Hetare King

Not every software company is a game developer, and even at home, not every programmer plays games on their PC that require a powerful GPU. If everyone working on, for example, business software all of a sudden needs a gaming PC that will be constantly running an AI model, that will greatly increase costs, both in terms of procurement of equipment and electricity. An increase in demand for powerful GPUs in fields that traditionally had no need for them isn’t exactly going to improve the hardware cost situation either.

Whether you run the model locally or you leave it to a company, someone is going to have to pay the cost.

What you’re saying feels kind of circular. Regulations and their enforcement being lacking doesn’t have any explanatory power for people’s behaviour, when those same people are actively trying to prevent regulation from happening. To begin with, someone with a shred of morals wouldn’t need regulations to try to minimise harm to their employees’ health on the work floor.

Also, lawmakers doing the bidding of evil people doesn’t make the lawmakers any less evil; it’s not as though their children are being held hostage or something.

I’m not really seeing any logic that here doesn’t ultimately boil down to “they’re evil”. Of course, being aware of the problems with current regulations is important and can lead to practical solutions that can improve people’s lives and be implemented in the relatively short term, but focussing exclusively on that obscures the more fundamental problem: that our society is set up so that greedy, unempathetic bastards float to the top. And that too, requires a solution.

I remember the CEO of NVIDIA making similar claims like 10-15 years ago (and being rather eager for that future to come), but I’m not really seeing anything that has changed since then that makes it more likely now.

There’s the increases in hardware costs, yes, but in order to get sufficiently low latency, you can’t just plop one or a few big data centers per country, you need at least one smaller one pretty much per town. And to ensure a good quality of service, those are going to need more redundant hardware than a more centralised solution would, and given that cloud gaming service providers are also affected by hardware getting more expensive, that would make for some very high up-front costs.

There’s the possibility of VCs throwing dumb money at these companies until they control the market, but for that to work, they need to actually be able to take control of the market. But they’re not just competing against the latest and greatest, they’re also competing against ten year old hardware that’s considerably more affordable. Things aren’t advancing as quickly as they used to, and I doubt the shrinking number of people who do always want the best possible are going to be too thrilled about having graphics littered with compression artifacts.

For small numbers of blocks, this would be more space-efficient, because it has no overhead and stores less information per block. However, Minecraft’s file format actually already supports compression, so once things get to a certain size, it would probably start catching up (depending on how structured the blocks are in relation to each other).

More importantly though, this program can only produce contiguous square pillars, so there are not a whole lot of practical uses for it (of course, this is not the goal of this project to begin with). Once you start adding in the complexities to make it useful, you’re basically reinventing Minecraft’s own file format, with similar space-efficiency.

And of course, there’s the matter that normally, Minecraft’s worlds are procedurally generated from a seed, so it doesn’t actually need to save all the blocks that make up a region, just the changes the players made to it. But this is something that’s impossible to do when storing a region as just a sequence of blocks.

It just reads out the raw bytes from the file and maps each possible byte value to a Minecraft block, which means that it works with any file. The flipside of this is that it’s rarely going to produce anything other than a garbled mess.

Not quite the same, but there is a game called Zwei!! (German for two). Its sequel is just called Zwei II, though.

Huh, that’s interesting. I already expressed my thoughts on consumers making their own physical media in another post (tl;dr: I don’t think it makes much sense), but if there are recordable discs with lifespans comparable with pressed discs, maybe a print-on-demand service could be a solution for developers that can’t afford to order a batch of a thousand discs and cases and rely on digital distribution because of that.

It feels like it would kind of defeat the purpose. If you like the ritual of physically picking a game and putting it into your machine, there are more practical ways of going at it, like those NFC card-based systems like Zaparoo. If you like having a physical collection, it seems to me that having a bookcase full of labels made with a home printer and with none of the legitimacy of them being “the real thing” would feel rather empty. You can’t legally resell it or lend it to others and if you care about preserving the game without being dependent on an external service that can go down at any time, you’re probably better off keeping several backups of your hard drive than counting on flaky recordable media (hard drives don’t last forever either, but you’re more likely to regularly check on the health of a couple of hard drives than potentially hundreds of BD-Rs and SD cards).

Basically, very few of the advantages of physical media would end up remaining.

All things decay, not even the doping of ROM chips will last forever, but I think the average lifespan of recorded optical media is like, 10 years? That feels rather short.

Doesn’t recordable optical media also have a pretty limited lifespan? Unlike commercially produced discs, where the pits are pressed into the plastic, CD/DVD/BD-Rs just have a dye that is made to change colour with a laser, and that dye degrades over time.

I always have a hard time putting things into rankings, because different games have different qualities that make them hard to compare to each other, but if I had to choose, there is one game that I keep coming back to and is pretty much my “comfort food” game: Twinkle Star Sprites.

On one hand, from what I understand, in the 1983 crash there was a real mood of video games being over, that the fad had passed. In North America, anyway. I don’t think there’s any risk of that happening today. On the other, the video game industry is a lot larger now, so there’s a lot more shrinking that can happen without it completely destroying itself. So in that sense it may actually be bigger.

That said, I’m inclined to believe that by laying off all these people, big companies are just creating their own future competition, because there’s no shortage of demand for video games. If anything there’s a greater demand for games that aren’t constrained by the arrogance of big company executives.

OK, so thus far we’ve had two major AI winters, where after some excitement over new developments in AI, funding ended up more or less drying up for decades, because it didn’t manage to go anywhere. I think a third such an event would go a long way in mitigating the harms from AI, because the AI space is currently very centralised, i.e. most users are very dependent on a handful of companies, and those companies are very dependent on continued investment. Also, running less capable open weights/source models locally has a fairly high barrier to entry and those models would become less relevant over time without the funding to keep up their training.

Thankfully for me, things appear to be moving in that direction.

As for regulation, I don’t think I mentioned or alluded to it at all. If anything, that seems more your thing, since you think holding people doing bad things with AI accountable is sufficient. Personally, I do very much think that the world would be better off without these forms of AI. But if there is to be regulation, I would argue it should focus mostly on the suppliers and not so much on the users. Not only because otherwise you end up in an unsustainable game of whack-a-mole, but also because there’s not always a user to hold to account. Like, if an AI chatbot encourages someone’s suicidal thoughts and they end up doing it because of that, what AI user is to be held to account?

Every single communications advancement has pushed that more and more towards the people, and I think AI will still favor the people over the elites for propaganda

It’s the opposite, this is very much a power-consolidating technology. It’s not like, say, reading and writing, where nobody can just change the alphabet to become incapable of describing subversive notions or something. With AI, what it’s capable of and what its biases are, is very much dependent on decisions of the people training and managing it. And because the cost of training a model the size of an LLM is so high, there are only a few entities that can afford to do that.

On top of that, people with good intentions cannot use AI at scale, because they need to verify all of its output. People with bad intentions, who don’t care about the details, only that the waters are sufficiently muddied, can use it at scale. And if they are or are backed by the elite, then they have the funding to use it even more at scale.

I can literally put Trump’s face on a chicken with no skills.

Only because the elite are currently subsidising it for you. Because they want you to accept this technology. Because they know it empowers them more than it does you.

You don’t seem like the type that’d be advocating teaching kids how to use it

I’m pro-education, but to teach someone, there needs to be something that can be taught first.

How does text prediction have nothing to do with programing?

The large majority of text is not code. To begin with, LLMs being able to generate code was more of an incidental discovery than anything, only then did you get specialised products for that purpose.

No, it’s not easy, …

That’s my whole point: if it’s not easy, then you have to justify the cost. And while yes, as you say, behind a lot of the problems associated with AI, there are deeper underlying problems that have nothing to do with AI, but that doesn’t mean that AI doesn’t significantly add to the problems and the cost to solve them. Let’s say you’re in conflict with someone, in what situation do you expect that it’s easier to resolve by talking it out: when they’re unarmed or when they have a gun aimed at you? Same underlying problem, different difficulty in resolving the problem.

Only if you buy into the idea that it’s something we have to move the world for. You claim you can see how little an effect it has had, but you simultaneously seem to think it’s a big bad demon.

I don’t understand how this is not getting through. Something can be worthless to one group of people and very valuable to another. If you care about truth, human wellbeing, creating value for society and so on, then AI (as in LLMs and other similar generative AI) is less than worthless to you. If you only care about gaining wealth and power and don’t give a damn about what is true or what harm you may be causing to society, then it’s the most impactful technology since social media. These two things can be true at the same time. And if you’re not aware of all the problems the second group has been causing at a scale that’s only possible because of AI, then quite frankly, you’ve been living under a rock. People getting laid off under the pretense that they can be replaced by AI is hardly the only problem.

Screaming regulate AI

Who said anything about regulating AI? I want nothing short of a third AI winter.

No, it literally is a programming package + some data.

What are you even talking about? LLMs, which is mostly what “AI” has been referring to in this conversation, are text prediction systems. You prompt it with text and it tries to predict what text comes next based on its statistical model generated from its training data. Add some cute pre-prompting and wrapping of user input behind the scenes, and you can give it an imperative interface that gives the impression that it’s responding to user queries and instructions. The output is entirely too unreliable and the input entirely too imprecise to be valuable to anyone with good intentions, but I digress. The point is that it has nothing to do with programming, other than that if code was part of the training data, the model can be induced to generate text that includes code, and some products exist that have been optimised to generate and interact with code.

“Agentic” (ugh) just means that the model was induced to generate output in a format that the client can parse and turn into actual operations, like deleting a database. I guess that kind of overlaps with what a pipeline or a script can do, but I’m not entirely sure where you’re going with that, I’m pretty sure I already made it clear that it’s not my view that AI can actually replace workers, just that it can be used to keep up the pretense that it can for long enough to cause harm.

First of all let’s not act like my exasperation with the world’s stupidity is some sort of lapse in my judgement. I’m getting the ability find algorithms and debug faster to write better code, and occasionally asking it for terms I can search for to find information. That’s it. I’m not revealing some deep political bias you can use to ignore my underlying argument.

It’s not just that line, everything you’ve been saying suggests that’s what your view is. You’ve acknowledged yourself that the real use cases of AI are very limited, but despite that insist that that alone is worth dealing with all the problems that come with it, and we should just hold the bad people using it for bad things accountable, as if it were that easy (not to mention that sometimes there is no person to hold accountable, because it’s the model itself causing the problem). All for a small sliver of value to you, that in my mind doesn’t even exist. How is AI itself not part of the problem here?

Yes, a lot of the problems associated with AI are just worse versions of problems that already existed. Workers getting laid off by the thousands to raise shareholder value, misinformation, assholes on the internet trying to get people to kill themselves for the “lulz” etc. And these problems need to be solved regardless of AI existing. But the existence of AI makes these problems worse and harder to solve. So you’ll have to excuse me that I’m not very keen on the idea of letting our enemies have mechanised infantry so we can have a toy to play with.

Most people are thoughtful and well-intentioned. But one, the minority that isn’t are enough to cause a lot of problems, and two, not all problems associated with AI are caused by bad intentions.

You say that we should just blame the people for how they use AI, but aside from that there’s not always a person to blame or the people causing the problems are often effectively invisible, people are already being blamed, there’s just nothing to make them care. So how do you propose we hold them accountable? The law? I mean, there’s not a whole lot of political will for that right now, but even if there were, a lot would be difficult to encode into law and even more difficult to meaningfully enforce.

Also, having to hold people accountable is also a cost that needs to be weighed against the basically non-existent benefits.

AI is a god damn programming package

No, it’s not. Other than a few specific products, it’s not marketed like that, it doesn’t have an interface that suggests that it’s that and its functionality (or veneer thereof) isn’t limited to that. What a strange thing to say.

Why do I have to lose a programming package because everyone - for no real reason - started blaming it for the reason they’re assholes to each other as if they hadn’t been jerks to each other before. It’s stupid.

And here we get to the crux of the matter: you are getting some use out of AI that at least in your subjective experience, for now, is positive, and that’s all the justification for its existence you need. All the problems in the wider world associated with it, you just magic away with the phrase “personal responsibility” so you can just stop thinking about it. But that’s not good enough.

I can’t offer anything other than armchair hypothesising, but Tezuka Osamu, the most important person when it comes to making manga and anime what it is today, and whose influence can also be felt in video games (e.g. Mega Man clearly taking influence from Astro Boy), liked to do things like having a character rip out the gutter between panels and break it in half in a fit of anger. So these kind of meta-hijinks have been part of modern Japanese popular media since the beginning.

You’re basically making a “if only people would just…” argument, but people are not going to “just”. There’s no world out there in which “AI” (again, using a pretty strict definition) exists, but there is no search results polluted to the brim with slopsites, idiots using AI as though it was a reliable source of information, automated propaganda, moguls trying to gaslight the world into thinking workers and artists have no leverage, deepfake porn, chatbots encouraging suicidal thoughts etc. etc. So you need to weigh the costs and the benefits. And the cost-benefit analysis of AI looks considerably different from the hammer’s.

Well, I say all that, but it’s not as though I believe the world can’t be made a better place. If we can set up society so that these bad behaviours aren’t incentivised to begin with, it would mitigate some of the worst problems associated with AI (though I would argue that such a society wouldn’t waste its resources on such a worthless application of the technology in the first place). But when is that going to happen, tomorrow? Worse yet, AI empowers the people who don’t want society to change in that way far more than it does the people who do. So until we finish fixing society with our hands tied to our backs, we’ll just have to suffer through all of AI’s problems, and for what? So you can feel smug about misunderstanding a sentence in a scientific paper, because you were in no position to determine whether the rephrased sentence an AI ejaculated accurately conveyed the information in the original sentence, because if you were, you wouldn’t have needed the sentence to be rephrased in the first place?

If someone creates a destroy-the-world button, is only the person who pressed it responsible for the destruction of the world? That’s an absurd example, of course, but the point is that you can’t just classify something as a “tool” and then leave all outcomes of the existence of that tool as a matter of personal responsibility, that’s just cessation of thought. You have to actually think about who is empowered by a tool and in what way. And if a “tool” largely and greatly empowers fakers, grifters, political actors with little respect for truth etc. and the value for most people is largely nil to negative, then I’d say the “tool” is very much part of the problem.

It seems weird to suggest that the supposed legitimacy of the AI hype bubble is caused by anything other than people with a lot of economic and political leverage using that leverage to will something that doesn’t really exist into existence. Which isn’t that hard given the people they need to convince of its existence are some of the gullible people on Earth, and they do have a product that, while not what they claim it is, does have the veneer of it being that. If a company’s shareholders are convinced that AI can replace many of the company’s workers (or are convinced that the other shareholders are convinced that it can), then the executives can fire thousands of people and it will just increase the shareholder value. Sure, it will sabotage the long-term health of company, but by that time the locust will have had their payday and moved on to the next grift.

Now of course, this sort of thing has been happening since long before LLMs and the like were a thing, but a key difference is that before, there’s been at least some connection between the executives’ and shareholders’ fortunes and the company actually doing something, given workers at least some leverage to gain and enforce their rights. But now the existence of AI allows them to keep up the pretence that workers have no leverage at all for a good while, so so much for workers’ rights. It’s not actually AI doing this of course, but the existence of a strategic weapon is still a problem even if the weapon is never deployed and even if it’s just a box of firecrackers and pinball machine parts. And that’s putting aside all the ways AI is being used as a tactical weapon, undermining education and democracy.

Reminds me of pattens, except they’re for shoes that already have soles.