Works fine for me, 10 minutes after this was posted. I'm not sure if this thread unrolled caches responses or not, but if it does maybe those who can't get through can read it here.
I managed to connect. I don't know how much I can copy, since it's quite a long thread, but here is the first message.
> Something exceptionally grim is happening on the Internet.
> In the last few months, the constant flood of algorithmically generated junk content has kicked into an AI-powered overdrive, and it is cutting a swath of destruction as it overwhelms search engines, filters, and moderation systems
> Call it Gresham's Law 2.0: bad content drives out good.
> I'm starting this thread to document it, because there is a lot happening all at once.
With everything in consideration, I think the real problem are incentives.
When human-created content competes and loses against AI-generated spam, the system is ill-designed.
This is the culmination of algorithms that have rewarded engagement over everything else. Creator platforms look at content made by experts that are high quality and take time to make and effort to consume, and judge it worse than trashy content full of controversial drama pushed every week without research or effort based on arbitrary metrics. Now we've reached the point where the creator is unnecessary and all that matters is engagement.
Creator-less engagement will be the future of the web. A technology that triumphed by connecting people around the world will die connecting you with nobody but an AI that acts as wall and filter between you and the rest of the world.
>> In the last few months, the constant flood of algorithmically generated junk content has kicked into an AI-powered overdrive, and it is cutting a swath of destruction as it overwhelms search engines, filters, and moderation systems
Nothing that wasn't happening already since some 15 years or so.
I don't know. People have been saying quite confidently that "AI" is "going to" "take over" the web since ChatGPT was announced. I live online and my only exposure to AI-generated content (that I recognize, of course, but that's another discussion) has been on purpose - when I look for it, use it myself, or others share what they're doing with it. I guess I'm still waiting for it to take over?
I think what people mean by "the web" is really just the 4-5 large social media sites. Those have been utterly filled with spam/scam/fraud for many years, so I fail to see a difference.
>I think what people mean by "the web" is really just the 4-5 large social media sites. Those have been utterly filled with spam/scam/fraud for many years, so I fail to see a difference.
No, as the linked thread shows, this is happening everywhere, not just the 4-5 large social media silos. There is another completely separate thread knocking around on HN somewhere about how AI spam is filling up OpenAI's chatbot store, and it's been well documented as a problem with Google image search, Quora, Linkedin and numerous other sites.
It's apparent that you yourself don't notice the problem but that doesn't mean the problem doesn't exist.
> when I look for it, use it myself, or others share what they're doing with it. I guess I'm still waiting for it to take over
Dunno if you follow gaming news at all, but AI spam has become the rule if you look for anything on new popular games
If you want to know how to solve a puzzle or compare two characters or something, it is almost guaranteed that you will come across pages of search results that seem like generated content
Current amount of SEO spam is nowhere near the 2007 levels, when the entire first page of Google results was your search query with stuff like "xxx", "2007 crack", or "download free" attached to it.
>If you want to know how to solve a puzzle or compare two characters or something, it is almost guaranteed that you will come across pages of search results that seem like generated content
This kind of content was never genuine. It was always low effort trash made to cheat search engines. It was either outsourced to copywriters regurgitating the same content over and over, or automatically generated when possible. Automating it with LLMs simply made it obvious.
Amazon book scams is the age-old tradition at this point, it exists for well over a decade and is a result of Amazon not willing to spend even one cent on manual curation.
OP might have a point, but it's still mostly typical microblogging ragebait with unrelated stuff mixed in.
> Dunno if you follow gaming news at all, but AI spam has become the rule if you look for anything on new popular games
At least in regards to game articles, this has been my experience for at least a decade. It wasn't AI spam, but it was spam everywhere non-the-less. I'm sure it's even worse now, but it was bad enough even 10 years ago that I've stopped bothering using google to find most game information years ago.
This was a plot point in the Cyberpunk lore. The Internet got overrun with AI to the point that they had to build a new Internet in parallel for humans, with mega-companies being the stewards of course. In a way, Pondsmith called this decades ago.
Everything that happens is cancelled out by the entropy increase generated for it to happen. And it will also degrade over time due to entropy seeping into the enclosure where it happened.
Nothing ever happens[0] indeed.
[0] Without costing something else coming undone. We can only pump low entropy around, we cannot generate it.
"Google is taking action against algorithmically generated spam. The search engine giant just announced upcoming changes, including a revamped spam policy, designed in part to keep AI clickbait out of its search results."
Even before the AI floodgates were opened, there was a huge problem with ads and SEO-optimized junk written by humans clogging up Google, and pushing the good stuff into the SERP netherworld. People complained for years (including here) with no discernible improvement.
It's one of the reasons HN turned into one of my go-to places for finding the good stuff - obscure blog posts and social media threads and websites that share facts, recollections, opinions, and expert insights that can't be found anywhere else. However, HN and a few other high-quality discovery sites can only go so far in surfacing such content.
Now it looks like Google, Amazon, and other Lords of the Internet are going to cause further damage with inevitable false positives as "AI spam detection" is unleashed on the Internet. Of course, there will be no recourse other than victims pleading for human intervention on a forum like this or hoping it goes viral on social media.
I mean they’ve had the opportunity for years, with spam and seo, already. They just chose to join in on the party and drive any reputation they did have into the ground.
Then they tried to play the reputation game again by paywalling articles, but using the dumbest mechanism of free-article-views that basically just continues to incentive seo spam from their own writers. It remains unclear to me why they haven’t adopted a model like LWN — paywall freshness instead of content altogether, and rebuild their reputation and relevance
I can’t understand why they don’t apply the same mechanism as how they filter spam out of gmail and friends: why can’t I the user block/derank/prioritize domains in a custom search results ordering? Aggregating this data at google etc scale would give a pretty good idea as to which domains are junk and eventually automatically tweak results rank this way. Isn’t it how gmail spam « detection » works (or worked?) for the most part?
Kagi already does that (minus the automated part) and it’s one of their best features imho.
1: Use Firefox.
2: Don't use Google anything.
3: If you can't do 2, at least do 1 and add a search blacklist extension like ublacklist.
4: Actually, use uBlacklist (or whatever) anyway, it will let you add other search engines, and block the annoying sites that clog the search results.
Here is my current search result blocklist, these sites earned a position on the list by either being intrinsically unwanted, (i.e. news sites, social media), completely useless (learn.microsoft.com, a site specifically written so that you can't learn anything from it (Oh, I define a few of the function for a specific thing but can't be searched and only rarely show any examples, hope you have a dev DC to practice your poorly written powershell scripts on! of course, we never tell you that there is a book series that actually teaches you microsoft stuff because otherwise more people would understand how easy it is to operate windows), or SEO spam optimized BS that prevents me from finding what I am actually searching for.
I somehow missed that article when it first appeared on HN[0], thanks for linking to it!
Surprised to "^F FTC" the article and comments, and not find any mention of the Federal Trade Commission. Given some of their prior rulemaking and caselaw around fake reviews and undisclosed compensation for endorsements, it seems the sort of thing that could be easily be prosecuted under false advertising laws. Unfortunately, I'm inclined to believe that the decision to let dying legacy media companies get away with fraud is probably a political one.
What's interesting to me is how quickly AI generated art came to be expected, unexceptional. Every random blog post has a couple AI generated images now, they are completely unremarkable and uninteresting. The novelty wore off basically instantaneously. They're just the new corporate art style.
It’s only been a year. As SORA shows, state of the art images and videos will soon be undetectable by human eye and it’s very soon coming to proliferate every “content” you interact with.
Does it really matter? AI will generate the images or video according to the introduces prompts so the creativity falls in the one entering them, you could be an excellent piano player but that doesn’t make you a composer.
>> What's interesting to me is how quickly AI generated art came to be expected, unexceptional. Every random blog post has a couple AI generated images now, they are completely unremarkable and uninteresting. The novelty wore off basically instantaneously. They're just the new corporate art style.
> It’s only been a year. As SORA shows, state of the art images and videos will soon be undetectable by human eye and it’s very soon coming to proliferate every “content” you interact with.
Even if the "AI" generated images will be "undetectable by human eye," they'll still be at least "completely unremarkable and uninteresting" and more likely crappy and stupid (even if all the people have the right number of fingers now).
I predict the floodgate of that "AI" generated stuff will reduce the production of actually good and creative art and images and make what is produced harder to find.
Sora costs a fortune, like dollars per minute. It's DOA. Also when handpicked demo content shows cats sprouting and losing limbs in a 15 second clip and that's the best they got, not sure anyone has to worry about undetectable fakes, certainly not with the total lack of creativity I've seen on the image side of things.
I'm perplexed that there are a significant amount of people on a tech forum that can't do a simple linear extrapolation from last year to one year into the future. That would get you in the right direction, except that we're on an exponential curve.
Not all problems are linear - people have been saying for over a decade now that self driving cars are imminent, but at this point it seems like most have actually given up on them.
Well… it’s possible we’re on an exponential curve there too. The early part of exponential can be pretty slow. And it depends on how you measure progress. We are seeing exponentially more training data collected in terms of miles driven every year. And likely exponentially more processing power applied to the problem space.
I think it is like if you gave everyone spray paint cans and said go paint, you wouldn't expect all the buildings around to have giant Basquiat inspired murals everywhere.
It would just be ugly, shitty graffiti, everywhere.
Maybe people have had enough of worrying about copyright infringement, licenses, attribution etc. and just want a quick sketch to get the idea across. Art is a tool for expressing ideas after all.
I ran into this today while searching for a way to merge channels in Slack. A google search turned up this result https://www.process.st/how-to/merge-channels-in-slack/ which describes an in-app method that does not exist.
It should also be banned from AI training data sets. Good luck to anyone trying to train an AI!
As far as I can tell, GPT-3.5 and GPT-4 are extremely aggressive about making up features of widely used software that don’t actually exist. I lost count of how many tables and columns in MySQL’s INFORMATION_SCHEMA they made up for me yesterday, often complete with elaborate SQL commands to query them.
The next generation of models will train on this crap.
Maybe it will put emphasis on the social aspect of internet. Maybe people will be more sceptikal. by having t know your source, having small but trustworthy network based on real people...
maybe that will change the shape of internet and lead to premium internet service where experts are paid to provide currated informations.
I had a family dinner last week, in the Pacific North West. One topic that came up was evil government regulations that required heat pumps, with the follow-up claim being that they do not work below -5C.
I cannot express the amount of shit I took for stating, un-movingly, that heat pumps work below -5C. I was shown Google search results, webpages, and was hit with professional experience from my brother, the red seal gas-fitter. I don't expect a "That's too bad" - I already know I'm right, and the US DoE agrees with me [0].
I think your comment is wrong. People don't search to find correct information, they Google to confirm pre-concieved biases and "verify" what they're already certain of. And I think computer-generated websites and answer boxes will only exacerbate the issue.
I wonder how much of this heat pump argument was miscommunication. My personal experience is that heat pumps don't work worth a damn in below-freezing weather, and always need to resort to resistance ("emergency") heating, which is extremely inefficient and expensive. Of course, my experience is with frequently decades-old units, or at best whatever cheap-o unit the landlord bought, not the very latest state-of-the-art system from the best manufacturer.
The other thing I've found with heat pumps is that, even when they are "heating" the air, it feels cold. The problem is they're efficiently heating the air only by a few degrees and blowing it around, but when slightly warmer-than-ambient air blows in your face, it feels cold. Gas-fired furnaces don't do this: they heat the air much, much more than ambient temperature, so it actually feels quite warm when blown in your face. (The problem here is frequently poor placement of vents, but this is common with homes and also offices.)
Sounds like you might be after a cold-climate [0] heat pump. They'll use advances in things like variable speed compressors and flash / vapor injection to draw useful heat from the air at well below -5C, without the need for auxillary heat sources like resistance heat.
I'm doubtful it was simply mis-communication. Another point of contention was that heat pumps can be over 100% efficient - to the point that asking "Will they be more than 100% efficient with outside temps at 15C or 25C, and in indoor target of 20C?" got an emphatic no. That's less a miscommunication than a poor education, IMO.
"Miscommunication" was probably poor wording on my part. What I meant was that their experiences, like mine, probably involved years or even decades of poor experiences with heat pumps in cold weather, so they were assuming that heat pumps haven't significantly improved in that regard.
For me Google Search is the big one. It has sucked for a few years, but in the last few months it’s been increasingly worse, to the point where I can’t Google anything anymore and get meaningful results. The top results (2 first pages) are systematically SEO/AI junk, and Google goes out of its way to ignore the keywords I add, even when I use natural language as we’re supposed to now.
I’ve completely moved my searches to ChatGPT (general questions), Reddit (consumer opinion, pros/cons), Kagi (when I need sourced stuff) and… TikTok, which is amazing for recipes and creative stuff.
58 comments
[ 4.0 ms ] story [ 109 ms ] threadhttps://mastoreader.io/?url=https%3A%2F%2Fmastodon.lawprofs....
or
https://unroller.zachmanson.com/tree/https://mastodon.lawpro...
>To use the Mastodon web application, please enable JavaScript. Alternatively, try one of the native apps for Mastodon for your platform.
That's very grim indeed.
> Something exceptionally grim is happening on the Internet.
> In the last few months, the constant flood of algorithmically generated junk content has kicked into an AI-powered overdrive, and it is cutting a swath of destruction as it overwhelms search engines, filters, and moderation systems
> Call it Gresham's Law 2.0: bad content drives out good.
> I'm starting this thread to document it, because there is a lot happening all at once.
> #greshamslaw20
When human-created content competes and loses against AI-generated spam, the system is ill-designed.
This is the culmination of algorithms that have rewarded engagement over everything else. Creator platforms look at content made by experts that are high quality and take time to make and effort to consume, and judge it worse than trashy content full of controversial drama pushed every week without research or effort based on arbitrary metrics. Now we've reached the point where the creator is unnecessary and all that matters is engagement.
Creator-less engagement will be the future of the web. A technology that triumphed by connecting people around the world will die connecting you with nobody but an AI that acts as wall and filter between you and the rest of the world.
Nothing that wasn't happening already since some 15 years or so.
I think what people mean by "the web" is really just the 4-5 large social media sites. Those have been utterly filled with spam/scam/fraud for many years, so I fail to see a difference.
No, as the linked thread shows, this is happening everywhere, not just the 4-5 large social media silos. There is another completely separate thread knocking around on HN somewhere about how AI spam is filling up OpenAI's chatbot store, and it's been well documented as a problem with Google image search, Quora, Linkedin and numerous other sites.
It's apparent that you yourself don't notice the problem but that doesn't mean the problem doesn't exist.
Well that's true, but likewise, just because people notice the problem more in their own lives, doesn't mean it's as widespread as some say.
(Neither of these comments are meant to say that this is not a problem - it is - I'm just tired of the crazy-overstated FUD surrounding it all)
Dunno if you follow gaming news at all, but AI spam has become the rule if you look for anything on new popular games
If you want to know how to solve a puzzle or compare two characters or something, it is almost guaranteed that you will come across pages of search results that seem like generated content
They are usually wrong, too
>If you want to know how to solve a puzzle or compare two characters or something, it is almost guaranteed that you will come across pages of search results that seem like generated content
This kind of content was never genuine. It was always low effort trash made to cheat search engines. It was either outsourced to copywriters regurgitating the same content over and over, or automatically generated when possible. Automating it with LLMs simply made it obvious.
Amazon book scams is the age-old tradition at this point, it exists for well over a decade and is a result of Amazon not willing to spend even one cent on manual curation.
OP might have a point, but it's still mostly typical microblogging ragebait with unrelated stuff mixed in.
Maybe not in your lifetime, but in mine, high-quality game walkthroughs with in-depth statistics were pretty ordinary for even fairly niche games.
At least in regards to game articles, this has been my experience for at least a decade. It wasn't AI spam, but it was spam everywhere non-the-less. I'm sure it's even worse now, but it was bad enough even 10 years ago that I've stopped bothering using google to find most game information years ago.
Believe it or not, sometimes things happen.
Nothing ever happens[0] indeed.
[0] Without costing something else coming undone. We can only pump low entropy around, we cannot generate it.
Even before the AI floodgates were opened, there was a huge problem with ads and SEO-optimized junk written by humans clogging up Google, and pushing the good stuff into the SERP netherworld. People complained for years (including here) with no discernible improvement.
It's one of the reasons HN turned into one of my go-to places for finding the good stuff - obscure blog posts and social media threads and websites that share facts, recollections, opinions, and expert insights that can't be found anywhere else. However, HN and a few other high-quality discovery sites can only go so far in surfacing such content.
Now it looks like Google, Amazon, and other Lords of the Internet are going to cause further damage with inevitable false positives as "AI spam detection" is unleashed on the Internet. Of course, there will be no recourse other than victims pleading for human intervention on a forum like this or hoping it goes viral on social media.
Information networks will increase in value over people that are knowledgeable from aggregation of third hand sources.
They'll drop the ball of course.
Then they tried to play the reputation game again by paywalling articles, but using the dumbest mechanism of free-article-views that basically just continues to incentive seo spam from their own writers. It remains unclear to me why they haven’t adopted a model like LWN — paywall freshness instead of content altogether, and rebuild their reputation and relevance
Kagi already does that (minus the automated part) and it’s one of their best features imho.
Here is my current search result blocklist, these sites earned a position on the list by either being intrinsically unwanted, (i.e. news sites, social media), completely useless (learn.microsoft.com, a site specifically written so that you can't learn anything from it (Oh, I define a few of the function for a specific thing but can't be searched and only rarely show any examples, hope you have a dev DC to practice your poorly written powershell scripts on! of course, we never tell you that there is a book series that actually teaches you microsoft stuff because otherwise more people would understand how easy it is to operate windows), or SEO spam optimized BS that prevents me from finding what I am actually searching for.
://.yelp.com/* ://.cambridge.org/* ://.linkedin.com/* ://.williampitt.com/* ://.sothebysrealty.com/* ://.whitepages.com/* ://learn.microsoft.com/ ://.microsoft.com/* ://.reddit.com/* ://.glassdoor.com/* ://.facebook.com/* ://.pcmacstore.com/* ://.yellowpages.com/* ://.superpages.com/* ://.mowerrepairshops.com/* ://.angi.com/* ://.bbb.org/* ://.freeconvert.com/* ://.pdfcandy.com/* ://.cleverpdf.com/* ://.convertio.co/* ://.png2pdf.com/* ://.pngpdf.com/* ://.xodo.com/* ://.cloudconvert.com/* ://.usinsuranceonline.com/* ://.elsevier.com/* ://.clarivate.com/* ://.google.com/* ://.usask.ca/* ://.jpg2pdf.com/* ://.smallpdf.com/* ://.pdf24.org/* ://.ilovepdf.com/* ://.11zon.com/* ://.pdfresizer.com/* ://.freepdfconvert.com/* ://.sodapdf.com/* ://.pdfcrowd.com/* ://.pdfsimpli.com/* ://.axios.com/* ://.insider.com/* ://.bhg.com/* ://.windows101tricks.com/* ://.stellarinfo.com/* ://.nucleustechnologies.com/* ://.wondershare.com/* ://.getmailbird.com/* ://.clean.email/* ://.minitool.com/* ://.cch.com/* ://.product-reviews.net/* ://.gbatemp.net/* ://.macbookproslow.com/* ://.iu.edu/* ://.hubspot.com/* ://.getdroidtips.com/* ://.shotkit.com/* ://.citrix.com/* ://.gosunoob.com/* ://.belenjesuit.org/* ://.computerinfobits.com/* ://.appuals.com/* ://.drivereasy.com/* ://.techradar.com/* ://.betebt.com/* ://.webstick.blog/* ://.anysoftwaretools.com/* ://.contactmonkey.com/* ://.regainsoftware.com/* ://.gottabemobile.com/* ://.easeus.com/* :/&#x...
The AI bullshit singularity
https://news.ycombinator.com/item?id=39422528
The internet feels fake now. It's all just staged videos and marketing
https://news.ycombinator.com/item?id=39551035
X becoming a 'ghost town' of bots as AI-generated spam content floods internet
https://news.ycombinator.com/item?id=39535310
The Internet Is Full of AI Dogshit
https://news.ycombinator.com/item?id=38952526
I at least have sympathy for the folks at: https://housefresh.com/david-vs-digital-goliaths/
However, mostly Google's included comment is personally hilarious. The weird craziness some people at Google must have to deal with.
> Lily Ray, 'when we told our client, they said "that's literally not going to be possible"'
> Google, 'Genuinely curious, what's impossible in guidance like "discuss benefits and drawbacks based on your own research"'
Surprised to "^F FTC" the article and comments, and not find any mention of the Federal Trade Commission. Given some of their prior rulemaking and caselaw around fake reviews and undisclosed compensation for endorsements, it seems the sort of thing that could be easily be prosecuted under false advertising laws. Unfortunately, I'm inclined to believe that the decision to let dying legacy media companies get away with fraud is probably a political one.
[0]: https://news.ycombinator.com/item?id=39433451
> It’s only been a year. As SORA shows, state of the art images and videos will soon be undetectable by human eye and it’s very soon coming to proliferate every “content” you interact with.
Even if the "AI" generated images will be "undetectable by human eye," they'll still be at least "completely unremarkable and uninteresting" and more likely crappy and stupid (even if all the people have the right number of fingers now).
I predict the floodgate of that "AI" generated stuff will reduce the production of actually good and creative art and images and make what is produced harder to find.
This is just the first attempt and you're criticising details. Last time they worked them out pretty quickly.
It would just be ugly, shitty graffiti, everywhere.
It'll be hard to identify a spammer who posts like a normal person but occasionally leaves a comment about enjoying a product.
It is just happening so much faster than expected.
For a little while there were studies about AI training data being poisoned by the AI generated content online. AI digesting AI.
It was expected to happen 'in the future' not 'the day after tomorrow'.
That entire website should be banned from Google.
As far as I can tell, GPT-3.5 and GPT-4 are extremely aggressive about making up features of widely used software that don’t actually exist. I lost count of how many tables and columns in MySQL’s INFORMATION_SCHEMA they made up for me yesterday, often complete with elaborate SQL commands to query them.
The next generation of models will train on this crap.
The internet is going to end up like a fuzzy mpg video that keeps getting compressed every time it gets reposted somewhere new.
maybe that will change the shape of internet and lead to premium internet service where experts are paid to provide currated informations.
does not sounds too bad tbh.
I cannot express the amount of shit I took for stating, un-movingly, that heat pumps work below -5C. I was shown Google search results, webpages, and was hit with professional experience from my brother, the red seal gas-fitter. I don't expect a "That's too bad" - I already know I'm right, and the US DoE agrees with me [0].
I think your comment is wrong. People don't search to find correct information, they Google to confirm pre-concieved biases and "verify" what they're already certain of. And I think computer-generated websites and answer boxes will only exacerbate the issue.
[0] https://www.energy.gov/energysaver/air-source-heat-pumps: "...air-source heat pump technology has advanced so that it now offers a legitimate space heating alternative in colder regions"
The other thing I've found with heat pumps is that, even when they are "heating" the air, it feels cold. The problem is they're efficiently heating the air only by a few degrees and blowing it around, but when slightly warmer-than-ambient air blows in your face, it feels cold. Gas-fired furnaces don't do this: they heat the air much, much more than ambient temperature, so it actually feels quite warm when blown in your face. (The problem here is frequently poor placement of vents, but this is common with homes and also offices.)
I'm doubtful it was simply mis-communication. Another point of contention was that heat pumps can be over 100% efficient - to the point that asking "Will they be more than 100% efficient with outside temps at 15C or 25C, and in indoor target of 20C?" got an emphatic no. That's less a miscommunication than a poor education, IMO.
[0] https://en.wikipedia.org/wiki/Air_source_heat_pump#In_cold_c...
I’ve completely moved my searches to ChatGPT (general questions), Reddit (consumer opinion, pros/cons), Kagi (when I need sourced stuff) and… TikTok, which is amazing for recipes and creative stuff.
If it was written by AI or not doesn't matter. As long as I cant check the sources it's all the same to me.