Tangentially related, I have a hunch, but cannot prove, that prediction markets are the driving force behind a lot of the bad information online, since they essentially monetarily incentivize making people misjudge the state of the world.
There's been a huge uptick in this sort of brigade like behavior around current events. First noted it around LK99, that failed room temperature semiconductor in 2023, but it just keeps happening.
Used to be we only saw it around elections and crypto pump and dumps, now it's cropping up in the weirdest places.
How do you misjudge the world based on web articles? If you don't have proper foundations where to source your information from you are already doomed.
Signal to noise ratio is getting *lower (EDIT: was higher) than ever. I don't see a way out of this other than "human certified" digitally signed authorship (e.g. by using eIDAS in EU). There could be a proxy to at least retain pseudo-anonymity, but trackable to a human. Tragedy of commons strikes again.
It could be interesting to have a search engine that only shows results that have human attestation via digital passports. Of course I'd prefer that to work without necessarily revealing the identity of the poster, similarly to how anonymous "sign-up tokens" for accounts would work, to prevent sybil attacks.
This is why my friends and I are setting up a mesh network in our town.
The open internet has been going downhill for a while, but LLMs are absolutely accelerating it's demise. I was in denial for the last few years but at this point I've accepted that the internet I grew up on as a kid in the late 90s to mid 2000s is dead. I am grateful for having experienced it but the time has come to move on.
The future for people that valued what the early internet provided is local, trusted networks in my opinion. It's sad that we need to retreat into exclusionary circles but there are too many people interested in making a buck on the race to the bottom.
Reminds me of a project idea I had. You'd get a little Raspberry Pi style board with BTLE and battery power (ideally lasting for weeks at a time) and covertly stick it in some communal location, e.g. a cafe or library. Then you'd have it run some local-only forum software and disseminate instructions for connecting to it. The point would be to have a digital community accessible only by direct connection and bound to a physical location by design, kind of in the vein of Community Memory.
It's probably too impractical to work as described, but I think that having a digital space constrained by physical access would be meaningful in a way that internet communities are not. The people you chat with would necessarily be the people in your physical environment, which would make it feel more like a local hangout than the typically vapid social media exchange.
(On further reflection, it would probably be easier to make a mesh network app version of this. Hmm...)
I had a similar idea a few years ago and tried to set it up, but failed at making it easy to connect to.
I wanted the phone to prompt you like when connecting to wifi hotspots where you have to accept some T&Cs before you can connect to the internet, but to then just show you the local services instead of actually offering internet. Honestly, this can't be that difficult, but at the time, I could not get it to work reliably.
I was recently running into this while playing the latest Hollow Knight game. Several sloppified sites which obviously were trying to tailor mechanics/items of the original game into the new one. The new release is only ~six months old, so there is just not that much hard content available to reference.
My question is -why? Is it really worth the ad revenue to trick a few people looking into a few niche topics? Say you pick the top 5000 trending movies/music/games and generate fake content covering the gamut. What is the payback period?
It comes down to Google's failure. Rather than outright defeating the SEO eldridge abomination by adopting a zero-tolerance policy to those tactics, Google made a mutually advantageous bargain with them of - course, leaving out a third party: us. They could do this because they had no competition. Now, the culture of enabling bad actors is, unfortunately, set.
Google did all the innovation it needed to and ever is going to. It needed to be broken up a decade ago. We can still do it now. Though I don't know how much it will save, especially if we don't also go after Apple, and Meta, and Microsoft.
I thought a lot last night about how we could protect HN, I didn't come up with a good answer except maybe you'll need to have someone with a higher reputation vouch aka invites. My internet community journey has mostly just been irc -> dA -> twitter -> HN. Too frequently these days I feel I might be putting emotional energy into something that isn't human on this site, hard to express how that makes me feel, but it's not pleasant at all. 힝
You never could trust the internet. The difference is that now the problem is so widespread that it's finally spurring us into action, and hopefully a good "web of trust" or similar solution will emerge.
It is true that as the cost to construct fake content has gone to zero, we need some kind of scalable trust mechanism to access all this information. I don't yet know what this is but a Web of Trust structure always seems appealing. A lot of people are going to be excluded, but such is life, I suppose.
If I were to be honest, going to where the fish aren't is also going to help. Almost certainly there are very few LLM generated websites on the Gemini protocol.
I'm setting up a secondary archiver myself that will record simply the parts of the web that consent to it via robots.txt. Let's see how far I get.
The future of the internet is going to be invite-only enclaves. I sometimes wonder is anyone working on the next generation of discussion forums, or if it'll be a return to PHPBB.
They'll instantly become infiltrated with bots and include people based on arbitrary politics. Either the content is such that it makes zero sense to game or spam or it is lost already.
This will create an interesting problem for newcomers to an area or hobby, since they will have few introduction points. It will work for some people, but exclude others.
AI is kind of like Skynet in the first Terminator movie. It now destroys our digital life. New autogenerated websites appear, handled by AI. Real websites become increasingly less likely to show up on people's daily info-feed. It is very strange compared to the 1990s; I feel we lost something here.
> The commons of the internet are probably already lost
That depends. If people don't push back against AI then yes. Skynet would have won without the rebel forces. And the rebels are there - just lurking. It needs a critical threshold of anger before they will push back against the AI-Skynet 3.0 slop.
I had a similar experience when I was looking for YouTube videos on the Intel i7-4790T, it's a relatively obscure CPU that was only found on small-form factor pre-builds during the Haswell era. The only recent videos I found were slop videos [1] narrating a script clearly generated by an LLM, with a link to their Amazon affiliate in the description. The CPU has never been put on retail sale! These channels upload a dozen times a day on random products just to get an affiliate commission.
There's a lot of people unhappy about this here. Presuming that the sentiment extends beyond HN, then it might be a problem that you could make some money by solving. (In the same way that Google figured out how to let the net tell it which pages were best, and made an insane amount of money from doing so.)
People want something real, not AI slop or shills or astroturf or corpo-speak or any of a thousand other flavors of fake. People want it rather desperately. In fact, the current situation is bad for peoples' mental health. Can someone figure out how to give people a much higher percentage of real?
I've been hitting this a lot lately in Kagi. I'll search for instructions on how to do a thing and some random website will have nothing but _hard_ AI slop going off about the thing I was looking up.
It must be easier than ever to build content mills these days.
And why not? We humans do things like this all the time. We act with powerful false beliefs. Misunderstand a situation or simply just the meaning of a word, and then build our world-view and lives around those false beliefs. Train your model on this, and replicate in those false beliefs.
39 comments
[ 2.1 ms ] story [ 71.3 ms ] threadPreviously you might get burned with some bad information or incorrect data or get taken in by a clever hoax once in a while.
Now you get overwhelmed by regurgitation, which itself gets fed back into the machine.
The ratio of people to bots reading is crashed to near zero.
We have burned the web.
Enshittification strikes again.
And it doesn't have appear to have any means to rid itself of the bad apples. A sad situation all around.
We must live on different planets.
There's been a huge uptick in this sort of brigade like behavior around current events. First noted it around LK99, that failed room temperature semiconductor in 2023, but it just keeps happening.
Used to be we only saw it around elections and crypto pump and dumps, now it's cropping up in the weirdest places.
Superconductor
1. Don't believe everything or anything you read or see on the Internet.
2. Never share personal information about yourself online.
3. Every man was a man, every woman was a man and every teenager is an FBI agent.
I have yet to find a problem with the Internet thats isn't because of breaking one of the above rules.
My point being you couldn't ever trust the Internet before anyways.
The open internet has been going downhill for a while, but LLMs are absolutely accelerating it's demise. I was in denial for the last few years but at this point I've accepted that the internet I grew up on as a kid in the late 90s to mid 2000s is dead. I am grateful for having experienced it but the time has come to move on.
The future for people that valued what the early internet provided is local, trusted networks in my opinion. It's sad that we need to retreat into exclusionary circles but there are too many people interested in making a buck on the race to the bottom.
It's probably too impractical to work as described, but I think that having a digital space constrained by physical access would be meaningful in a way that internet communities are not. The people you chat with would necessarily be the people in your physical environment, which would make it feel more like a local hangout than the typically vapid social media exchange.
(On further reflection, it would probably be easier to make a mesh network app version of this. Hmm...)
On the internet no one knows if you're a dog, human or a moltbot.
My question is -why? Is it really worth the ad revenue to trick a few people looking into a few niche topics? Say you pick the top 5000 trending movies/music/games and generate fake content covering the gamut. What is the payback period?
Google did all the innovation it needed to and ever is going to. It needed to be broken up a decade ago. We can still do it now. Though I don't know how much it will save, especially if we don't also go after Apple, and Meta, and Microsoft.
If I were to be honest, going to where the fish aren't is also going to help. Almost certainly there are very few LLM generated websites on the Gemini protocol.
I'm setting up a secondary archiver myself that will record simply the parts of the web that consent to it via robots.txt. Let's see how far I get.
I wonder if there is another way.
> The commons of the internet are probably already lost
That depends. If people don't push back against AI then yes. Skynet would have won without the rebel forces. And the rebels are there - just lurking. It needs a critical threshold of anger before they will push back against the AI-Skynet 3.0 slop.
And at that point does it even matter? Zuckerberg wins.
[1] https://www.youtube.com/watch?v=YpHUBC681iU https://www.youtube.com/watch?v=0w5a33Jeen0
But it's the date at which it is no longer possible to discern reality you can't actually observe.
People want something real, not AI slop or shills or astroturf or corpo-speak or any of a thousand other flavors of fake. People want it rather desperately. In fact, the current situation is bad for peoples' mental health. Can someone figure out how to give people a much higher percentage of real?
It must be easier than ever to build content mills these days.