at all?

This requirement for conformity paired with uniqueness and originality requires a balancing act that many people who are not familiar with the task of copywriting might not understand at all. I think to some extent, LLMs are capable of creating the impression of conformity that clients expect from copywriters, but they tend to fail at the "uniqueness" part.

maybe they’ll figure a way to squeeze suckers out of their money in order to keep the charade going

I believe that without access to generative AI, spammers and scammers wouldn't be able to successfully compete in their respective markets anymore. So at the very least, the AI companies got this going for them, I guess. This might require their sales reps to mingle in somewhat peculiar circles, but who cares?

It's almost as if teachers were grading their students' tests using a dice, and then the students tried manipulating the dice (because it was their only shot at getting better grades), and the teachers got mad about that.

This is, of course, a fairly blatant attempt at cheating. On the other hand: Could authors ever expect a review that's even remotely fair if reviewers outsource their task to a BS bot? In a sense, this is just manipulating a process that would not have been fair either way.

To me, the idea of using market power as a key argument here seems quite convincing, because if there was relevant competition in the search engine market, Google would probably have had much more difficulty imposing this slop on all users.

I disagree with the last part of this post, though (the idea that lawyers, doctors, firefighters etc. are inevitably going to be replaced with AI as well, whether we want it or not). I think this is precisely what AI grifters would want us to believe, because if they could somehow force everyone in every part of society to pay for their slop, this would keep stock prices up. So far, however, AI has mainly been shoved into our lives by a few oligopolistic tech companies (and some VC-funded startups), and I think the main purpose here is to create the illusion (!) of inevitability because that is what investors want.

Completely unrelated fact, but isn't the prevalence of cocaine use among U. S. adults considered to be more than 1% as well?

(Referring to this, of course - especially the last part: https://pivot-to-ai.com/2025/06/05/generative-ai-runs-on-gambling-addiction-just-one-more-prompt-bro/)

Stock markets generally love layoffs, and they appear to love AI at the moment. To be honest, I'm not sure they thought beyond that.

Yes, they will create security problems anyway, but maybe, just maybe, users won’t copy paste sensitive business documents into third party web pages?

I can see that. It becomes kind of a protection racket: Pay our subscription fees, or data breaches are going to befall you, and you will only have yourself (and your chatbot-addicted employees) to blame.