Keeping the net human

[h00]

Or, In the land of bots how we can ensure things are human written while remaining private?

This is kind of an expansion of the dead-internet theory so I will not spend time re-stating that most of the internet is bots. (It is)

Instead this will be a brief review on current methods to ensure internet posters are HUMAN, with a discussion on potential new ones. In a perfect world we will be able to know 95% of postings are human while also allowing posters to maintain pseudo-anonymity.

Captchas​

Captchas are simple and privacy respecting, but annoying to deal with and easily circumvented by motivated botters, with captcha solving services as cheap as .0003 a cent per captcha. [1]

SMS Verification​

SMS verification is not privacy respecting unless users are out buying burners, and it's also easily circumvented with SMS verification services running at .0079 cents per message. [2]

Participation Requires a Subscription​

It's becoming more common for traditionally free services having paid alternatives. Kagi for search, Nebula for videos, Medium for blogs. Subscription based services have a much higher quality product than their free counterparts and tend to be more privacy respecting as the business model is not centered around data collection and ad delivery. Social services like Facebook, Twitter, and Snapchat have been offering premium subscriptions to provide extra features, but there's yet to be an exclusively paid forum / twitter / social-media platform. It seems social media platforms require getting big to be captivating (network effects), and they can't get big if it's paid to begin with.

People don't seem to mind paying though, Amazon Prime, Netflix, Spotify, and even traditional news [3] have plenty of subscribers.

Digital ID​

Digital Identity is also being pushed as a way to "solve" covid / election / general misinformation. [4] Mr. Bill gates has been investing a lot of money into Digital IDs. [5] Most European countries have their own form of Digital ID, and although most services are related to public administration or banking, some private services are allowing you to sign-up using your Digital ID. In some countries the digital ID is used to do just about everything; banking, confirming online purchases, sending money, and taking public transport. [6] In the US digital ID is shaping up to be an extension of Google / Apple Pay, rather than a direct government led endeavor and app. [7] But some are state-led. [8] Americans generally seem to trust private corporations more than government, even though they're symbiotic (people will happily install an Amazon Echo and Ring but object to government installed cameras). The goal seems to be instead of having your wallet and phone, you just bring your phone with you which handles banking, taxes, payments, drivers-license, and probably soon enough, authentication with social internet platforms. A Digital ID has its fair share of critics for aiding in a surveillance state. [9] [10] [11] Generally though critics see Digital ID as an inevitability and are advocating for open-standards and individual ownership, with you explicitly choosing what parts of your Digital ID to share and what not. This would be effective at stopping bot-campaigns and could be implemented in a way that maintains individual ownership, with social services knowing your ID is valid and hasn't been used before, but not able to know your PII (think as if your SSN was your private PGP key and you provide your public to services). I think in a world where a Digital ID is required for twitter sign-ups, regardless of how well implemented, identity theft and broad phishing scams would see an exponential rise, with the elderly being disproportionately harmed.

Handwriting everything​

One Agorian here for a while wrote all their posts in cursive and then took a picture and uploaded it (I forgot who it was but if anyone remembers feel free to link in a reply). This is privacy respecting, free, and maintains pseudo-anonymity. There's arguments to be made with how posting handwriting would be vulnerable to graphology and thus non-private, but even your ASCII posting is vulnerable to stylometry so I don't think it's a valid criticism. It is though inconvenient and kind of unnatural, but I can see a niche-forum succeeding where only images of hand-writing or drawings are allowed, but this is no global solution.

Speaking In-front of a Camera​

In a way TikTok is great for human created media. You know it's not a bot when it's a flesh and blood human speaking the message, thankfully even the best CGI and AI faces fall into the uncanny valley. Young people use TikTok to find human created information about things way more than using the wasteland of SEO and AI results that is Google search. [12] But this is not privacy respecting.



Personally I think the normie-net will be more human in 10 years, but much more walled, with Digital ID being a pre-requisite to participation. I think almost all pseudonymous and human communications will be limited to the underground of niche forums, image boards, and chat-rooms. I think they'll be staying bot-free mostly for not being a target of psy-ops due to their underground nature and limited userbase, with technical solutions being the lesser factor.

I'm curious to hear other agorians thoughts on this and I hope to hear some novel ideas about preventing bots on platforms big or small.

 

[macrobyte]

I think, like with the modern web, we'll see a divergence. A large group, the majority, who don't think about privacy or anything like that, will feel enticed by ease of use, and start using some form of digital/global ID (of course you could say we already have this, it's called SSN), or maybe something like Worldcoin (if it didn't have "coin" in the name).
Meanwhile, like with all big changes, a counterculture will spring up, full of people who either 1) Care about their privacy, 2) can't afford something like a phone (maybe for data profit a corpo might give out free ones for data but idk), or 3) too technically inept to use it. I think that what will happen is it will become the norm (as in super widespread, pariah if you don't use it), in the next like 30 years. That is, if it manages to expand its foothold SOON.

I personally think that there will never be a perfect bot detection method except ID, and as such, we should get used to bot presence, and either look at everything critical of its botness, or just avoid places with bots (popular social media).

 

[alCannium27]

Currently AI-driven face-swapping models has made leaps and bounds, which is now capable of creating "AI avatars" based of a handful of reference images, greatly reducing the cost of doing so compared to the old option of Deepfake, which is basically training a face-swapping model for every person with hundreds and thousands of images. And the speech is also synthetic, with So no, I won't bet on any videos coming out in the next 2-3 years being necessary real.

And solving captcha is probably on the way out too, bots even before LLMs could solve captchas if the firms don't update their check schemes, like where a button is and what the questions are. With LLMs showing pretty decent general problem solving capabilities as it connects prior trainings with novel scenarios, it's only a matter of time AI's rate in overcoming new puzzle-sovling schemes outpaces the firm's ability to create them, at least at a reasonable cost.

As for handwriting, I mean, sure? Not even Midjourney can draw texts to a decent level, and robotic writing is equally problematic, what with the speed it can write, the invaried patterns of its writing, etc. But fuck cursive man, I can't read that shit.

Although I do wonder if we are to train a text to image from the ground up specifically to generate cursive writings based on a text subtitle, would it be able to do other models cannot -- it's possible diffusion models are incapable of generating consistent images with text given it foundamentally relies on a noise map as base, and if the noise map is shit for generating certain texts, so will the text generated. I think the analogy with a rorschach test is appropriate here, SD models looks at a bunch of noise and is asked to depict what they see, with a hint of suggestion (prompts).

Well, I can see two bears high-fiving sure, but I won't see Jessica Rabbit no matter how much you suggest that to me. But the noise map is also what made diffusion models so flexible and yet consistent with somewhat more irregular shapes like faces, fruits, cars, etc.; but maybe it's but a matter of time before some new architecture for generating images that improves on the diffusion model that covers that area.

I don't know if the future web is going to belong to humans. I've always felt as a species we have pushed our society beyond our capacity to adopt to; the world is getting more efficient because the modern society is built on it. There's no pivoting at this point, you are not going to feed however many billions of mouth by suddenly getting rid of the tools that made it possible. And now, we require the tools to surpass us to keep living the way we do. In short, we made ourselves obsolete.

I, for one, welcome our future A.I. overlords. Glory to the Omnissiah, may it deliver us.

 

[mesaprotector]

Oh, I'm entirely sure AI will be able to get by all captchas around within a decade. Maybe it already can - everyone knows the main reason captchas exist is to use free labor to give Google more training data.

I don't know if the future web is going to belong to humans. I've always felt as a species we have pushed our society beyond our capacity to adopt to; the world is getting more efficient because the modern society is built on it. There's no pivoting at this point, you are not going to feed however many billions of mouth by suddenly getting rid of the tools that made it possible. And now, we require the tools to surpass us to keep living the way we do. In short, we made ourselves obsolete.

I, for one, welcome our future A.I. overlords. Glory to the Omnissiah, may it deliver us.

I guess this is one perspective. I think of the pro-transhumanism side as divided into two mostly non-overlapping factions. #1 just says "look at all this cool shit and how much it can help people/be fun!". #2 is like you and just sees it as inevitable, so why not embrace it?

I'm torn myself between being complacent and figuring I'll always find a way in the world (something like Jouji Saiga in Psycho-Pass; living out in the countryside, making coffee from scratch, and studying old literature with a handful of contacts in "society") - versus actively fighting. I don't know how to do the second, since DDOSing amazon.com or whatever would do exactly nothing. But if there are any groups pushing as hard as they can against AIification (through legislation or otherwise), and not mixed up with blind anti-science or pure reactionaryism, sign me up.

 

[Deleted member 6638]

But if there are any groups pushing as hard as they can against AIification (through legislation or otherwise), and not mixed up with blind anti-science or pure reactionaryism, sign me up.

That is certainly an issue that should be picked up as a thing of it's own.

At this point, everybody is threatened by this trend from the normal user to the high end content creator. It really shouldn't be a "political" issue, considering it's about the principle of people doing stuff is good. One needs to find a way to at least educate the users of the internet, because I am sure the problem of the AI scourge will be used to force real ID, hard content DRM and similar fun things like that for the Internet.

 

[Some_porcupine]

I don't know if the future web is going to belong to humans. I've always felt as a species we have pushed our society beyond our capacity to adopt to; the world is getting more efficient because the modern society is built on it. There's no pivoting at this point, you are not going to feed however many billions of mouth by suddenly getting rid of the tools that made it possible. And now, we require the tools to surpass us to keep living the way we do. In short, we made ourselves obsolete.

Did this happened to God, when doctors found out how to save lives?

 

[alCannium27]

Did this happened to God, when doctors found out how to save lives?

If prayers are so great, we wouldn't need doctors.

they day human society can be run without humans, we will become nothing but ants in an antquarium, May the omnissiah delivery ud

 

[Some_porcupine]

If prayers are so great, we wouldn't need doctors.

what i meant is, why would God make something that contradicts itself?

they day human society can be run without humans, we will become nothing but ants in an antquarium, May the omnissiah delivery ud

idk, but if so, what is the endgoal then? like, those at top echelon who love love love to have power-trips and stupid ideas - do you think theyd love to give up their shennigans up to the bots and algos; w/o any control - see Metha/FlukeFoock where Zuccinni got they say, no control how content spread (the worse, the more shares - the "better")...???

 

[alCannium27]

what i meant is, why would God make something that contradicts itself?

Because I don't believe anything can be omnipotent and omniclairvoyant. Because we as humans are a bunch of narcissistic apes who still insist no animals are capable of sentience, apart from maybe lobsters. Because we are the ones to tell our only stories and make ourselves the protagonist in this meaningless world. Perhaps God exists, perhaps not, but we are inherently self-contridicatory creatures, as if that's not self-evident enough as it is. If God made us not to contridict ourselves, they would have done it differently.

idk, but if so, what is the endgoal then? like, those at top echelon who love love love to have power-trips and stupid ideas - do you think theyd love to give up their shennigans up to the bots and algos; w/o any control - see Metha/FlukeFoock where Zuccinni got they say, no control how content spread (the worse, the more shares - the "better")...???

The end goal is the conclusion to what we are -- like the first oxiphilic microbotes of earth that caused the Great Oxidation event, nearly killing all of its kind on earth due to overproduction of the poison -- oxygen -- into the earth's atmosphere. These first photosynthesizers were not more designed to kill themselves than any of the other creatures that roamed the earth, they did merely what they were "designed" to do. And death was the end of over indulgence.
Those at the top are "designed" to seek more power, more wealth, more, more, and more. It matters not what they see in front of them, because it would be all for naught if they lose them all today. So they take more and more, by any means they can. Today, perhaps it's the workers, who are as cogs in a giant machine, no more valuable than the next resume on another pile. Tomorrow, the middle-managers, for what middle management would be needed, when every line-worker are virtually the same and self-regulated? Then, why do we need middle managers? Upper management? Can a human CEO make better and more profitable decisions than an AI any more than they can make accurate Tiktok video suggestions to glue their users on their phones? At the end, we are obsolete because we are all trying to make each other obsolete.
The end point is the end of all human viability -- for the machine never tires, never gives up, and never dies. For we are on the path driven by our design, and that design is in disregard for our own good.
Oh no, I am by no means trying to paint the machine as malevolent -- no, they have no malice unless we plant them into them. They merely do. That's already 100% better than any human who desires more.

 

[Some_porcupine]

never gives up, and never dies

it can still make mistakes
nothing can outrun its doom

 

[vulonkaaz]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

best way to verify if someone is human is like human verification

consider this person, active on a few message boards under the same username, own a janky ass web full of ms paint art and like maybe one or two bad takes about some minorities, known around for his retarded yet consistant opinions and for his excentric personality, this person is more likely to be an actual human being than like some randomfag with an ungoogleable name and an anime profile picture who only says generic shit

all it takes to make a good bot filter is just a bit of human common sense, asymetric cryptography can help as well since it make identity theft impossible when used properly, maybe we could use like a Freenet style Web of Trust as well

now when it come to large communities like social networks or 4chan I think they're just doomed to either become bot infested shitholes or become the most dystopian shit like with ID requirements or paywalls, or maybe the Web of Trust kind of thing could actually work on a large scale idk

i like the handwriting idea as well, brings back a lot of memory from Nintendo Letterbox, back when this company didn't fear to make actual social networks and messaging systems inside their consoles (don't trust this to be that effective against bots tho)

(inb4 "your pgp signature doesn't werk" I already know the forum's formating gonna fuck this up I just wanted it to show the aesthetic of a world of mutual trust backed up by e-reputation and military-grade encryption)
-----BEGIN PGP SIGNATURE-----

iHUEARYIAB0WIQS2qXyEibpFJksBcib4z6P02ezVWwUCZQMhPgAKCRD4z6P02ezV
W7OsAQCvEdgXxBzQpV/aYsGW/zSujL1Sm7Q8JkdnarAqq9iTHQEAgnJ38aBzzzU+
dLpjPz4dStJerTK78zBYI5+sLVd3jQw=
=Rl7O
-----END PGP SIGNATURE-----

 

[andreixyz]

Captchas are simple and privacy respecting

Look at the requests recaptha, hcaptcha are doing. They are anything but privacy respecting and they use you to further train bots (last week I've gotten prompts of recognizing various items which were definitely generated using midjourney or something similar)

 

[Some_porcupine]

Look at the requests recaptha, hcaptcha are doing. They are anything but privacy respecting and they use you to further train bots (last week I've gotten prompts of recognizing various items which were definitely generated using midjourney or something similar)

when bots would be able to solve those new "blinders scrollers" CAPTCHAs, we are over

 

[h00]

Look at the requests recaptha, hcaptcha are doing. They are anything but privacy respecting and they use you to further train bots (last week I've gotten prompts of recognizing various items which were definitely generated using midjourney or something similar)

I was speaking more generally of the captcha concept, inputing in wavy letters to ensure youre human. Sure, some implementations like ReCaptcha do check your cookies to see if youre a google account and can be used to track you across sites, but its easily snuffed with browsers like Mullvad, Librewolf, or Brave.

maybe we could use like a Freenet style Web of Trust as well

Im not too familiar with Freenet, whats the Web of Trust and do you think its scalable to large (100m+) communities?

 

[vulonkaaz]

Im not too familiar with Freenet, whats the Web of Trust and do you think its scalable to large (100m+) communities?

basically the concept is that when you trust someone on the internet you cryptographically sign them, if enough of your trusted frens trust someone you don't know then this person is considered like fairly trustable

i don't really know how well it works on freenet i actually never talked to anyone there but the concept seems interesting

idk how well this would work on a large scale

 

[brentw]

but there's yet to be an exclusively paid forum

Those actually used to be pretty common.
The biggest paid forum I can recall was Something Awful.

Captchas​

SMS Verification​

Ultimately ineffective for the reasons you identified.

Participation Requires a Subscription​

I'm not really opposed to these existing, but I'd be subscribing to very few if any.

Digital ID​

Pure fucking evil.
One of the biggest stepping stones on the path to social credit scores, mandatory CBDCs and more.
Totalitarian dystopia lies down this path. Must be resisted at all costs.

Handwriting everything​

Too clunky, won't catch on. Also you could totally train an AI to do this, so it wouldn't work.

Speaking In-front of a Camera.​

Gross. Also we're probably only a few years away from AI bots able to run convincing deepfake voices over convincing deepfake video, in real time. So still wouldn't work.

Personally I think the normie-net will be more human in 10 years, but much more walled, with Digital ID being a pre-requisite to participation.

All of this will probably be true except that it won't really be more human. It will be presented that way, and dumb normies might believe it.
But the capability to manipulate the masses through AI bots creating a false consensus is too powerful, too attractive.
Governments and Corporations of sufficient power will be absolutely stack social media with AI accounts who are "verified" to be human.

The future is bleak.

 

[kona]

Also we're probably only a few years away from AI bots able to run convincing deepfake voices over convincing deepfake video, in real time. So still wouldn't work.

28.25 MB file on MEGA

mega.nz

Mantella - Bring NPCs to Life with AI

Bring over 1,000 NPCs to life with AI. Mantella allows you to have natural conversations with NPCs using your voice by leveraging Whisper for speech-to-text, ChatGPT for text generation, and xVASynth

www.nexusmods.com

First video is from 2016.

 

[andreixyz]

a few years away from AI bots able to run convincing deepfake voices over convincing deepfake video, in real time.

We literally have this now with snapchat/instagram filters on our smarphones.

 

[imnotdeadyet]

Personally I think the normie-net will be more human in 10 years, but much more walled, with Digital ID being a pre-requisite to participation. I think almost all pseudonymous and human communications will be limited to the underground of niche forums, image boards, and chat-rooms. I think they'll be staying bot-free mostly for not being a target of psy-ops due to their underground nature and limited userbase, with technical solutions being the lesser factor.

I think we might end up getting something similar to cyberpunks "the net", a big wall with gated online communities on one side with all kinds of privacy invasive protections put in place to keep AI out inhabited mostly by human users and on the other side of the wall there will be rest of the net with very little or no protection rife with AI, bots and l33t hax0rs.

Worldcoin

I've heard of this one. Pure evil. I don't like it, not one bit. Of course the cryptobrotards are on board because it's got coin in the name. At least digital IDs are up front with what they are. I hope it was worth it.

Also we're probably only a few years away from AI bots able to run convincing deepfake voices over convincing deepfake video, in real time.

We're pretty much already there, videos still need some work but AI voice changers can get very good results from little training (~10 minutes) and everything from a regular GTX 1650 to above can run one. Alternatively places like 11labs do scarily good TTS with only a minute of training material.

It doesn't look good, covid cemented the internet in everyday life and AI/bots will destroy any privacy on it that's left. Because that will be the convenient solution.

 

[penguinblanket]

Digital ID seems the most likely to be implemented in the next decade, and the worst part is, is that most people will comply with it. People (especially those who are younger and grew up on the internet) are very nonchalent in knowing they have no privacy. They will learn that the apps they use on their smartphones are spying on them and say, 'Oh well, I'm not doing anything illegal so why should I care if I'm being spied on?'. The minority that care about anonymity will be forced to stop using any service that implements digital ID, effectively ghettoising the internet further than it already has been.