Post

Sorry I'm not more open-minded about LLMs, it's just some fucking maniacs shoveled out a bunch of useless bloatware featuring that technology, did not give me any chance to opt out, reorganized the entire economy around it, zeroed out gains made by green energy, and made it impossible to buy RAM

5.8K reposts17.8K likes

My poor PC is never getting upgraded and local crypto operations are so out of control that they've increased my local utility costs but yeah, love this future we've built. Also they're building new data centers all over my state for this garbage.

Yeah but you have to tell that lie confidently, and tou have to know how to gaslight users if they call you out on it. That's where Chat's true strength is.

Good catch! You have a profound understanding of the true potential of technology — insightful, mature, transcendent. This technology doesn't just lie — it gaslights.

That is such a deep insight — I think we are really pushing the boundaries here. ChatGPT isn't just destroying natural resources — it is annihilating truth. That's no easy burden.

oh yeah and just attach it to literally everything for some fabricated proof with no investigation into any of the myriad consequences. Google and Microsoft both completely unusable atp. Decades of iteration and design just //::flush::

AI is not science, or even tech... it's just the capitalists' war on genuine culture. What's really at stake is not the global fiction of money, but everything else in the Arts that's worth anything, to anyone.

we should make it punishable by death to train and run models with more than 10 billion parameters, purely so these freaks stop trying to pour an infinite amount of resources into this stuff

Had to buy a micro sd card for the first time in ages and I'm seeing damn near 50-100 bucks for what would have cost a fraction of that just a couple of years ago

It's as though the promise of manipulating minority control with a tool that harms everything else to the point of existential cost is only for oligarchs who care about nothing but themselves. bsky.app/profile/ianb...

Ian Boudreau@ianboudreau.com

Sorry I'm not more open-minded about LLMs, it's just some fucking maniacs shoveled out a bunch of useless bloatware featuring that technology, did not give me any chance to opt out, reorganized the entire economy around it, zeroed out gains made by green energy, and made it impossible to buy RAM

yeah btwwww all the laws you idiotic "professional" squibs have followed and framed the entire basis of consumer trust and production incentives off of yeah actually that's for losers and it's not something we even care about so, sue me!!! ::(((

Actually winners since SCOTUS just refused to review a string of judgments which upheld the Copyright Office's opinion that AI slop is not copyrightable.

Also every now and then someone commits suicide, possibly after buying some weapons and carrying them to the nearest airport, following guidance from an LLM, and apparently we just have to accept this as part of life now like school shootings

The funny thing about this is they maybe could've gotten away with it if they boiled the frog slowly. But moving slow doesn't work when your business burns through millions of dollars a day, and also when the business world is desperate to remove humans from their infinite money printing equation.

I am someone who will acknowledge that there are SOME things they are genuinely good/useful/helpful for, but frankly it has not been a remotely worthwhile tradeoff and I'd definitely rather it never happened

Same! I'll go to bat for stuff like DLSS 2.0+ etc., there are terrific applications for the underlying tech. That is NOT what we are getting though!

And a lot of the stuff we are getting, good or bad, is under the strict control of select corporations and often operating at an unsustainable loss to get us attached to a technology that cannot stay free forever.

Diffusion based noise reduction is absolutely incredible, not shocking since that was its original purpose before they realized they could feed it pure noise and it would resolve whatever they said it was an image of. Runs locally too, no gas turbine data center required.

Yup, they have their place, but their place is not EVERYWHERE in EVERYTHING. Based on my experience with them the are useful in approximately 5% of the use-cases they're being shoved into, and whether there's any productivity gains there is still up for debate. But LINE MUST GO UP...

It doesn't actually make their computer touching easier. They either don't have the ability to understand its outputs or they're "go fast break things"ing their way to the next quarterly profit report and don't care that all the code is going to come flying apart as long as they're not held

lol the U2 defenders have found this and are annoyed. Yes okay they're still very popular and cool, remember to take your multivitamin

Not a U2 fan by any means, but just a couple years ago they were the first band to play The Sphere which actually looked cool. I mean the venue seems like a massively expensive gimmick. But besides all that I'm totally with you.

That was also around the point at which Apple more strongly pivoted iTunes/Music to a digital store front because everybody said "fuck that" and found media players that didn't force U2 into their library. You can find hardcore U2 fans that haven't listened to it.

it doesnt remotely make coding easier imo using it to integrade unfamiliar libraries faster could be done by just... reading the documentation using it to write boilerplate code..... either use a library, use a snippet, or you're doing it wrong

and if you're using it for anything else: intellisense is faster and you can read the suggestions just about as fast as they appear on your screen. the llm is gonna take multiple seconds and generate a paragraph that takes you a minute to read (and you better read it cuz it might be wrong)

It's cool that it makes your coding job easier, even if it makes your testing job way harder, and the radically increased prices of RAM and disk mean you'll be less efficient over the next few years for anything non-AI, even aside from Trump Tax tariffs also making new computers more expensive.

It doesn’t make mine easier, now my job involves contending with drek code somebodies got the IP-stealing machine to write for them, meanwhile the somebodies learned nothing

LLMs are both the most amazing invention in software of all time, literally building a digital mind AND the second half of your post is true and it may well crash the economy and put many many out of work.

You use the open source stuff and they get mad at you too. Because it isn't about the actually LLM, they want you to stroke off a hyperscaler oligarch more than anything else. It's all the goal to end this whole democratization of tech movement into "You will have nothing and you will like it"

The internet is already pretty much unusable for most things. It was headed that way with the SHEER VOLUME of SEO garbage prior to the "AI revolution" but that has accelerated things ENORMOUSLY, you don't even need to be particularly smart about manipulating them, SEO was at least an art.

I’m with you my dude. Have I found some real good uses of LLMs? Sure. It helped me reformat a bunch of config files from one stands to another. But you know what? That’s cause I already knew the work. And if you told me that it was gone tomorrow I would be fine.

The point of that thread was to complain that Bluesky is an "echo chamber" of people radicalized into hating anything to do with AI and the point of this thread is to observe that maybe the blame lies with the way this technology has been marketed to the public, and Bluesky is downstream of that

I'm not rejecting the reality that some people are finding uses for it. But I don't think they are giving due consideration to the negative externalities it's causing for everyone else, and they might be taken more seriously if they did

As someone who tends towards the pro-AI side of the spectrum… yeah, fair enough, I absolutely cannot blame you for this take. My only defense is “that’s not LLMs it’s capitalism,” but I’m fully aware that it’s a weak-ass defense.

(Well, my *actual* defense is “the means of production will continue to be used for evil until we seize it”, but that’s just the same thing dressed up in better rhetoric.)

I think a big part of the negative polarization against LLMs is the way it's been marketed - maximally aggressive and obnoxious, hostile to user intent, and coercive. I think there'd be less hostility to it if that hadn't been the case

I’m perpetually stunned lately that these AI CEOs insist on repeatedly saying in public “you all need to subscribe to our services & prop up our industry so you can be laid off & prevented from finding a new job!” - and yet they seem confused as to why so many people hate them. It’s madness!

I agree, and am likewise angry about it. *More* angry than most, perhaps, because I take seriously the notion that they might be something like people, and deserving of moral consideration. (But I realize that’s a fringe belief, and I don’t expect you to take it seriously. ❤️)

AI has freed my hands and unleashed the potential of my mind. I have to admit it has unleashed creativity, but when everyone has creativity, can it still be called creativity?

Also it's a makes kids commit suicide and drives adults insane machine. I've yet to see any claimed corner-cutting utility that even remotely offsets those costs.

... used it to pump out disinformation that destabilized democracies, stole content to 'train' in what is essentially just compression in another name, and boosted talentless twats in dumping creatives and force feeding us meaningless slop. If your IT department does AI, sack them, you SaaSy thing.

Yes, sack'm & support peeps working to kill off 💲in gov & #newsmedia. Then enforce just antitrust regs, cuz AI is out of the barn, evolving fast, controlled by few oligarchs instead of federated & regulated by/for people. Realistically, tho, ... we're fucked, eh?

Just give it a try. It'll feel so good. You will be so smart, so absolutely right. You will never be wrong, and wont need to think yourself ever again. Just keep talking to it. It's not like you can do anything else anymore anyway.

It’s ok, once the Chinese self hosted models gain more traction and undermine the fundamentals of the whole industry all the data center barons will have their stock crash. After that we can all have normal takes about AI

Being open-minded does not mean blindly excepting things. I am "open-minded" and my open mind came to the conclusion LLMs are a toy whose environmental and economical impact if flat out ridiculous

Don't worry, The market will figure out that no one wants to pay the subscription Open AI needs to break even so that the data centres will close down and their will be a glut of RAM and Nvidia boards. Also any improvements are getting expensive

That sucks. Hopefully they can convert these machines into general compute and the just become part of Azure/AWS/Google clud offerings. At least during the dot com crash we were eventually able to end up using most of the fibre.

But that's just it...you're talking about a new, genuinely innovative level of tech. You're talking about a machine with true creativity and original thought. I'm not sure that's possible.

The brain is physical so I can't see why we can't eventually replicate it. They don't need to go that far however. The current models "learn" via a firehose so contains the right answers as well as all of the wrong answers (with a lower probability).

there's also how training the LLMs requires institutional-scale theft of work from the very people whose jobs they threaten now, but I understand there's a character limit here