/int/ - International

That anyone using the term "AI slop" should fuck right off back to the 4chongs where they belong

>>11236 Using 4chongs unironically is probably a bigger crime than using a term coined by people who try to understand the advent of automated content production under human supervision and its consequences.

Our grandkids will look back on human-produced content like we look back on drawing water from a well, making fire in your kitchen and traveling to the library to look up information. Incredible what effort people had to put into a pretty basic version of entertainment or education. How the way there will play out is only interesting for us witnesses.

Me and a buddy were shooting the shit yesterday and we suddenly realized that in the near future people will wear augmented reality glasses with filters to make partners or potential partners look more attractive. I don't know why exactly, but the thought horrified me.

>>11235 I hate AI defeatists more than AI. Stop crying and add it to your skillset.

Test

Wenn I use chatgpt for work problems it feels similar on how google used to feel back when it wasn't as shitty as today. And you get your answer way faster. Only problem is that you can't trust it completely. Also it is pretty decent in writing bydlon scripts.

>>11288 was that a test? Did we pass?

>>11287 What skill would I add to my skillset? Wasting time on getting a machine to do something I could have done better in the same time?

>>11291 Good that you have the self esteem and skill to claim AI cant do better...... usually the people I see complaining never did anything artistic and can fuck off.

>>11294 > usually the people I see complaining never did anything artistic In my perception the loudest complaints about "AI" come from actual artists

>>11300 Who gives a shit? I don't see any jobs that will reasonably survive the next 20 years, artists are just the first in the execution line.

I hate it and I worry about what it will do to art.

Waiting for the moment when AI content is better than human.

>>11338 It will never be better. But people will still use it because it's cheaper and the world will once again be a worse more boring place because of the maximization of profits.

>>11344 Arguably that had already happened with the introduction of stock image/music databases, hobbyists invading Social Media and the emergence of "gig economy" platforms around fifteen years ago. That was, relatively speaking, the worst hit for most professionals and has driven very large numbers of photographers, illustrators etc. out of existence and kept many from getting into the profession to begin with. Bernd wanted to become a photographer but decided against it after a professional animal photographer who used to get booked for expensive calendars explained to him what had happened. Twenty years ago he would get paid very handsomely for taking custom pictures for a calendar. Fifteen years ago the calendar company switched to stock photos. Ten years ago they started looking for semi-professional and amateur photographers on Social Media and bought their pictures for little money, or no money at all ("paying" in "visibility"). He is actually no longer a photographer today, his actual business now is taking rich people on photo tours to exotic locations. All his own photography is basically a hobby subsidized by being a travel agent. Most professional photographers who aren't lucky enough to work full-time for a company or agency end up like that. The few illustrators, writers and musicians I know have had to fight for scraps on platforms like Fiver for years. I know people who make CGI for movies and AAA games who haven't been able to make an actual living for years, living at least partially on government subsidies. All of that happened long before "AI". Generative "AI" makes it even worse, but the situation was already so bad before, the relative change is not that big. A lot of people also still prefer to hush around an artist slave over having to find out how to prompt the stupid AI correctly. A very decent, hand-drawn music video by some Nepalese guy on Fiver is literally 50 bucks now, and for those 50 bucks you can order him around for a week and get exactly what you want. On most "AI" platforms 50 bucks don't buy that many tokens if you start requesting changes.

>>11344 of course it will be better and it won't take long now

>>11389 How could a thoughtless, emotionless copymachine with no sense of humor ever produce better original content than a human?

>>11398 with science and mafmadics

>>11344 >more boring place Profit driven use of AI will help bring an era of mediocracy and some romantic youth will take their own photos again and have real human partners instead of entry level gf avatar

I really wonder if LLMs will hit a hard wall and the bubble will bust. There probably isn't much human created culture left to train LLMs on, and if you train them on media made by other LLMs you get a negative cycle.

>>11401 That's not how it works. It's just an algorithm of already created material. So at best it can be as good as humans, never better. >>11418 It will hit a wall eventually but it's already good enough to fullfill many tasks. So I don't think it is a bubble, just not as great as some people will think it will be.

>>11419 >never better where are the proofs >>11418 you can always shoot more video and synthesize more data as you go

>>11418 > if LLMs will hit a hard wall They already have, see the GPT 5 disaster.

>>11419 I can't wait for spinning jenny to allow factories to employ less people while I ask Claude about photography composition.

>>11419 > It's just an algorithm of already created material. So at best it can be as good as humans, never better. Let's assume "better" can even be defined in this regard: Yes and no. The algorithm starts on a randomly generated set of data and then tries to turn it into something that meets the constraints. This same mechanism that makes it "hallucinate" can lead to something that looks "creative". Arguably the "hallucinations" are "creative". Some of the stuff that comes out is just as unique as something a person could have made. The iasue is that it's not directed on any way like a human would. A human would recognize somehow that what has been created is in some way unique and purposefully add on top of it, e.g. enforcing key consistencies between works. The current "AI" algorithms can't do that. It all stems from complete randomness at the first step. >>11421 > synthesize more data No, you can't. What would be needed is less data but annotated in a much, much, much better way. That would be magnitudes more expensive than just scraping the whole internet, though.

I don't have any. I think they're fun.

>>11546 I wish they didn't have the piss filter though.

Its funny that people often have nostalgic rose tinted glasses and on the future they will be vindicated, because culture from now on will be an endless stream of recycled half broken half assed LLM generated shit. It's harder to fix and go over LLM generated content than to create it.

>>11557 You can actually fix that. https://gpt-tone.com

>>11422 You are a retard and have no idea. "GPT-5" was an efficiency consolidation + PR smokescreen to implement the router (give customers less compute for the same amount of money, adjustable as necessary). The real GPT-5 in a technical sense was released as GPT-4.5 after it was too expensive to be even remotely tenable yet (you *could* call that a flop, but the patrician's answer of anybody who could afford to use it and has taste (almost nobody) is that is still the best model out even though its posttraining was cut off before they could RL it to benchmaxx). The real GPT-5 in terms of impact was o3 and it was an off-the-charts retarded business move not to release that as GPT-5. o3 was as much of a jump from gpt-4 as gpt-4 was from gpt-3 and not even slightly a flop.

>The company lists 4,749 AI-generated shows on Spreaker under its Quiet Please brand. It claims its podcasts have been downloaded 10 million times since September 2023 and it is publishing 3,000 new episodes a week. >In an interview with The Hollywood Reporter, Inception Point AI CEO Jeanine Wright said that “in the near future half the people on the planet will be AI, and we are the company that’s bringing those people to life.” She said that critics “who are still referring to all AI-generated content as AI slop” are “probably lazy luddites