Blinkie: Evil Mode

there's somethign that riley quinn from the trashfuture podcast keeps saying -- "if your job can be replaced by AI, it was already being done by AI". which is to say, that jobs most at risk from AI replacement are ones that were borderline automated anyway. like, i say this as someone who used to write, not for the website buzzfeed itself, but buzzfeed-adjacent Slop Content for money -- i was already just the middlewoman between the SEO optimization algorithm and the google search algorithm. those jobs vanishing primarily means that middlewoman role has been cut, computers can tell other computers to write for computers.

& similarly this is why i keep saying that, e.g. stock photographers are at risk from this, because the ideal use case for generative AI content is stuff where the actual content or quality of the image/text doesn't matter, all that matters is its presence. and yknow, living in a world where many people's livelihoods were dependent on writing and art that is fully replacable by inane computer drivel is itself indicative of something about culture under capitalism, right?

& to some degree, like i'm always saying, the immiseration of workers by advancement in technology is a universal feature of capitalism -- i recommend you read wage labour & capital to see how this phenomenon has persisted for well over a century. it's simply nothing new -- like, the stock photographers who are most at risk from this already are already employed in an industry that itself decimated in-house illustration; think about how any dime-a-dozen reomance novel you can pick up at a store nowadays has a hastily photoshopped stock photo cover when fifty years ago it would have had a bespoke cover illustration that an artist got paid for.

Buddy, it's an environmentally destructive brute force algorithm that feed upon plagiarized data and can't produce a coherent documentation pathway of its processes or parameters. That by itself, makes it completely useless in collaborative projects or ones rhat need a lot of fine tuning and different development iterations.

This is why as an engineer I'm ALWAYS against the use of AI crap because the reality that most of y'all would never need to use it, and yet you cram that shit everywhere unnecessarily. Are you optimizing robotics controllers? Are you doing path finding algorithms? No? Then no, it's useless. You are chasing after the latest shiniest thing just like those capitalists with more money than sense.

Machine learning is best used to add finishing touches to an already developed main mechanism. But you idiots want it to be the substance itself. The best way to describe it is that nearly ALL of you are trying to making a shoe out of shoe polish.

it's less environmentally destructive for someone to talk to chatgpt 24/7 for a year than to eat 1 burger. yes to local communities too, not just in ghg. training a *frontier* model in the first place is about as ecologically costly as regular light car usage for a year. these costs are both going down per unit of work btw, not up

who cares about copyright. are you serious? are you for real? "plagiarized" oh no someone right click downloaded my art! you are not selling nfts. if you are, lol

coherent documentation pathways are likely possible but why, in a competitive market determined largely by model performance, would any company training their models reveal their training data? you are criticizing company secrets, at which point, you should be criticizing IP and the market. what you are criticizing has basically no relation to AI at all

I don't know about you but I like it when I can communicate intent to the computer and have it spit out diffs that make my intent happen irl. and when I'm employed, 90% of my time pre-AI was communicating and establishing intent shared across multiple teams. now that I have AI, 99% of my time is spent communicating and establishing intent shared across multiple teams. wow!

I build programs out of text and you'll never guess what the second l in llms stands for

Analytical AI, is the one that can detect cancer and save lives

Generative AI is the one that steals art to make it worse, and gives you a wrong answer every time you google something

generative AI is also used in drug discovery technologies that can also save lives

complex technologies cannot simply be decried as good or evil

For example, it’s very reasonable to be concerned about the new Twitter data center that’s poisoning Memphis right now, and single out the LLM, grok, as a source of it. However, generating an image using locally run stable diffusion does literally nothing to contribute to this. Conversely, using Twitter without using grok absolutely contributes to the polluting Memphis. If we try to simply boycott AI entirely and nothing else, we still contribute to that data center. However, if we single out Twitter/X as a company, then we can actually have an effective boycott.

If your problem is the environmental cost of all data centers, and you single out AI as the source, but still continue to stream video (ie Netflix, YouTube, TikTok etc), use social media, etc, you are still contributing to the stuff that’s at least 90% of current datacenter output. I know this is news to people but data centers didn’t just. Suddenly become a problem post-AI.

People pointing this out are not being contrarian for no reason. Current anti-AI thinking has done literally zero for actually combatting the issues we need to care about, because a lot of these problems go beyond AI. This is the tech equivalent of banning straws.

The point of this isn’t to go “oh and yet you participate in society…interesting”. The point of this is to show how you need to be thinking systemically about these issues beyond the superficial shit you hear in op-eds and tweets, which by the way, frequently misreport the actual data/info regarding the situation.

i am differentiating between criticisms of genAI that relate to the concerns of workers: how it is integrated into workplaces, who might become unemployed because of it, who might have their work deskilled and get paid less because of it, who is mistreated and exploited in its creation -- and criticisms of genAI that relate to the concerns of the bourgeoise: ie its inadequate respect for property rights or perceived property rights.

Imagine you build this in-depth fictional setting over the course of decades with your friends, and then decide to turn it into a game for fun. You end up creating one of the most celebrated and well-written RPGs of all time and become celebrated for reviving a dying genre.

A few months later, you're fired from the company you founded for criticizing the new direction its taking. The company has the copyright not only to the game you made, but literally everything in the world of the game, which spans decades of lore you built for no profit with your group of friends.

the people using AI are not doing this. using AI is carbon-equivalent with using google search or texting your friends. Watching netflix is ~10-20x more energy intensive than using AI. training AI is relatively (vs using it) energy-intensive, but it's also nowhere near many other, much less useful industries (crypto mining, animal agriculture, war, automobile/roads, golf, american football, football, etc). Meta's Llama 3 series of AI was trained using 11,390 tons CO2 (equivalent). For comparison, *just car use* in California generates 323,276 tons of CO2 *per day.* So, at the equivalent of *one day, of one state's use of gas, solely for single occupant vehicles,* you can train 30 Llama 3's. Llama 3 was also one of the more expensive models to train out there; Deepseek, for example, used less than half that much energy.

Eating one (1) pound of steak is equivalent to more than a month's worth of average ChatGPT use. Driving your car to work for 15 minutes is just as bad.

"Analytical AI" is the same technology as ChatGPT. It's the same underlying math, it's the same computational intensity, it's the *same thing.* Drawing a distinction between "analytical" and "generative" AI is literally nonsensical.

To be crystal clear, I do think artists should be paid more. My problem with this chain is that you are angry at the wrong targets. Why are you upset at the consumer? I don't eat meat, but I don't get mad at people who do, because they aren't the ones making superprofits by torturing, raping, and murdering sapient beings. Omnivores are simply doing what's normal and convenient in their society. When someone uses C.AI or ChatGPT, they are doing what is normal. There is push back, but this technology is here to stay. Why? I can run (an equivalent to) ChatGPT on my laptop without an internet connection. How do you stop that?

What do you want, politically?

Stronger copyright? This will only hurt you, the artist. Do you, @thankyourluckystars13, want to be able to continue making Legend of Zelda artwork? It's beautiful, I'm really happy I live in a world where I get to see your art. A future where AI is litigated due to copyright will prevent you from doing this.

Destruction of AI? I can run ChatGPT on my laptop, and weaker models on *my phone.* How do you stop this?

For it to be weird and socially ostracized to use AI? So how do disabled people who legitimately need it factor in? Do you want someone with a TBI to be unable to communicate effectively with the outside world due to language processing issues? Do you want someone with a physical dexterity disability to be unable to draw?

He's literally the core artist, he used some stupid tool to generate a tertiary asset (a snarkier person would note how funny it is that you're worried about the "background" in a touhou game of all games), nobody was directly impacted monetarily (outside of ZUN apparently possibly losing profits because people are upset about this), and like...

(The post that documents this info)