this post was submitted on 06 Oct 2023
178 points (100.0% liked)

chapotraphouse

13924 readers
606 users here now

Banned? DM Wmill to appeal.

No anti-nautilism posts. See: Eco-fascism Primer

Slop posts go in c/slop. Don't post low-hanging fruit here.

founded 4 years ago
MODERATORS
 

I was looking for information on insect vision for a report I'm writing and came across an internet article. The first thing I noticed is that there was a picture under the header that the article called a mantis, but the anatomy was all wrong. It was covered in ocelli, for one.

Then I started reading, and as someone that has used chat GPT before... Yeah it was clearly written by an algorithm...

This is the site. https://healingpicks.com/

Look at the pics closely and you'll see they're messed up.

How long until the whole internet is just this shit? This is why there needs to be rules and shit about this, I don't care if it stifles the 'entrepreneurial spirit' of AI or whatever. The droves of lazy copy and paste style news and 'top ten' sites were already bad enough, but this shit is bringing the internet to a new low in quality.

I can't wait until this shit is banned by China or something.

top 50 comments
sorted by: hot top controversial new old
[–] KobaCumTribute@hexbear.net 67 points 2 years ago* (last edited 2 years ago) (3 children)

I used to think the cyberpunk genre trope of finding information online being some difficult task requiring a trained professional to be quaintly anachronistic, but the proliferation of AI generated nonsense has given the concept a new life to the point that "ok yeah soon enough it really will take a skilled professional with up to date detection tools to actually parse through this sea of literal nonsense won't it?" actually seems like a realistic future.

And yet all the discourse on it gets narrowly focused into the useless dead end of fucking property rights, a fight that's lose-lose for the public, win-win for business, and does nothing to stop ad farming/well poisoning/astroturfing spam bullshit at all. Like the property rights thing needs to be solved by making generative AI a poison pill that prevents a work from being copywritable at all and is retroactive poison against it as well (Disney used a deepfake AI in Star Wars once? Star Wars is public domain now, because fuck you; a script used AI autofill? The entire property and all licenses attached are now public domain, because fuck you), and the rest has to be solved by criminalizing ad farming spam and making the use of generative AI in it an escalating factor that turns it into a more serious crime.

[–] FourteenEyes@hexbear.net 42 points 2 years ago (5 children)

What I know from playing Cyberpunk 2077 is that being a fixer is half figuring out Google results and half putting together slick 3D powerpoints and putting them on a datashard in a fancy little box that you present ceremoniously to someone

[–] keepcarrot@hexbear.net 29 points 2 years ago (3 children)

The other part of your job is knowing a guy. Tbh, full time job.

load more comments (3 replies)
[–] Tankiedesantski@hexbear.net 18 points 2 years ago

A Cyberpunk fixer is just a venture capitalist for violence.

load more comments (3 replies)
[–] ClimateChangeAnxiety@hexbear.net 38 points 2 years ago

Disney used a deepfake AI in Star Wars once? Star Wars is public domain now, because fuck you

sicko-wistful

[–] CliffordBigRedDog@hexbear.net 35 points 2 years ago

Im gonna be a netrunner

[–] BigBoyKarlLiebknecht@hexbear.net 54 points 2 years ago (1 children)

An archive.org backup from like 2019 is going to be the low-background steel of human knowledge

[–] Tachanka@hexbear.net 25 points 2 years ago (2 children)
load more comments (2 replies)
[–] TrudeauCastroson@hexbear.net 53 points 2 years ago (2 children)

I can't wait until it snowballs into itself and AI is using AI generated text for its own models and gets shittier.

The internet is going to be useless lol. I should learn how to read/find real books in a library to find info.

[–] Frank@hexbear.net 46 points 2 years ago (2 children)

I'm told this is already a problem for LLM model training. There's so much algorithmicallyl generated garbage now that they can't scrape the internet for training sets without feeding garbage in to the garbage machine and getting even noisier garbage out.

[–] Sir_Kevin@lemmy.dbzer0.com 34 points 2 years ago (1 children)

I had a gig for awhile training LLM models for exactly this reason. They wanted human made content to train the models with. Problem is, many of the human trainers are using ChatGPT to get the work done faster. So even the companies going out of their way to avoid using shit content, are using shit content.

[–] GaveUp@hexbear.net 19 points 2 years ago (1 children)

They're publishing books written by AI too

[–] Beaver@hexbear.net 44 points 2 years ago (4 children)

Kojima's prophecy from MGS2 was that AI would create a version of the internet where all the trivial stuff is filtered out from the “truth” so human society doesn’t regress in the then-new social media era.

...instead, we got a nightmare inversion where AI turbocharges the creation of trivial and outright hallucinatory content that slowly overwhelms and annihilates all sentience on the internet.

[–] SkingradGuard@hexbear.net 30 points 2 years ago (1 children)

The difference is in Kojima's art, that "AI" is actual AI and not overly bloated statistical models trained on their own bullshit.

Like you say, we're so much more doomed.

load more comments (2 replies)
[–] hexaflexagonbear@hexbear.net 43 points 2 years ago

How long until the whole internet is just this shit?

They've been the majority of top results from before when chatgpt was a thing. Usually if you search up a release date, or a comparison between two products it's bot generated shit. Hell, they seem to have gotten worse. Probably because more mainstream places do it now, or they're just cocky and instead of spec dumps they're doing harebrained stuff like letting the AI write that the higher megapixel camera is automatically better.

[–] SkingradGuard@hexbear.net 43 points 2 years ago* (last edited 2 years ago) (1 children)

The LLM industry is going to crash and burn so hard and I'm all here for it. A bunch of techbros hyping up what are just big statistical models was never going to last.

[–] Smeagolicious@hexbear.net 40 points 2 years ago (3 children)

Less LLM articles but “AI” shit has made finding art reference material so annoying. Search results are 50% sponsored ad links and a good 10% are shitty stable diffusion prompts. Even worse if you’re looking for generic scifi fantasy bs, then it feels like 70% of results are peoples ai robo waifus sometimes

load more comments (2 replies)
[–] Owl@hexbear.net 37 points 2 years ago (6 children)

I warned everyone this was going to happen and nobody believed me.

Being right doesn't really make me feel any better about it though.

load more comments (5 replies)
[–] Maoo@hexbear.net 37 points 2 years ago

Microsoft going all-in on LLM is just a win-win to do sneaky layoffs and destroy Google by making search absolutely worthless.

[–] Awoo@hexbear.net 31 points 2 years ago* (last edited 2 years ago) (1 children)

Shinigami eyes extension but for AI content could be a useful browser extension. Enable people to opt out of even clicking links that has AI content or allows AI content.

Or alternatively, what about going whole hog? Something that functions like adblocker except ai-blocker. Remove the content from search results and being seen as linkable content entirely.

load more comments (1 replies)
[–] Mardoniush@hexbear.net 31 points 2 years ago* (last edited 2 years ago) (1 children)

What gets me is that it would take 5 min to vet these articles and make minor changes, and they're too lazy for even that.

That said, cheap content mill crap was already 99.9% of articles. Most content is written purely for SEO>

[–] dannoffs 28 points 2 years ago (1 children)
[–] D61@hexbear.net 26 points 2 years ago

For a brief moment, I thought we'd done away with websites that were just random walls of text attempting to take advantage of early stage search engine crawlers. yea

[–] Melonius@hexbear.net 25 points 2 years ago (2 children)

Yeah it's really bad. I was looking up information on healthcare coverage and it's "8 things you need to know" or "36 questions you might have" before selecting a healthcare plan.

I was thinking it'd be great to have an option to filter out any list results for a search engine.

load more comments (2 replies)
[–] MerryChristmas@hexbear.net 25 points 2 years ago (2 children)

I'm assuming someone has already mentioned the problems this causes when googling health concerns, but have y'all tried looking up animal advice recently? I was trying to find some info on a rare parosphromenus species I'm working with. In the old days, this would have taken me to some niche forum threads full of posts from dedicated conservationists. Instead I found dozens of articles with care requirements listed that would have killed the fish in a matter of weeks.

At present it's still pretty easy to recognize GPT text when you see it, but it's enough to fool web crawlers and by the time they roll out GPT5 it will probably be a lot more difficult for humans to tell it apart.

load more comments (2 replies)
[–] SorosFootSoldier@hexbear.net 23 points 2 years ago

Oh it has an orgone section, that's how you know it's good quack shit.

[–] WafflesTasteGood@hexbear.net 22 points 2 years ago (4 children)

Understanding How Long Does It Take for Miralax to Work

Some of theses pictures really dont seem to match the article very well. Also im loving the really funky words and numbers that look like old captchas.

load more comments (4 replies)
[–] dragongloss@hexbear.net 19 points 2 years ago (2 children)

This hand coming out of this dollar store Edward Cullen has got me rolling.

load more comments (2 replies)
[–] bquintb@midwest.social 17 points 2 years ago (1 children)
load more comments
view more: next ›