Consolidated AI Thread: A Discussion For Everything AI

Game Development Writing and Content
537

The claims aren’t outrageous; they’re fact. The need to create more and more data centers due to the usage of AI, and the carbon emission footprint they have on the planet, is something that companies like Google have reported themselves. It’s also worth noting that, yes indeed, farming creates about 10% of the earth’s greenhouse gas emissions. But that is not the exact same as carbon emissions. Carbon emissions are a specific type of of greenhouse gas emission.

You wanna know how much of that 10.6% is made from carbon emissions from running the farms and producing them with electricity? 0.6%. ( Reference is found here https://www.ers.usda.gov/topics/natural-resources-environment/climate-change/ ) The majority of greenhouse gas emissions from farming comes from Nitrous Oxide, which is used in soil and manure management. Frankly, it’s a bit wild to me that someone would compare the need for AI data centers to the need for agriculture.

Is the agriculture industry perfect? No, of course not. But it’s an industry that feeds billions of people. It is, by all accounts, a necessity. Without it, you (the general you, not you specifically) wouldn’t be able to buy eggs or milk at the grocery store.

Again, data centers produce about 3% of carbon emissions just from the amount of electricity it takes to run them. That is only carbon emissions. And again, that’s not me or other people coming up with random numbers. It’s based on real numbers that the companies using them are reporting.

13 Likes
538

“No one knows the future of it” is kind of why it’s primed for failure in its current state, though. If you’re going to ask for billions of dollars each year, you have to demonstrate some practical applications which can generate actual profits down the road. The issue is that none of the current AI developers have been able to do that.

So long as your stock bubble is pumped up with nothing but hype, it’s bound to eventually burst with the investors jumping onto the next shiny thing that can be pumped up to give them the quick buck.

The internet analogy is poor in this regard. 3D TV or the VR would be perhaps closer, with how much was sunk in them based on grandiose ideas and Vision™, that inevitably and repeatedly got a reality check.

7 Likes
539

I mean in a best case scenario, if I am being realistic, I see the future of AI as a marginal but useful technology when it comes to tasks like quickly summing up large documents, a better grammar checker, scanning large amounts of data, producing what amounts to clipart or stuff for powerpoints, and better functions for digital art, like generative fill, background removal, and stuff like that. Things where the AI won’t be noticeable, and which will help people who perform those tasks to work quicker and better. I don’t think any of this will be labeled as AI in the future because the name will be so tarnished, just like terms like Web3 or blockchain are useless as selling points, even if parts of the tech might still be in use.

Of course it will also be useful to more quickly produce scams, fake books, cheap art and a million other things that scammers were already doing (like making books from Wikipedia articles and the like). Just a lot faster, more automated, and in far greater quantities. All this will lead to people disliking AI even more, because just like spam e-mails or robocallers, they will experience far more of the bad tech and the annoyances that come with it. And now we have spam filters, and nobody answers their phone unless they know the number…

Also, none of these things will generate any huge income, which means that the data centers will be economically unfeasible, and the more showcase chatbot AI’s will cease being free. Novelty only goes so far, and people are already tiring of it, even if they don’t care about the ethical concerns.

17 Likes
540

My main issue with AI, regardless of any other decent point that’s been mentioned for or against, is that it plagiarizes indiscriminately. And no one thus far has been able to produce a generative AI that doesn’t rip off hundreds to thousands of creators. Once you can earnestly say the generated content is reliably owned by you and your machine companion, then I’ll take less issue with it. It’ll still be undesirable at best to me personally, but that’s neither here nor there. Many creatives may be left with a shrinking pool of options for fair compensation and an increasing demand for insane production rates to compete with I Robot, but realistically, we were always facing that dilemma in some form anyhow. And I do not believe all is or ever will be truly lost. Maybe I’m stubbornly optimistic.

4 Likes
541

I don’t personally have a problem with training sets that consist of public domain works, but that’s only my personal opinion of course.

4 Likes
542

Yep, precisely! That is where my complete hatred for AI comes from as well-- that and the fact that people try to pass off the plagiarized generated dumpster fires off as their own (and get seriously butt-hurt when you tell them "hey, you didn’t make/draw/write that!!) with a complete disregard to the fact that AI does not cite, reference, credit, or otherwise obtain permission to use those works.

Public domain is another subject in its entirety since it’s not legally protected by copyright. CC0 is also another subject, though similar, since both do allow their works to be used, altered, derived from, et cetera and in the instance of CC0, the creator does not always even require attribution.

However, AI is a thieving tool that rips off anything and everything it can find online, with no regard for where it came from, what the terms of that creator are (licensing requirements, credit requirements, whether that creator allows their works to be edited, referenced, used by another person, et cetera) or anything else.

Not just “may,” they are. I’ve read about numerous strikes in the news: actors, musicians and voice actors are being taken advantage of constantly by companies that think it’s better to bottom-feed via AI than to pay the people for their work, or even to credit them (in the instance of a big-name video game development company recently, they aren’t wanting to credit the actors and actresses whose voices and likenesses have been fed to AI.)

I hope you’re right and that it will end up dying down very quickly. I’m not so sure, given human nature and how corporations always want to screw people over in the name of making a quick buck, so yeah, I’m pretty pessimistic about the whole thing. But if luck is on our side at all, it might hopefully crash and burn as people turn to the next big shiny thing. As always though, I am very much a firm believer in the phrase “things will get worse, before they get worse.”

4 Likes
543

Kind of funny if you’ll need an AI to train AI.

On a general level I disagree - it is by no means necessary for an AI to use everything online only because it’s AI (you could totally train your own if you had enough material without resorting to scraping - I was on a university course that included training an AI to categorize images) but speaking of the current services that people usually mean when they talk about AI, I agree.

1 Like
544

OK, one thing of note, this concept that AI “plagiarizes” things is simply nonsense. AI puts together sentences from collection of individual words annotated with how often those words come adjacent to other words, and based on this frequency. If the AI writes

“That’s ridiculous,” she snorted.

It doesn’t do it because it’s lifting that sentence from some specific works, but (simplifying) because it has recorded that in hundreds/thousands works it has processed the verb “snorted” was very commonly used after dialogue.

For the same reason it can’t “cite, reference credit etc” any specific sources for its output, because there’s no “sources” for individual words, and that’s all it operates with. It’s akin to demanding a human writer to reference and credit every single book they’ve read in their entire life, since that’s what shaped their understanding of how to put together sentences. And complaining that they don’t.

1 Like
545

Mmh. But in some contexts, you really need to cite your sources as a human, and when people use ChatGPT to generate their answers/essays/research papers/news articles, it poses a huge problem.

3 Likes
546

Oh, absolutely. It’s not a coincidence use of AI is far more restricted (to the point of being illegal in some places) when it comes to academic works.

Yup, which is quite understandable, given that it doesn’t internally hold any “facts” it could give, much less references for those. I can’t blame people who take what it tells them for granted, because they’re basically duped by companies which created these things that they’re interacting with a “thinking” entity with vast knowledge.

2 Likes
547

There’s also the little detail that you really shouldn’t use it for fact-checking (which some people do), if it can generate content out of thin air and not give references so that you could check what is true and what is not.

5 Likes
548

It does blatantly plagairize and is not nonsense whatsoever:

The Authors Whose Pirated Books Are Powering Generative AI

Stephen King, Zadie Smith, and Michael Pollan are among thousands of writers whose copyrighted works are being used to train large language models.

– “The Atlantic”

Recently, a new Substack called The Rationalist lifted analysis and writing directly from Big Technology. Its plagiarized post went viral, hitting the front page of Hacker News and sparking a conversation with more than 80 comments. It would’ve been a terrific debut for any publication, if it was authentic.

– Observer.com

Almost every author has heard the adage that to be a great writer you must be an avid reader. It turns out that the same maxim applies to language-generative artificial intelligence (AI). In order to mimic human writers, AI technologies such as ChatGPT (OpenAI’s chatbot) and LLaMA (Large Language Model Meta AI, a tool for AI developers created by Facebook’s parent company, Meta) were fed millions of copyrighted books, articles, essays, and poems, according to the Authors Guild. Revered authors such as Margaret Atwood, Stephen King, George Saunders, and Zadie Smith are among those whose books are being used to train generative AI, the Atlantic reported in August. These and other texts are responsible for AI’s uncanny ability to reproduce high-quality prose and verse on command, setting off alarm bells in writing communities worried about the exploitation of their work.

– www.pw.org (Poets & Writers)

Yep, that is true, too. Which causes its own set of issues, like unnecessary slander, defamation of character and simple misinformation–all three of which can be very dangerous.

549

To be fair, the AI can’t plagiarize because it has no intention or will, but the companies who stole the data to train the AI certainly can, which is usually what people mean when they call the AI a plagiarization machine.

11 Likes
550

…“The Plagiarization Machine” sounds like it could be a great sci-fi book.

4 Likes
551

So, how does the claim from this article and the others you cite, about AI producing high quality prose, squares with your repeatedly stated belief that AI produces nothing but nonsensical dumpster fires…? (that a human can spot at a glance)

Shouldn’t this set your internal bullshit detector alight, at least?

552

Is it copying their plots, settings, and/or characters? I mean, I definitely am against using copyrighted works without permission, but I’m a little confused as to what’s counted as plagiarism here.

(I learnt the general tone and structure for writing thrillers by reading Jack Higgins and Robert Ludlum, does that mean I’m plagiarizing them when I write a thriller?)

4 Likes
553

An AI can produce high-qualityish prose (if we count it as grammatically correct and florid), while still creating nonsensical dumpster fire stories. Writing short sections such as paragraphs are usually fine, but the ability to remember and make connections are just not there, which makes it easier to spot where the story fails. Just like a picture might be lovingly rendered with folds and shadows, but when you look closer there are issues with symmetry, fingers or text.

6 Likes
554

Or with those folds and shadows, for that matter.

1 Like
555

Absolutely not.

I am a human, my opinion on quality is different than anyone else’s and that will not change. The world would be a horribly boring place if everyone agreed about the same thing and no one’s opinion was different than anyone else’s. What I despise, someone else will view as high quality and vice versa.

In the heart of what was once a vibrant metropolis now lies a haunting shell of a city, a dystopian landscape ravaged by decades of neglect, turmoil, and chaos. The skyline, once adorned with sleek skyscrapers and modern architecture, is now a jagged horizon of skeletal remains that jut out like the broken teeth of a long-dead beast. Shattered glass reflects the sun’s dying rays, casting a kaleidoscope of bruised colors on the asphalt below—a cruel reminder of a beauty long forgotten. The air is thick with the acrid scent of burnt metal and decayed concrete, mingling with an omnipresent layer of unsettling silence. Occasional gusts of wind carry the faint echoes of a world that used to pulse with life, now replaced by a chilling stillness except for the soft, eerie whispers of distant sirens and the occasional crash of debris cascading from a crumbling building. Many structures are reduced to hollowed shells, their walls charred and blackened, graffiti spiraling across what remains; crude, desperate messages left by those who once roamed these streets.Nature is both a reluctant survivor and a cruel invader here. Wild weeds and stubborn vines claw their way through the cracks in the pavement, while towering trees, fuelled by the darkness of the disaster, spew toxic spores into the air. Battered park benches and rusted playground equipment remain as ghostly silhouettes of innocence, overtaken by moss and briars, inviting the specter of desolation to entwine with nostalgia. Amongst the wreckage, remnants of human life litter the streets—abandoned cars sit with their windows shattered, paint peeling under the oppressive sun. Their tires are rotted and flat, some turned into makeshift homes for the desperate and displaced. Darkened storefronts with shattered displays offer nothing but darkness inside, empty shelves echoing the sounds of lost commerce and a world that once thrived on trade and interaction. he former heart of the city, a once-bustling central plaza, is now a gaping void filled with rubble and refuse. Faded remnants of advertisements cling desperately to a crumbling wall—the last vestiges of consumer culture lost beneath an abyss of soot. A rusted fountain, long dry, stands at its center, the figures of child-like angels locked in eternal poses, their features softened by soot and time, yearning for the laughter of children that will never return.

This complete bullshit is something I just had Talk.ai spew for me. Even if I wasn’t the one who had generated it-- had I read it as a post somewhere else-- the way that is written is clearly, blatantly AI; not only can you tell by the style, but by many of the phrases used.

Also, not that it probably matters to anyone: I am currently in the middle of an extremely heated debate with my roommate-- she is enrolled in creative writing classes, but has been attempting to pass AI shit off as her own writing repeatedly. She is beyond pissed at me that I have told her seven times over the past week that I can tell the shit she has asked me to proofread and edit for her is 100% AI (she has admitted that she has not done anything but “put her ideas into it, but she is writing! Because those ideas are hers!” (Bullshit. It’s not your writing and ideas are fucking cheap.)

So yeah… between first-hand generating the bullshit to try to see why people like it (I still don’t fucking get it) and being asked to read “writing” that is 100% generated multiple times, I am confident in being able to tell what is and isn’t AI-vomited.

And no, I will not proofread, edit, or touch the trash she has asked me to. Not one sentence has been written by a human and if she fails her course because she refuses to do the writing exercizes on her own, let her fucking fail, she deserves it.

Exactly.

1 Like
556

So, AI can produce passable sentences. I’d argue that being simply grammatically correct is very far from high quality, and florid can be counter-productive here if anything – as the purple prose has been mocked for literal centuries. But as long as these sentences are but nonsense on a higher level, how can we even talk about this plagiarizing anything, especially the most renowned writers? Unless one believes that those writers also create nothing but disjoint nonsense.

So you did, but i don’t see what makes you think that. Plagiarizing generally requires you produce and present as your own something that can be reasonably tracked back to the original creator(s) you’ve plagiarized from. But this really isn’t the case with AI.

I have read works of multiple authors. I’m sure you’ve done it as well. Do you feel you’re plagiarizing them (and all of them simultaneously to boot) when you write your game, because you have knowledge of how those authors put their sentences together, and on some level this influences the choices you make?