Just going off open source models, most of them are at least somewhat better than that out of the box, and certain strains of model have a bunch of tools built around forcing it into a more reasonable state (like modern descendants of SD1.5 - a model from two years ago - fail hard on their own 9 times out of 10, but there’s a mature ecosystem of tools to control post-NAI-leak SD1.5 and cover for its flaws). The most recent big model, Flux, is terrifyingly accurate on its own although it still tends to get proportions subtly wrong in very offputting ways.

Closed source models are impenetrable and no one knows what they’re doing under the hood since the public only interacts with them with prompt boxes and every company is being super secretive about whatever they’re doing.

That said, I do agree that it feels like there’s something fundamentally wrong with the way “AI” is currently being focused on. It’s like what’s being trained now are eyes (and backwards eyes that reverse the process of seeing to produce an image from how that image might be parsed) and speech centers, and there’s all this hype that the speech center bits can just be made big enough to start being smart instead of just being a potentially useful bit of language processing, and I really can’t help but feel like it’s just a flawed overfixation on one novel bit of tech kind of like how rockets became the standard for space launches because primitive rocket tech looked neat and had already been developed.

Going off the pictures, it looks like it was just suggesting “just increase the time limits to solve the time limits being hit” as a solution, it wasn’t trying some sort of subterfuge. Looking at the exact language used, it sounds like it got the concept from people suggesting increasing the timeout threshold as a solution to connections timing out or the like, which is very funny because it’s learning “instead of fixing the problem, is ignoring it and hoping that’s good enough an option?” from engineers.

Asking the chatbot if person of interest is gonna do a crime, and continuing to nag it until it says yes and gives me the official AI-prediction warrant to detain them indefinitely.

*tinny midi jingle begins playing* “In hard times like this…” *slideshow of pictures harvested off social media of likely family members of employee begins playing* “It’s important to remember all the people who are relying on your employment, and how failure to meet your quota could affect them.” *midi jingle continues playing on repeat for several minutes as part of a pavlovian training of the employee* “Your mood control break has now concluded and in compensation to your employer for this unscheduled break ten more call resolutions have been added to your target quota.”

You’re not thinking insidiously enough. It’ll be an excuse to raise prices across the board and then provide incentives for constant consumers that return prices to a lower, but still higher than now, point. Instead of a cold and calculating “someone making 3x as much will eagerly spend 50% more on this without thinking about it” plan it’ll be “everyone pays 50% more, but good corporate paypigs get a 25% discount off of that, and if their loyalty credit score with the vendor is high enough it’ll be 35%!” incentivizing both consumption and subservience while extracting more profit off of fewer commodities.

It’ll be the evolution of existing “loyalty programs” and the like, not the sorts of weird idealist “market efficiency” thought experiments the high priests of capital think up when swimming in numbers and models built on flawed and nonsensical foundations. Because as everyone quickly sees, this whole concept is nonsense that implodes from outrage or gets gamed immediately - it only works in on a small scale with passive participants who don’t share information or know it’s happening. But the sorts of infrastructure the concept requires can easily be used the same way as existing manipulation tactics, just amplified to their logical conclusions.

GHub is awful, though fortunately one can just use the old software that the G502 originally used (Logitech Gaming Software) instead.

I will note that the G502 has had varying build quality over the years: the original ones were amazing workhorses and mine has been in constant use for ~8 years without any issues except for some of the LEDs burning out, but some subsequent runs have defective switches in the right mouse button so they quickly wear out and break - the part itself is apparently both cheap and easy to replace with a better switch, but that requires doing some maintenance on it if it happens to you.

You can replace GHub with the older Logitech Gaming Software which was also bad but actually works, has a better (though still bad) UI, and interfaces with the G502s onboard memory to set the profiles, instead of being some weird software based profile system. Just search for Logitech Gaming Software, download it from logitech’s support page, and then uninstall GHub so it lets you actually use it.

I should try ComfyUI; to be blunt, I went with Automatic1111 because there was a decent walkthrough on how to set it up on a Radeon card

IIRC setting up comfyui for AMD is the same as with A1111, it just only works on linux because it doesn’t have directml support. I just cloned the conda environment I used for A1111, IIRC I uninstalled the pytorch libraries from it, then let comfyui install the versions it wanted. I don’t think there was anything more involved than that, just setting up a launch script with the right environment settings to try to mitigate the frequent hard system crashes I get with ROCm which may or may not have been solved by switching from ROCm 5.7 to 6.0, changing a certain kernel value, adding the right environment settings, and underclocking my 6800 even more than it already was - I haven’t crashed since then but I haven’t stress tested it either.

I’ve played around a fair bit with Stable Diffusion running locally on my PC and can confidently say it has an absurd potential as a streamlining tool for art because unlike the simple prompt remote tools you have tons of tools available to control composition, pose, details, etc instead of just giving a vague prompt and hitting “generate” over and over like it’s a slot machine and you’re addicted to the rush of seeing if the latest pull will be a winner or more garbage.

That said, if anyone’s actually using it well they’re keeping it to themselves: the hobbyist AI community itself is compromised of the most vapid, talentless dipshits you’ll ever see who want pats on the head for asking the inscrutable machine “pls give good image” while using something like A1111 or Forge because ComfyUI’s extremely simple flowchart UI is too hard for them and they want a single text box they can ask for their skinner box treats instead. And don’t even get me started on civitai, a site that I think I’ve actually gotten addicted to raging at, like I’ll check it multiple times a day just to about its userbase.

I’ve compared it to things like Poser or Daz3d before, which caused floods of absolute dogshit CGI content made with prefab assets that still haven’t completely subsided to this day, which is made all the funnier because of how many people seem to want a terrible faux CGI look out of image generators. That, the weird sort of soft-focus glossy style that makes me think of like sketchy airbrush art on cars for some reason, and the creepy grasping for photorealism are all weirdly popular with them and it’s all uniformly awful drek.

Still, I paradoxically think it has a ton of potential for more minimalist aesthetics where it can serve as a labor amplifier for an artist that can feed it a sketch with maybe some crude shading, and then fix up the output in short order, it can create out-of-focus filler for backgrounds, etc. I strongly believe that open source models should be embraced and exploited for that purpose by independent artists and the left in general, because the corporate use of proprietary models is not going to stop and we should adapt to the way that’s transforming the landscape and seize upon new capital to compete and survive. The only alternative would be en masse, and there’s just not a militant anti-AI movement that could do that nor is one going to form because of how abstract and unemotional the problem is.

It it weird that I find a giant company making money off of strangers words without consent a little rapacious?

This is exactly what I’ve been harping on about: all the buzz about AI training “stealing” from rightful property holders has always just meant that the enclosure of the commons with AI will be legitimized and secured with token licensing fees paid out to huge corporations, the content hosts, social media sites, and media companies, without the actual creators of the content seeing a cent or having a say in it. After all, they don’t own what they’ve produced, some huge corporation claims to own it instead.

touch screen phones prove you do not need to have fixed analog buttons

Touchscreen phones and tablets are all horrible and notorious for not working right whether that’s in the form of presses not registering or registering in the wrong spot. Like think of all the design that goes into a decent keyboard, the way keys are differentiable without pressing them, how there are physical marks that tell you where your hand is touching it, and how all this combines into an input device that you don’t have to look at to use quickly and accurately (hell, my keyboard no longer has letter markers on half the keys because they’ve worn off over the nearly 20 years I’ve been using it, and this doesn’t matter because keystrokes are even more ingrained into my hands than literal written text is, but this relies on the tactile feedback of them).

Meanwhile a touchscreen is a flat, featureless surface where nothing has a fixed position, any input may or may not work, and you have to watch it to see where it wants to put a button and whether that button is reacting correctly or prompting another input. Operating traditional controls that require a hand to be removed from the wheel, like for a car radio or the AC, is already considered a dangerous hazard that’s only tolerated because it has to be; making that at least an order of magnitude more distracting is a catastrophically bad idea.

Touchscreens are ludicrously dangerous interfaces for cars that have basically only proliferated because of techbro dipshits mucking about with things they don’t understand. Cars need fixed controls that provide physical feedback and which don’t require reading or light to operate. Like that’s not a matter of taste, that’s a basic “this is usable and safe” thing.

“Researchers who do not speak native English are using ChatGPT a lot, as an aid to writing and to improve the English language,” says Delgado Vázquez, a researcher from the Pablo de Olavide University, in Seville, Spain.

So in a lot of these cases it’s people who don’t speak English or don’t speak it very well having ChatGPT work over a machine translation of their actual work, and seemingly in other cases it’s a knock-on effect of researchers who are still learning English interacting with it a lot and incorporating its idiosyncrasies into their own English.

The researcher jokes that his colleagues have congratulated him on the meticulousness of his report,

I love that even his coworkers made that exact joke.

Is this about the guy I posted about a couple of weeks ago in the dunk tank?

Holy shit it is, lmao. https://old.reddit.com/user/BryanVision

The clearest thing I can remember is that Socialism Betrayed: Behind the Collapse of the Soviet Union despite mostly being a blow-by-blow of the reforms and political situation of the 80s talks a bit about the further-industrialization vs consumer-goods camps in the 50s. I haven’t come across anything that goes into the same sort of detail about what the actual material policies and conditions at any given point in the USSR were as, say, Sorghum and Steel does for China up to the 70s (caveat for Sorghum and Steel: it was written by ultras who are vocally anti-CPC, but it’s still the most nuanced and sympathetic piece I’ve ever seen about China because of how much it goes into detail about why this or that decision was made and what the material conditions behind it were).

The Chinese economy really wasn’t planned in any but the broadest and fuzziest terms, if only because the central government simply lacked the material ability to plan out production and logistics directly on the scale of the entire Chinese economy. That left individual regions highly autarkic and free to experiment: the model of the Great Leap Forward came out of the experiments of individual rural communes, for example, and the central government only popularized it after the supposed success it was having (to mixed results that varied heavily by region, with it doing ok in some places and proving catastrophic in others). For urban factories, the organization ranged from traditional strict management-led patterns that any capitalist would expect of a business, to radical worker-led systems where individual workers had considerable leeway to set their own pace and procure their own tools and materials on behalf of the factory - both models functioned about as well as the other, and both ran into the single biggest problem China faced which was the crushing lack of industrial capital.

That last point is really the core of the whole issue: China as of the revolution was one of the most underdeveloped places on earth, with less total industrial output in 1949 than Russia had in 1917, and that problem was a heavy influence on every decision the CPC made since. They received an influx of industrial capital from the Soviets, but then Khrushchev happened and the USSR shifted its focus from producing industrial capital to further industrialize towards producing consumer treats since the left-liberal bloc with Khrushchev thought that was better (it wasn’t and it led to stagnation and increasing dissatisfaction from the liberal bloc who wanted treats and couldn’t understand why they weren’t getting as many treats as white americans were and this led to Gorbachev and Yeltsin), and with the Sino-Soviet split that potential solution ended entirely. Ultimately their revisionist cooperation with the US solved that problem: they got the industrial capital they needed to actually develop their economy and they got a flow of resources in exchange for the industrial output of their large and well educated labor pool - by solving the need of American Capitalists for ever larger labor pools and a market to sell lots of brand new industrial capital in, they solved their own problem.

And now they have their own modern struggle in the CPC between the communists who want ever more aggressive reforms and a move towards socialism with their now-developed economy and the liberal opportunists who will maybe concede to some socdem style regulations and welfare programs but mostly want to keep the status quo and keep lining their own pockets. From afar it seems like the left is winning that, and things like the massive infrastructure projects and anti-poverty campaigns of the past decade shows that at the very least they’re to the left of any modern socdem party which would clutch pearls over the cost and then surrender to liberal demands for austerity instead.

Last I saw you could disable it through group settings.

In this case “AI” means “advanced relative of OCR,” not a chatbot. If it’s the same tech I’ve seen talked about before, it’s basically using imaging techniques that penetrate or pass through the material and then trying to analyze the resulting data to figure out what’s ink and what’s paper and what layer it’s on and turn that into something researchers can look at and decipher. The script and language are both known (it’s just some classical greek dialect), the problem is that the scrolls are so heavily damaged that they can’t be looked at normally.

Yes. It’s also been done, like 15 years ago, with invasive brain probes and IIRC with less invasive outside-of-the-skull electronics as well. Just like all the “experiments” neuralink was doing on monkeys was just stuff that had already been done 15 years ago.

So far neuralink’s entire thing has been trying and mostly failing to do stuff that actual researchers did a long time ago, with tech that’s far behind its competitors. The only question is whether they’re grifting Elon, or if he’s in on the grift and they’ve got other marks.