But I wonder how much this is AI and how much we've sort of curated a slop pattern even before AI:
- Video game tips web pages with massive chunks of text / ads before you get to the inevitable answer of "hit A when X happens".
- The horrendous mess that Quora became with technically correct in some ways but also misleading answers to historical content.
- Medium articles about coding that are filled with irrelevant pics and blocks of text that are "not wrong" but also "not right" followed by weirdly specific code...
We had all that before AI.
The problem is that dopamine addicts generate outsized engagement. I know a literal crack mom who spends a solid 90+ hours a week watching accident videos to keep her brain triggered. The algorithm caters to her. Send promotional emails daily or more, constant notifications, recommend the same few videos over and over. Gotta get in there before she clicks another car crash video.
IMHO: Marketing is a top societal evil right now. If the media machine wasn't so desperate for content, AI wouldn't be a fraction of the problem it is. But with everyone obsessing over the next piece of content, fake AI presentations are mandatory.
Spelled "profitable". This is definitely something that's already happened/happening; see algorithmic timelines and the widespread sudden legalization and acceptance of gambling.
is
Though I’m starting to think that AI might improve faster than us so there might only be a diminishing margin of opportunity left to do this.
Eventually we’ll realize we can just send the bullet points and generate the prose on the receiving side. This will be great because most of the time the AI’s will be able to say “let’s just be a nop.”
I am less optimistic than the comic.
The algorithm is being trained solely for engagement. It is horrifying.
The Slop is the wordy vapid garbage that maximizes SEO.
Also, we have to look at the incentives: advertisement. Somehow, this is acceptable to consumers, profitable for companies and profitable for publishers. How, is absolutely beyond me... and it won't change so long as Google has a majority in the "search"-space as they are directly profiting from this.
It’s the reason advertising costs have ballooned digitally, and also the cause of many lawsuits that Google continues downplaying in the public eye.
What kills me is I have to hunt for the answer in Quora now. I just treat quora like I do pinterest, just back out and never return.
For me, it's one of those "quality of life" things that really improves my search experience (and is therefore worth paying for).
What AI gives us is vastly cheaper slop, so now it can be produced at a scale unimaginable to prior generations. No more paying some schmuck a penny a word to bang out "private label articles," so they were only practical as SEO. Now you can have a unique slop for every spam email, every search query!
Truly, we are making the world a better place.
Wot now? Somehow I managed to completely miss that.
Edit: ah, seems like it's mainly a twitter thing. That explains it.
But, consider that all that crap ended up on the web for a reason, and wonder how long before AI just injects it itself into its own results.
(And people have been giving me identical response with "advertising" in place of "marketing" for years now, so I'll say yes to both; the terms are effectively interchangeable in both motte and bailey cases.)
I do find myself sometimes even prompting for a shorter answer after it hits me with a blob of text ;)
It make sense to come up with terms to describe common patterns: Chain-of-Thought, RAG etc. are good examples of this. But the passion some members of this community have for being intentionally confusing is tiresome.
- 10 seconds intro
- 10 seconds yoooo guys wassss up
- 30 seconds build up
- 30 seconds showing what the answer will do
- 30 seconds encouraging you to post comments to the video
- 2 seconds to explain the answer
- 20 seconds yooo don't forget to pound that like and subscribe
If this is really what's optimal for the sacred algorithm, then that algorithm needs a serious tune up.
But that's not the scary part.
The difference with AI slop is just the enormity of the scale and speed at which we can produce it. Finally, a couple of data centers can produce more slop than the entirety of humanity, combined.
At work people often ask me for help with documents or translation. Or I see some friends' conversations. While Polish grammar is pretty difficult, it's not surprising to see messages with orthographic errors in 5 out of six words. You just live in a bubble of people who can read and write well.
There is absolutely everything wrong with that if it consistently invades and drowns out the voices of the well-educated elite.
The worst tyranny in this world is the tyranny of the ignorant against the learned. In its worst form, it can lead to mob justice and other horrible things.
Maybe that's not your worldview, but it is the view of many, and it's just as legitimate as yours.
But they're saying something. The characteristic feature of slop is not informality: it's fundamental meaninglessness.
Think only about your own consumption for a second. You're not going to engage with slop, are you?
I'm imagining that whatever your filter process is, that you manage to heavily engage with content that is mostly good and well-suited for you. Discounting Google search becoming crummy, of course.
AI in the hands of talented people is a tool, and they'll use it to make better stuff that appeals to you.
I wouldn't worry about other people. Lots of people like rage bait, yellow journalism, tabloid spam, celebrity gossip, etc. There's not much you can do about that.
My master's thesis was on a topic that nobody else researched (it wasn't revolutionary, just a fun novel gimmick), so I had to write filler just to have a chapter on a topic possible to find references to, in order to get the citations count, even if the chapter wasn't relevant to the actual topic of the thesis
So yes, I think that the push to create slop was there even before computers became a thing, we just didn't recognize it
Its not a horrendous mess for me. It works very well. Everything depends on what content you interact with as the algorithm heavily shapes your content depending on what you interact with. Its no different from any other social network.
When I was a kid we used to spend summers with my grandmother. It was an idyllica pastoral setting and we used to chase the goats around and catch butterflys.
[snip 3000 words]
...when I asked her for her recipe, it turns out she made cinnamon rolls by buying pillsbury ones at the grocery store! So if you don't want to be like grandmother, use 2 cups of flour...
Other examples include those books where each chapter generously estimated has a tweet worth of thought padded out with 35 pages of meandering anecdotes that just paraphrase the same idea. Like it's very clearly a sort of scam, the padding is there to make it seem like the book has more information that it does when you look at it in a digital bookstore.
Just like simple template generated SEO, template-written "content", etc. before.
In fact, a lot of writing about AI slop could be considered just as much slop...
They look for detailed pages, so pages are bloated with irrelevant information. They look for pages people spend a lot of time on, so the same thing occurs. Plus, the hellscape that is modern advertising means that rushing content out quickly and cheaply is encouraged over anything else.
AI will probably accelerate the process even more, but it's already been a huge issue for years now.
Google incentives don't matter much for honest website operators. They're only relevant when you want to abuse the system to promote your worthless bullshit[0] at the expense of the commons.
I really wish society started to treat marketing hustlers with the same disdain it has for robbers.
--
[0] - If it was worth anything, you wouldn't be worried about SEO all that much, especially back before it all turned into a race to the bottom.
“The term [‘slop’] has sprung up in 4chan, Hacker News and YouTube comments, where anonymous posters sometimes project their proficiency in complex subject matter by using in-group language.”
“Some have identified Simon Willison, a developer, as an early adopter of the term — but Mr. Willison, who has pushed for the phrase’s adoption, said it was in use long before he found it. ‘I think I might actually have been quite late to the party!’ he said in an email.”
The first substantive discussion of the word here seems to be this:
If anything, it originally started as calling things "goyslop", which you might be able to deduce is a portmanteau of "goyim" and "slop", the implication (given it's 4chan) of course being that it's low-quality stuff made by Jews that is foisted upon the "goyim" (non-Jews). To the point that I usually see people calling it "AIslop"... specifically to differentiate it from "goyslop", so pervasive is the use of the term.
I'm honestly surprised "slop" (in this specific context) is hitting the mainstream (apparently) given it's so closely married to anti-Semitic undertones. I assume it's kind of like Pepe? People see the cute frog or the edgy designation of things as "slop" not knowing that's kind of a minced version of how it's actually used on 4chan.
(This is in contrast to Pepe, which was popularized principally on 4chan and then exported by reactionaries.)
It would be strange if the slop industry didn't try to take over the word since they exist to preempt all charges against their authority.
Actually "engineer" as well in countries where it's not protected. Though the confused look on my German housemate's face when an "engineer" turned up to connect the cable broadband cable to the property was pretty funny.
So things that are woke get called goyslop (as opposed to Jewish as you imply).
Didn't claim it wasn't antisemitic, it is, a lot!
The way 4chan lingo seeps into mainstream Internet discourse is so annoying. It happened with "degenerate" too (which, of course, 4chan borrowed from the Nazis' Degenerate Art Exhibition).
One such idea was how the Internet was filled with garbage by all these agents (which were implied or stated were AI, I can't recall). They would subtly change things to be wrong. Why? Essentially to sell you a solution to this that filters out all the crap.
Currently we rely a lot on altruism for much of the information on the Internet (eg Wikipedia). AI agents will get harder and harder to differentiate from actual humans making Wikipedia edits. I don't think we're that far away from human-vs-AI Wikipedia edit wars.
I really wonder how much human knowledge will be destroyed by (intentional or otherwise) AI vandalism in the future.
> “Crap, you once called it,” I reminded him.
> “Yes-a technical term. So crap filtering became important. Businesses were built around it. Some of those businesses came up with a clever plan to make more money: they poisoned the well. They began to put crap on the Reticulum deliberately, forcing people to use their products to filter that crap back out. They created syndevs whose sole purpose was to spew crap into the Reticulum. But it had to be good crap.”
> “What is good crap?” Arsibalt asked in a politely incredulous tone.
> “Well, bad crap would be an unformatted document consisting of random letters. Good crap would be a beautifully typeset, well-written document that contained a hundred correct, verifiable sentences and one that was subtly false. It’s a lot harder to generate good crap. At first they had to hire humans to churn it out. They mostly did it by taking legitimate documents and inserting errors-swapping one name for another, say. But it didn’t really take off until the military got interested.”
> “As a tactic for planting misinformation in the enemy’s reticules, you mean,” Osa said. “This I know about. You are referring to the Artificial Inanity programs of the mid-First Millennium A.R.”
> “Exactly!” Sammann said. “Artificial Inanity systems of enormous sophistication and power were built for exactly the purpose Fraa Osa has mentioned. In no time at all, the praxis leaked to the commercial sector and spread to the Rampant Orphan Botnet Ecologies. Never mind. The point is that there was a sort of Dark Age on the Reticulum that lasted until my Ita forerunners were able to bring matters in hand.”
I always pictured the Ita as wearing elaborate robes with hoods darkening their faces due to their secretive nature. But in fact he describes Sammann as looking basically just like Gilfoyle from Silicon Valley, including the way he dresses. Which is amazing given the roots of the Ita (he describes the word as coming from Information Technology and the meaning of the A is lost to time, but it's obvious to a 20th century Earth-born reader it comes from IT Administrator).
There are so many delightful details in Anathem, it's well worth a second reading.
All the more reason to use local models and curated feeds from now on. Local LLMs can clean / firewall the bad stuff, and follow our guidance. They will be like the new anti-virus software. I've predicted early in 2023 that in the future operating systems and browsers will all sport a small LLM that will ensure we don't get abused by the internet and provide a "room of our own", where we have total privacy. It's already a reality.
No we can't. How many websites from 1998 survive today, in a form you can actually find (e.g. not the Wayback Machine)? In ten or twenty years, most pre-2022 data will be inaccessible.
[1] https://www.businessinsider.com/ai-spam-google-ruin-internet...
[2] https://www.ghostery.com/blog/how-to-prepare-for-ai-spam
Spam is trying to sell you something, e.g. an unsolicited email peddling a supplements.
Slop is low-quality content, e.g. someone taking a bunch of bird pictures off Google and posting it in a birding Facebook group.
Spam is an ad, slop is not. With AI, it is now much easier to generate slop.
Computer generated fake texts are spam, doesn't matter if they're selling pills to me or my attention to ad networks.
Slop = technically what you asked for, but intentionally created just to fill space and increase traffic/hits, generally of the lowest quality rendering it unusable
Would you still consider that "spam"?
Maybe the algorithms will be so good, and enough creative people will use these tools to generate truly exciting content that they wouldn't have been able to otherwise but it just looks totally dire to me for creatives at this moment.
The "content" industry (books, music, movies, all of it) has a systemic issue of which we are only just seeing the beginning. Namely, there is now so much content, and it is all so easily accessible, that the relative value of any one piece of content has fallen way, way down. There are only so many hours in the day, and only so many days in a lifetime, and only so many humans on the planet, and growth in that aggregate content consumption capacity has been far outstripped by the growth in content production capacity.
There's just obscenely more high quality new-to-you content than you can ever consume - and an increasing proportion of it is available very cheaply, or even free. Anything new faces an uphill battle against everything old - and now, against AI too.
This is going to get a lot worse before it gets better (and it may never get better).
The value of generic / impersonal content is rapidly approaching zero. The only thing that still has value is a particular creator of interest posting their next video — like your favorite YouTube channel, you’ll watch that.
It seems like the only way to succeed in this new environment is to be a real human person who builds a following / cult of personality around themselves and their content with its signature that is unique to them. It’s something like ‘releasing content that is personally signed’ where the person’s signature has value to a certain audience. The audience is ‘captive’ because they can’t get that ‘personal signature’ anywhere else. Even AI can’t deepfake it, because the perceived value of it is specifically that it is coming from a particular real human person.
Correction: pretend to be a real human (or even a hyperreal human), not to be a real human. This is the game YouTubers and Instagram influencers have been playing for over a decade - there's a team of people building a brand around the face of the vlogger/influencer, making them seem like a really nice and interesting human, where in fact the opposite is the case. The point of it is to exploit human vulnerability to parasocial relationships, creating a captive audience primed to be receptive to the deluge of advertising that follows.
Yes, this is one of the few ways for "content" to keep value these days. Which is ironic, given that the net value of it to the consumer and society is squarely negative.
Are you sure about this? How are you measuring quality?
For me, if something resembles advertising I consider it to be of very low quality. There are some exceptions, for example some of the movie work by Roy Andersson, but they are very few.
As far as I can tell, ad-discourse and ad-style permeates pretty much everything in contemporary "content". Every time I go to my library and open something older than me the language is like a fresh air, it's clear that someone put some intellectual work into it and there is a distinct character to the text, personality imbued by the typographers, authors and editors. This is very rare on the Internet, and whenever I come across it the typography is usually ad-adjacent anyway.
I hope this is the start of walled garden human internet. Web-rings, moderated forums, ect.
A common cyberpunk trope is a trashed net and a private net.
People who pay for content demonstrate that they have disposable income and are willing to spend it, which makes them prime population for advertisers to target. By paying, they're distinguishing themselves for the net-near-worthless population of free users. There's a huge pressure for advertisers to tap into that juicy population of paying users; it takes only so long before any given service succumbs to that pressure.
The problem is with the spam and slop.
Any sucessful walled garden will need to keep this crap out. banning adds is just one way. Content review is another.
There are some, but it's a rounding error compared to the ad-funded slop. Basing your business model around advertising creates strong perverse incentives, pitting you against your users/customers. Sometimes, a company gets enough return on advertising investment they can fund side ventures off their marketing budget - this creates an UBI-like low pressure space for quality content to grow. But stuff funded directly from ads? It's safe to assume the content there is secondary at best, and exists primarily as a vector to bait you so you can be exposed to ads.
> banning adds is just one way. Content review is another.
Right, but they're complementary. Content review gets easier when you cut out 99% of crap at the source by fiat, and with a hard line drawn, the reviewers are much harder for marketers to compromise either.
Agreed. Walled gardens cut out the crap. Invitation only.
Then they will complain that the internet is full of trash content that doesn't suite them.
This will happen regardless if you paid for content or not. The natural world is filled with parasites, it is an effective evolutionary strategy.
They are willing to view ads (proving they value the material) but as a rule are unwilling to pay any cash.
For a real world example, just look at the scientific journal system - researchers pay upwards of $5k to publish (after spending however many tens to hundreds of thousands on the science, out of their own pocket), readers can pay $50 per article or their institution can subscribe for tens of thousands of dollars (if they're lucky and have a good negotiator). Journals do nothing of value aside from hosting the PDFs (which absolutely does not cost $50/download) and facilitating anonymous peer review (which amounts to sending emails to a few academics who will review it for free, at no cost to the journal).
Even content that is worth paying for, like research, will quickly reach an equilibrium that maximizes profit while minimizing effort.
That's completely the wrong framing for this idea.
The problem is that you don't know if it's valuable until you've bought it, unless the seller gives it for free and just trusts you to pay later in the event you did.
If I see a new-to-me fruit in the supermarket, I can buy one to see if I like it, and I can be reasonably confident that my first taste will be a reliable indicator of if I should buy more.
People used to do this with entire newspapers, but (1) newspapers have been derided for taking nonsense for basically as long as we would even recognise them to be newspapers in the modern sense, and (2) dividing them up into separate web pages per article makes the challenge greater, as it's gone from 50p for the entire broadsheet based on the front page headline as an advert, to a "please subscribe" banner after seeing the headline and generic intro paragraph for a random article you were probably linked to because someone else thought it was interesting.
Not that there is no sub-groups that will happily pay.
For all their problems, I trust numerous media brands to not give me slop: The New York Times, The Financial Times, Monocle, Matt Levine, Linus Tech Tips, The Verge, hundreds of YouTubers and Twitter users, even Hacker News. Let media companies and creators who want to set fire to their good names get on with it, because it'll hopefully mean anyone doing a good, consistent job will rise.
Caught red-handed faking test results, the official response "if we actually did the tests we wouldn't be able to publish videos fast enough"
I still trust LTT more than a channel pumping out faceless review videos or one so unknown that there aren't enough viewers to even provide any scrutiny. To butcher the eponymous Linus's law: given enough eyeballs, all mistakes are shallow?
The other part is that practically speaking, there's enough good free content in most fields that paying doesn't get you anything better. It's not surfaced well enough by the algorithms, but it does exist, and it makes it so in most areas, there's very little reason to pay for anything.
Might not be the main reason, just that these unsatisfied users are so vocal, idk.
You literally can't evaluate content before you consume it in it's entirety or the at least the amount you wish.
Given that copyright should be abolished, adverts should be banned and content industry should move to entirely post-paid voluntary financing.
Something like gaming piracy where you play the whole game for free and if you really liked it you "buy" "a copy" to support developers and their investors.
1. People could forcibly seize and redistribute content already made in years 1980-2020. Even that would be better than slop.
2. People could read only those - like scientists or open source and public domain authors - either funded by the state or otherwise willing to publish their works for free at high quality without the monetization slop text. The content exists, but monetized slop hides it.
3. Actually good AI can compress multiple slop articles into useful, non-sloppy content.
I'd be perfectly fine if the internet consists of just math textbooks and science papers, and actually good articles automatically distilled from slop.
It has the potential to give us what we need.
The problem is that slop hides all that!
The problem with public domain stuff is that there is more than enough of it, but you cannot access public domain information because monetized slop has superseded it in the search results.
I believe that automated AI engines will eventually help individuals find non-sloppy public domain articles, or assemble them from slop directly.
But piracy is always a good option.
- Slop will continue to become cheaper to generate, and people will only notice the obvious stuff
- Hyperpersonalized content will abound, yet authenticity will run dry
- The lack of authenticity in electronic channels will drive a small segment of people offline into less fakeable (for now) social contexts
- Humans online will walk a treadmill of increasingly convoluted shibboleths / Gnirut tests (reverse Turing tests ;)) to self-identify as likely not AI-generated, i.e., subtly run-on sentences that are intelligible but slightly non-conformist to prevailing AI model outputs, and usage of old-school emoticons and other quirks
- Humans will walk on similar "Gnirut treadmills" for visual art, speech, video, and music
- AI models will gladly chase humans along these Gnirut treadmills, filling in canyons and sections of the Uncanny Valley with fractally sophisticated humanlike content
Though I try, I fail to think of a comparable scenario in our past, at least as relates to language. You can look around whatever room you are in and try to identify an object that was made by human hands rather than a factory process. That's a fact that always makes me a bit sad. I think we're headed in a similar direction with the language we consume. Craftsmanship falls by the wayside, and our world loses even more of the human touch that connects us with one another.
I think this segment might start small but I think it will grow rapidly if the utility of the internet is dwarfed with low quality crap. The belief that non-technical people won't catch on to the shenanigans and simply look elsewhere is a bad bet some are making and I think everyone living on the internet during covid gave non-technical people an intuitive feeling for all the manipulation and tenuous quality of the internet as a tool/public utility they can trust in any form.
The kicker is that if you ask GPT-4 about it, it spits out the same incorrect information, meaning that GPT-4 was likely trained on this bad data. FWIW, GPT-4o gives a much more accurate response.
And beyond slop, there will be AI models that do product placement. OpenAI's "publisher partnerships" deck explains https://news.ycombinator.com/item?id=40310228
So soon you'll go to a news website and get the political filter bubble that reinforces - or outrages - your prejudices to maximize your engagement. And in the middle of it, the AI will slip in that the brand of grill that caused the fire was rumoured to be ${insert_name_of_competitor_here} etc?
The big future for AI is to move slop beyond outrage and into intimacy territory. If rage was the engagement of the last ten years, then ending up only talking to AIs who pretend to care will be the even more addictive engagement of the next ten :(
Then it'll build up habits and routines with you. You'll feel good that your AI slop is actually good AI and is trying to benefit you.
All the while, the verified reports and data always supports your expectations and worldview. It's almost perfect, aside from the fact it sometimes cites The Federalist more than it should. And it convinced you to throw out all your pillows and buy a whole new set for your family. But you're able to look past it, because it's actually good AI.
It's better than what automated scripts could produce and cheaper than human generated copy. It has expertise of the whole Internet and its reasoning capabilities are often worse than those of a three year old.
Similar for sure, and yet something different.
They sent us new copy for our core marketing pages last week, and many sections simply sound non-sensical. As in it just didn't make cohesive sense, and it was clear a human being didn't write (or even review) the content.
This is the problem with new AI "slop". In the past, blogspam / SEO spam was at least reviewed and written by a human. Now, we have content getting published straight out of the mouth of text generators.
The quality of long-form content from text generators is significantly worse than even mediocre $10/hr copywriters in many cases.
On the other hand, there's indeed a lot of popular human-made fake content lately, especially in TikTok, like fake guitar playing and fake nonsensical DIY videos. So it's not an AI exclusive phenomena.
People should call it for what it is. Tried to find some answers on Google earlier in the day, and the first result pages were 100% generated slop. Funnily enough, any AI summary of the slop would be slop squared.
It's everywhere, and I hate it. What ways do people have to combat it out of their day?
https://web.archive.org/web/20240611214752if_/https://www.ny...
I’m not necessarily seeing the slop problem. People should always have been skeptical of content on untrusted websites.
Now, if reputable sources start trying to pump out content with AI, that’d be a problem. I suspect for those who try, they’ll quickly lose their reputation.
Yes, but people's output is limited by their ability to type words on a keyboard. LLMs and other generative A.I. aren't bound by this limitation, and can put out significantly more.
> People should always have been skeptical of content on untrusted websites. Now, if reputable sources start trying to pump out content with AI, that’d be a problem.
How do you define untrusted websites, or reputable source? Especially when Google - which should be a trusted, reputable source - starts pumping out garbage as they did?
On the second point - this is precisely what I’m talking about when I say if reputable sources start churning out junk, they will lose their reputation. This is a negative publicity event for google. If it keeps happening, people will no longer trust the information coming from google.
But there is a difference whether the ratio of good to bad articles is 1:10 or 1:10,000 one is tedious but managable, the other is hopeless.
With generative A.I., this kind of slop can be pumped out at an industrial scale.
It'd be like equating your neighbor dumping a bag of garbage on the roadside, with the industrial plant down the road pouring out thousands of gallons of toxic waste per minute into a river.
That's for the very lowest-tier video.
The usual formula, is that for every minute of runtime, you have at least 30 minutes of editing.
With professionally-produced video, I think it's triple or quadruple that.
Now used to describe anything that looks half-assed, poorly put together, etc.
I think of the buckets of scraps you would feed a pig. The stuff the humans don't want to eat and would just as well put in the garbage bin.
Irrelevant, but also fantastic.
At this velocity it’ll make some categories pure noise by end of year
there will be a backlash against robotic phone assistants in support centers.
support businesses that dont put out slop or ai-garbage.
https://www.sciencedirect.com/science/article/pii/S000768132...