it's like two mirrors pointed at each other
it's like two mirrors pointed at each other
Comments
139-
-
The internet I loved died like 15 years ago anyway. It got replaced by ads and misinformation bots and hatred.
-
I'm deeply jealous of people who experienced the early internet. For me it has always been like this.
-
My dude, YouTube used to not have ads
-
"You don't really know what it is you have until it's gone. Gone...gone."
-
People say that early internet was like the wild west, and I agree. And I feel like a weathered old cowboy who saw my homestead on the frontier get surrounded by buildings, cars, industries, smog, noise, and the never ending growth of the population.
Capitalism fucked us all, and it keeps fucking us, and it will kill us.
-
In the early days, there were a lot of porn ads before ad blockers came along. There’s a nice sweet spot somewhere between that invention and the internet of today.
-
I was there when people when people were sending around Goatse pics and all our info came from Alta Vista.
It was pretty sweet, apart from the prolapse.
-
I lived through the transition from BBSes to the publicly available internet. I had to go in person to pick up the equipment and software on a floppy.
The first thing I found was the Church of the Subgenius which I found hilarious.
-
-
-
I love love love model collapse. It's the absolute perfect conclusion to our Internet and this AI bubble.
The make AI. They realize that we're not ready for it. But short term profits overlook any and all ethical concerns. So they business bro and release anyway, unleashing a flood of slop and garbage content. All of which is impossible to tell if it is human made or not. But the only way to make their business continue is to train new models, which needs more human created content, of which they already destroyed and made imperceivable from their slop.
It's honestly just pure poetic justice. The ones who are hiring the next generation of AI models are AI companies.
-
In part, this is what Microsoft Recall is about: scraping end users' data at will to sort and feed to its LLMs, without the user ever seeing what is being scraped or having any real, lasting ability to shut Recall off and keep it shut down.
While I am aware that MS insists none of that is true, it is fact that 1) the snapshotted and OCR'd Recall data is now stored in an encrypted database that takes higher than average user skill to get into; 2) even users who turned off Recall saw it turned on again at the next Windows Update; and 3) even after MS said they were backing off Recall, MS continued to partner with hardware makers to create computers bundled with Windows 11 on top of the extra GPU necessary for processing all these Recall snapshots without making that sluggish Windows bloat even more sluggishly bloated than it already was.
So why all that money and effort, even as they claimed to be backing away from it, just to help a hypothetically forgetful user here and there? Data harvesting was always part of the payoff, why they were and are very willing to piss off a huge part of their own consumer base around the world by ending Windows 10 unnecessarily, and why even now they keep ramming Recall shit down the pipe when literally NO ONE wants it.
They get your data. At will. And as much of it as they like, without you ever having the opportunity to oversee what they're getting, much less curate it. And after feeding it to their LLMs, they get to aggregate and broker it to their "partners" as well. Never forget what MS did in Palestine and the partners they can and will gladly work with, all based on massive collections of quietly gathered user data that either should not legally exist, or is not known outside of MS and its partners to exist at all.
-
You know Microsoft isn't about the user experience the moment they removed free games from their distro
-
Calling Windows a distro, while technically true, feels offensive
-
-
In part, this is what Microsoft Recall is about: scraping end users’ data at will to sort and feed to its LLMs
That's also what Google has always done. Want a large data set of emails? Look at our new free service, Gmail! Need a lot of images to train machine learning vision models? Check out our newest free backup tool, Google Photos! and so on. When they want one particular data type, they launch a free service that just so happens to collect this exact data type from millions of users.
-
-
Explains why my personal blog, wiki, and git repo keep getting hammered by hordes of AI company scrapers. If AI was intelligent, they'd download a single snapshot every month or so and share. But no, eight different scrapers using thousands of different IP addresses (to evade my
fail2banmeasures) each have to follow every single blame and diff link when a simplegit cloneoperation would get them the hundreds of megabytes of content in one go.They are getting better, though. More hits are to RecentChanges on my wiki, so there seem to be some optimizations going on. But I refuse to increase my operating costs beyond a few USD/month to serve AI bots when I know barely anyone human visits.
-
The open internet will become divided into verified websites, and the rest will be left for bots to fight on forever.
It will be used as an excuse by our governments to force a ID verification system tied to your real life person, refuse it and fight it in every possible way.
-
Meanwhile we need enclaves of not corporate bullshit (like fedi but also) including bringing back webrings and old school chats. And usenet. And irl word of mouth.
Because fuck all that shit.
-
I mean, verification doesn't really help because in the end it's still mostly human posting the AI slop (I think).
That said, we'll probably need some sort of reputation system. Something like a revamped GPG or Web Of Trust, where you a) can tag users/websites you find trustworth and b) can see what other people you trust think about someone/something.
-
Who verifies the verification though?
-
-
So at this point we're gonna have to go back to reading books...
-
Books made before 2019. Amazon is absolutely filled with AI generated books nowadays.
In fact, this whole "consume media only from 2010 and earlier" idea is getting more appealing by the day. I'd rather watch an anime from the 80's where each frame was drawn by a human hand and somebody spent a week encoding it to extract all the details from the original analogue source, and the subtitles were made by a person who considered each nuance carefully as if their life depended on it, rather than watch a 2025 sequel to a prequel to a reboot of an existing IP where half the assets are AI, the subtitles are AI, the script is AI, and it's just the most generic mass appealing thing ever made.
-
It's pretty sad, creepy and hopeless, but this is exactly how I have been feeling lately with YouTube videos. If it's made post GPT, I am not inclined to watch it unless it's a channel I know has a stated anti-AI position, because at least then I know I'm getting human input. They are building up Plato's cave around us stone by stone, we can't even move out of the way, we are getting imprisoned, and the few of us who see it happening and shout are being drowned out by the noise of the billions around us who happily hum along.
The system itself needs to come down, with violence, or we won't make it, none of us, and honestly even if it does, I am not sure we are going to survive. I feel like I'm playing the fiddle on the Titanic.
-
The 80s and 90s have some of my favorite anime, movies, and tv shows anyway. I can't really think of any recent masterpeices aside from Interstellar and The Martian.
-
-
If you are a USian then you probably have access to a public library.
If you go to the website of said library, usually a city.gov or countystate.gov, you can create an account.
After you create said account, you can download apps(they will tell you but normally Libby or Hoopla, Hoopla hasn't been work for me but might be a me issue.).
Use the ID from the website on the apps and you can check out books to read. You also have access to comics and audiobooks. The Invincible comic is good, you should check it out.
-
Weird that third party apps, made by corporate entities, are needed for this. They're public libraries funded with public money, it should be one unified backend with libre applications.
-
You likely need to physically go to the library and prove your residency but yes. Then you can do the above.
-
My username isn't random coincidence.
-
-
Most new books are AI generated now.
-
Human forums. Throw in a tar pit for any scrapers
-
-
The Hapsburg chin of the digital age
-
I’m stealing this
-
-
How is it a crisis? I'm expecting/hoping LLMs will just get increasingly worse as they are fed on their own slop, until they collapse into unusability and the world finally returns to sanity.
-
The crisis is that companies will have to pay fair prices for human labor again, which will lead to less profits for the precious shareholders
-
Not just precious shareholders, it's a massive bubble that will impact the economy and peoples retirement when it bursts. They keep dumping more money into AI even though there are no returns.
-
-
The crisis is that if the internet is bad enough that they can't train LLMs on it, it's also useless for us human. And if the LLMs leave because there is no information and we come back, eventually we will generate enough content that it will be worth it again
-
This is said as though it isn't an immensely expensive endeavour to run these things and the only reason they're this prevalent right now is the overspeculation and starved growth of US tech companies.
-
Eventually is doing some heavy lifting. The costs of these data centers and the power plants to run them has to come from somewhere.
I don't know if or when the chickens come home to roost, but it could go badly for the US and China to find out their spent trillions to make a million versions of AI shrimp Jesus.
-
-
-
-
This is a superbly placed gif.
-
-
Its almost like its predictable outcome of prediction algorithms being used to generate content
-
I feel so bad for Gen Z. Their future looks so bleak.
-
It affects us all. Our future looks bleak, not only theirs.
Even if we got to enjoy the good times a bit.
-
Zoomers and Alphas have more years of that bleak future to enjoy.
-
-
I feel bad for millenials in their 30-40s who do not have unreplaceable expertize.
-
the percentage chart gives a timeline but not much else. makes it look like humans are writing less, I wonder what those numbers actually look like, if people are actually writing less or not
-
Yea. The Y-axis should be in absolute numbers to make a better comparison.
-
-
It is exactly a mirror, since all the percentages add up to 1 at any time
-
A very small percentage of articles are written by cats walking on the keyboard.
-
-
On top of this, the scrapers that feed the AIs are creating more and more traffic, and therefor load on sites that did not have them before.
-
AI articles cost <$0.01. Human writers cost $10-$100.
Buying a (human) slave for $100 sounds way more profitable than renting an AI if you ask me.
-
AI revenues are like 1/100th of what they need to be in order to pay off their debts. More if local communites stop subsidizing datacenters
-
You are thinking like someone caught in last century's economical paradigm. Now that they own the government, they can use tax payer money to subsidize it, and if it were ever to fail, they would use even more tax payer money to bail them out. In the words of the late, great, George Carlin: they have us by the BALLS.
-
-
yells at computer* NOW GO DO MY LAUNDRY, SLAVE!
-
Thank god, capitalism will save us!
-
-
If training on AI-generated text results in "model collapse" when AI models do it, I wonder what happens when humans do it.
-
Societal collapse.
-
Holy shit maybe that's what's been happening since the invention of mass media.
-
A form of group-think. A tendency towards absolute extremes. Like falling into a black hole.
I mean, we are genuinely fucked, aren't we? I mean we are, there is nothing we can do to stop this, because even if we do, not only is the damage already done, we have dug the pit so deep that if we remove one piece of the jenga tower, it will all collapse and bury us all underneath.
I've been thinking of joining a nomadic group of yak herders in the Himalayas, but I honestly don't think I have any skills they need, nor do I think they want my company.
-
-
-
How could they have ever known about the possibility of the thing that literally everyone everywhere told them almost immediately a couple of years ago when this fad really started charting (or perhaps sharting is more appropriate)?
-
Everything converges to generic sameness
When I hit this sentence in the OP, I realized AI is going to remain very popular with the average joe for a long time.
People who are tech literate, actively curious about the natural world, and do crazy shit like care about humanity (so most people reading this, most likely) will still reject it for the junk it is. But it seems the vast majority of people around me are not like Lemmy users. I mean, they are called normies for a reason, and I don't mean that in a derogatory way.
Generic sameness seems to be what the rat race pushes people towards. Maybe being burned out and having the economy constantly innovating new ways to bleed you dry makes the pre-packaged commoditized comforts from the advertisements too easy to accept. I look around and I see people anger-driving their pickup trucks and luxury SUVs to their jobs that they hate so that they can afford the cars they drive to get there. Plus they need to be able to afford beer and snacks for the game after they fill up those gas tanks!
These people don't care if the stories and tiktoks scrolling past their glazed-over eyes are AI generated. It only matters if those things can shake that stubborn drop of dopamine loose that just won't fall from the faucet. Just get through the day so we can do it again tomorrow.
-
I can not prove it but i think that already since a long time that most articles are copy + pasted and some kind of summary. Which might have happend automatically. Like someonr might write a real article on anything and then hundreds of sites copied it, without adding anything
-
The Huffington Post Effect.
Unfortunately, there's another side of this coin. The "original" content sources value freshness to grab attention and hyperbole to generate interest. The end result is Drudge Report / Brietbart / Alex Jones / Joe Rogan journalism that mills out innuendo, conspiracy theory, and quack medicine as Breaking News. And that becomes the "original journalism" all the other copypasta outlets reproduce ad nauseum.
-
Yeah I get the main theory you're going with and it's been rampant without AI bullshit. When sites became traffic driven to an extreme they started trying to grab whatever was the most rage enduring, verified or not. I forget the exact example but there was a guy that back traced a rage bait article and it was like 10 articles deep, and at the bottom it was a misquoted tweet that was complete misinformation.
-
-
The incumbents have their training data that can also be used for the next generations of AI. This is just them helping others to pull up the ladder to avoid more competitors.
-
Could you guys stop dumping your trash in the forest please? It obstructs my garbage trucks which I send to the forest to dump garbage in.
-
There's no way an AI generated article costs less than a cent.
So humans writing articles cost more than the billions they're pumping in to AI? I very much doubt it.
-
The problem is that the resource they consume to feed the AI, (human generated content) has become a limited resource, completely mined.
they could pay people to write, IE, news agencies pay writers to write and AI site are one of their clients.
you should get DMs from anthropic offering 50$ for your weeks posts and comments...
Instead they want to pretend they have still room to grow for free. but they can't
(That is just basic economic theory, I want those companies to fuck off already)
-
Mirrors face, fingers point.
-
Well, technically, the AI companies aren't making any profits so the actual cost is higher, and also the revenues from the AI articles are declining because people aren't interracting with them.
-
I feel like all of this research is extremely dubious. There's basically no way to know how much of the internet is AI or not. It's indistinguishable in most cases and especially at scale.
-
It is a fully rational conclusion, a logical inevitability, and what we are all seeing. But I guess if you can't prove it by some ever changing subjective standard, it isn't happening.
-
I'm not saying that it isn't happening. The opposite, I personally believe it is happening on a large scale. But I feel that it's extremely hard to measure and I'm not convinced any of these numbers are correct.
-
-
On scale using tools that detect likelihood of AI would work quite well.
For me the biggest challenge is sampling. I guess if you got all posts within a few minutes from each over and repeat that (to deal with time zone bias). I think it's possible but I'd want to look at the methodology from the primary source
-
-
If AI replaces humans then why would they need to continue training them? You don't believe in the AGI hype do you? You don't think our system is about anything other than efficiently racing to the bottom, do you?
Besides, humans will always figure out a better way, a new way, a fresh take. And there's your training data.
I don't get why it says "training AI on AI content makes it dumb" when people literally use synthesized data to carefully tune models. Today. I'm sure they'll improve training by the time the internet is fully replaced.
This all assumes that it won't take a human-like AI to replace humans. Whatever replaces us will be pitiful, but it will be supported by our institutions, trillionaires, and enough personal data to create fun headlines like "Elon knows so much about you that if each datum was a grain of sand it would be bigger than Saturn!"
-
Those of us who work on maintaining their critical thinking skills and book knowledge are gonna be kiiiings
-
Time to start working on the Blackwall
-
The good news is that AI will eventually eat itself. The bad news is that all we need to do to make that happen is not exist anymore.
-
-
Except they know this and the vast majority don't train ai with ai data.
Where do you guys get this stuff from?
I'm old enought to remember when people were freaking out the exact same way when the internet came out. History repeats itself I guess.
https://surrealevents.com/wp-content/uploads/2023/05/Internet-6-1024x694.png
The march of progress stops for nobody, yall are gonna be like boomers who still can't use a computer.
I'm still baffled about the expected outcome of replacing all content creation with AI for new events or ideas, like news articles.
If there are no humans to write the story first, where are the bots going to get the new information to plagiarize?
The expected outcome is internet collapse. The internet is a mass communication tool that allows bothersome peasants to organize and educate themselves if they are inclined to do so. Which means it must be destroyed.
The poor, disenfranchised, and uneducated are significantly easier to manipulate. So the goal of the 1% is to break society, eliminate the ability to educate, eliminate the ability to organize, and the use their near total control over mass media to assure everyone that they are better off under the thumb of the technocratic state.
That's a good point. The internet collapse also helps this new age of " information capitalism" to spread even further: any information that you may seek will be behind a paywall or intermediate by an AI that you have to pay a subscription to access. In fact, that's the best case scenario for all these AI companies: to kill free and reliable information.
This is not a conspiracy to destroy the internet. This is a tragedy of the commons. It doesn't take many bad agents to ruin it for everyone. AI slop is so cheap and has the potential to make profit, so it's a no brainer if you don't have a better way to make money.
The only reason the AI companies are crying is because now they need humans in the loop somehow. They'll need to source data from only known and verified real-person sources (like reputable news sites, magazines, personal blogs, and podcasts), manually verify data is not slop, and/or pay people to generate data for them.
Apart from being hit with what seemed like random ban hammers, that was one of the reasons I left Reddit.
It just seems so “the same”. Every comment chain is filled with the same predictable comments and they’re all predictably commenting on the same predictable content.
I hear lots of Lemmy newbies saying they stay on Reddit because of the user base in their niche communities. I hope those same people post or comment here a little bit in their niche instances before buggering off back to the black hole of bots.
I love that I can spot individual users again without tagging them.
I love the guy who uses that ye olde symbol instead of typing the letters “th”
This place is awesome because of the limited user base IMO. Very rarely have I been met with negativity here.
Yep. That about sums it up. It's a bit worse than that though, because they can also actively poison the AI and make it express more positive right wing sentiments too. For now that's still a bit tricky, because the AI goes mecha-Hitler, but don't worry, they will fix that glitch in the coming months.
It doesn't even matter if it does a poor job. I see people scrolling endlessly, 24/7, they're already zombified. The amount of people who can filter the bullshit or think critically about what they are fed online are in an extreme minority. It doesn't matter for the ruling class if there's some single digit percent of the population that don't buy it, fooling most people most of the time is absolutely enough to rule absolutely.
Anyone else inhabiting a very dark mindspace lately? We should start a club.
It's wild that most people don't understand this is happening. And it's being manifested in a Mafia type way.
I honestly have no hope for society. People really are ignorant and arrogant to the point that they are basically like farm animals.
Once Ai becomes Super Aware AI^TM then we will have no need for original stories......We just need a trillion more dollars of funding, all your power, fresh water, and for you to rewrite all intellectual property laws. Trust me bro, we are sooo close.
I swear tech bro billionaires forgot they were the ones who wrote the checks to the marketing department that dream this stuff up and have just been getting high on their own supply. The amount of people in silicon valley that actually believe AI is going to evolve into some version of the 40k Omni Messiah is too damn high.
I think it's the greatest sunk cost fallacy the world has yet seen.
"We've poured trillions of dollars into this, and convinced a bunch of people it's going to fundamentally transform humanity, so it MUST be a good idea!"
None of them want to be the first to admit that it was a bad bet.
That’s the thing that I wonder about. Like I know people who think all coding is going to be ai in the next couple of years. But where do they get data? Come up with better programs? Llama are, like you said, machines that just steal what’s already there. So if they are doing all the coding, we’re stuck right now? Nothing new or better?
Llama being the LLM corrected to is amazing
Why do you think they store and analyze everything you do on your computer, like when you use Windows? "Would you like a permanent spell checker that you can't turn off make sure anything you write is spelled correctly? No? Too bad, go use Linux. Oh shit sorry, it's in the hardware now, you can opt out if you want, but not really, we are going to read everything you write on your computer and your phone and analyze every conversation you have on your phone and it's all going to be processed and then eventually sold for scraps after we've extracted what we want from it".
That's where, and it's been happening for way over a decade already.
Most likely the only goal is getting short term massive investment like everything else. Just the capitalism playbook of find resource, hype it up and exploit it, draining all value before discarding it. Pump and dump.
There are legitimately AI researchers who are investigating "how can we make models that can be trained on AI generated data". These researchers know about model collapse (LLMs recursively trained on generated data will degenerate after a few iterations) for over a year now
... I hope I was joking
Wait, maybe that's the sustainable endgame. If the problem is that there is no new human contect for AI to learn from, then the AI companies can hire some humans to create literature, art, music, memes, etc. to carry on a stream of data for the AI to consume. They said that once AI had taken all the shit jobs, humans would be able to spend their time on art - maybe this is how it all ties together!
Haaaaaa...