so you may have heard of this Dark Forest hypothesis the idea is that basically the reason we haven’t found any aliens out there in outer space no matter how hard we look or how many messages we send out there is because they’re staying quiet they’re staying hidden maybe because any life form out there in space that makes too much noise they get wiped out they describe space as a dark Forest filled with arm Hunters stalking through the trees like ghosts so this post has been making its rounds online and more and more people including people on YouTube have covered it it’s called The expanding Dark Forest and generative AI proving you’re a human on a web flooded with generative AI content and this post and the surrounding conversation about it has a few excellent points it also makes a few points that are kind of nonsense or perhaps outdated at this point this was planted as they say about one year ago and I think it’s missing a few key details so Maggie Appleton who wrote it I think did a phenomenal job and really opened up my mind to a few of these things the conversation around it but there are a few points that need to be dark forest internet added so what is the Dark Forest of the internet the Dark Forest theory of the web points to the increasingly lifelike but lifeless state of being online most of the open internet is overrun with Bots advertisers trolls data scrapers clickbait keyword stuffing content creators and algorithmically manipulated junk what’s even more annoying some of the stuff is flowing into other areas of our life the other day I was trying to open up my garage and I had an app for it and when I open up the app I got hit with a promotional message like I had to interact with it and exit out of it before I could open up my garage to get into my house and all this is only going to get worse and continues that all the living creatures that’s that’s us that’s me and you are hidden beneath the ground or up trees if they reveal themselves they risk being attacked by automated Predators so you might have seen more and more people people talking about Discord channels slack channels Etc where it has to be invite only where it has to be a little bit more exclusive wild off from like the rest of the internet and here’s kind of how they sort of Envision that so here’s you know at the top is the Dark Forest of the clear web so this is Facebook LinkedIn Twitter Etc one layer below you have the digital Gardens email newsletters RSS feeds so these are kind of the curated things this is where somebody’s going through and if if you choose to hear from them they give you kind of like that content that you’re looking for and this tends to be unpolished Loosely organized notes published on personal wikis tended over time as an evolving knowledge base then we have the Cozy web gatekeeper protected Enclave communities run through chat streams like WhatsApp slack Discord Snapchat Etc so more and more I’m seeing this these small cozy web communities as they referred to and then below this whole thing you have like the dark web onion routers so basic basically I mean use specific Tech to get onto the dark web but basically that has its own ecosystem and its own set of problems let’s say and so what they’re saying is that the Dark Forest that kind of clear in it it’s about to expand large language models that can constantly generate content well they just join the party they can pump out advertising copy blog posts emails social media updates and marketing pages and they’re really good at it so this is saying a few weeks ago opening I released Chad GPT the next version of GPT 4 so keep in mind that this is a a little bit outdated and that’s why I think there’s a few additional points that that we have to make as an addendum to this because things have changed since this was written and of course they mention other things that can generate images by now you’ve seen Sora I mean you get it there’s more and more stuff out there that can produce content really good content you thought the first page of Google was bunk before you haven’t seen what happens when the SEO Bros pump out content of Chad GPT marketers influenc and growth hackers will set up opening I to zappier pipelines that auto publish a Relentless stream of LinkedIn motivation Monday posts it goes beyond text two video essays on YouTube Tik Tok Clips podcast slide decks Etc so what’s funny is I see more and more people on YouTube getting accused of just using AI voices to create content now i’ I’ve tried a lot of AI voices a lot of video generators and stuff like that I’m going to say we’re not quite here yet and I’ve seen certain YouTubers try to generate AI voices but we’re still not quite to a point where it’s indistinguishable like if you know what to look for you know what’s real what’s not but I mean that’s going to rapidly change like for example this guy uses AI generated video I’m pretty sure he records his voice though but then he just has it sort of animated with an AI generated Avatar if you look at enough of the videos you can tell I think what happened was his microphone was in slightly different positions through several of the videos that he used to train the AI model on and so take a look what happens as he’s talking you see how it’s like just rocking back and forth and kind of just shifting from place to place I doubt there’s like some earthquake or something like that going on rocking it like that it’s just all the different videos that were stitched together that were used to train the models probably in slightly different positions so you see it kind of rocking back and forth you can also see this ra go in and out of focus take a look at what’s happening here it goes in and out of focus kind of like even though the rest of the camera doesn’t go in and out of focus and I can give you other examples but the point is I don’t think we’ve quite reached it there yet the voices that are out there they’re not quite there yet the inflections and the pauses and stuff like that are very very difficult to get right so if you have like a short thing that you need to to say like a little oneliner that is going to sound great if you want them to do a long form video like a podcast it’s not quite there yet like we’re we just don’t have that at a level working fake people but soon we are going to have people that are completely AI generated but we’ll go over a few ways to prove that you’re online at the end of this post you’re going to hate all of them so they’re saying we’re about to drown in a sea of pedestrian takes an explosion of noise that will drown out any signal I I feel like we’ve been there already before Chad GPT I think a lot of the optimization of the internet killed a lot of that there was a brief moment of time where search results were great podcasts were great because it was just real people talking about stuff they were interested in and then the money flowed in the optimization the corporations like it really went down quite a bit many people say we already live in this reality oh that was literally her next Point yes this is I’m one of those people and forchan apparently proposed dead internet Theory years ago that most of the internet was taken over by artificial intelligence passing the reverse touring test so what is the reverse touring test though the touring test is this idea of like how at what point will artificial intelligence be sort of smart and advanced enough to trick us into thinking it’s human so if you hear a conversation between two people and they’re talking discussing various ideas and at the end you have to guess which one is the robot and which one is human at what point will you no longer be able to guess after the forest expands we will become deeply skeptical of one another’s realness every time you find a new favorite blog or Twitter account or Tik Tok personality online is this really a whole human of a rich and complex life like mine next they say before you continue pause and consider this how would you prove you’re not a language model generated predictive text what special human tricks can you do that a language model cannot now on this Channel we’ve covered examples where people would kind of quote unquote jailbreak these models to solve captures the more advanced models do really good at those little intelligence tests where you kind of have to follow a certain pattern and guess what comes next there was this camping ground where they had a few incidences where the Bears would be able to get in there and kind of get at the trash that people would throw out they would be able to open whatever security systems that the campground set up to you know keep the Bears out and they were facing some criticism along the lines of how hard is it to create a security device that a bear can’t you know navigate around cuz they weren’t breaking it they were opening they they were figuring out how to get inside right so they were saying to these people like you should be smart enough to build a security device that can you know keep a bear out and the guy that was in charge of creating those security templates he said well the problem isn’t keeping the bear out he said the problem is that the smartest of bears this is sort of the intelligence distribution of bears right so this kind of your average bear and so over here you kind of have your you know the smarter bears and here you sort of have your intelligence distribution of you know the people that visit the park the campers there’s this unfortunate overlap to where the smartest of Bears and the not so smart camper where it overlaps so building a device that keeps out all bears will keep out a certain percentage of humans and it’s kind of the same problem here right can you build a capture that currently will keep out all of the AI models maybe but what percent of human beings using the net won’t be able to you know solve those captures so here in this blog post they give some ideas of how to make sure that we can kind of find the humans online right Tri triangulate objective reality right they’re saying well we can have Vision sound taste we’re part of cultures communities and these AIS they’re just a language model in a box if a some event or person isn’t part of their training data they can’t tell you about it they don’t know about the events that happen after a cut off date number two is one of the best ways to prove you’re not a predictive language model is to demonstrate critical and sophisticated thinking or we can develop creative language quirks dialects memes and jargon and they’re saying no language model will be able to keep up with the pace of weird internet lingo and memes so everything there I’m pretty sure we can safely say it’s outdated since that was written we had massive advances in Chad gbt Gemini 1.5 Pros demonstrating incredible in context learning abilities learning to speak a whole new language and again this is this is post trainining right so after it was trained we gave it some documents about this language that has fewer than 200 speakers worldwide and it learned to speak a brand new language just cuz it you know read a book in in that language and other people are demonstrating similar or the same abilities with Claude 3 so it’s important to understand that a lot of the stuff that people are saying well we can just do this that assumes no progress that these language models won’t get better no language model will be able to keep up with the pace of the weird internet lingo and me yes maybe in you know 2020 or 2021 whenever this was written but it’s 2 years later now it could do so much more we have to assume that they’re going to get better two years is a long time and AI time step four is consider institutional verification so we need verification by centralized institutions so like a DMV but for posting on Twitter you then get some kind of a special badge or Mark online legitimizing you are a real human it may or may not be on the blockchain somehow and they’re saying this could have problems susceptible to abuse and ultimately impractical and the other kind of idea here is show up in Meats space which is a great way of putting it basically you know enjoy the great outdoors go and meet your friends in real life grab a cup of coffee so I enjoyed reading this essay I think the first half of it is spoton I agree with pretty much everything the internet has been getting worse and worse I think objectively if if you’re a user the experience has been getting worse and worse there’s more popups and cookies and those stupid cookie messages and tracking and just advertisement is shoved everywhere now I think this article does Miss a few key items so most places find it near impossible to have an open web and not have Bots there are almost no ways to defend against that no matter how thorough you are in defending against them there’s one way though that is really effective and that is collecting payment even if it’s just a little bit it doesn’t have to be a lot of money which the fact that you have to pay with a credit card makes it easier to to eliminate Bad actors it could be a dollar it could be five bucks few people have access to an unlimited number of credit cards even if you do you know they’re able to match your name and all the other stuff it’s still difficult to continuously get them and pay for stuff you can lose a 100 automatically generate profiles and not really have a problem with it but if a hung of your credit cards get blacklisted I mean it’ll be difficult to kind of recover that and certainly we’ve seeing that with Twitter they’re moving more and more towards getting paid to you know you have to pay to be a verified member on there and get your blue check mark If you will so certainly I think this would increase the amount of content and communities that people would pay for to be a part of but that still doesn’t prevent something like a Cil attack this is where an attacker creates multiple fake identities known as Cil nodes to gain control of a network of a system so it’s kind of like let’s say you you’re trying to vote for something and you create multiple identities or if there’s like a raffle of some sort creating multiple raffle tickets right you’re sort of changing how much power your vote has or what the chance of your winning ticket being selected and there are instances where you might not need a million Bots you could just go and create you know 10 different accounts verify that they’re a real human right let’s say there’s five different networks you pay a dollar to join each you show them your ID so they know you’re a real person but you still have 10 identities 10 nodes to do whatever you will that’s still a problem where this problem gets really big is if you can get some sort of money paid to you for each and every single account so imagine a situation where the government gives out some money to every real person if the person could set up 10 different accounts they would receive 10 proof of humanness times the payout now do you know who has two thumbs and is already working on a solution this guy I couldn’t find a picture of him pointing two thumbs at his face but yes Sam Alman the founder of open AI some might say the person kind of potentially may be a little bit behind some of these issues he is also working to address those issues now he has a solution he has a really good solution that will fix this Dark Forest of the clear web problem he has a solution that will fix this Cil attack problem but here’s the thing you may not like it because it has to do with harvesting your eyeballs list allow me to explain so really fast I have to make the disclaimer so I’m not trying to convince you anything I’m not trying to get you to buy anything or do anything I like many individuals have my own reservations about this right so I’m not like Yay problem solved that’s not the point of this there’s this really good quote there are no Solutions there are only trade-offs so kind of keep that in mind as we go through this because this would certainly solve some of the issues but it has some trade-offs if we go with a solution like this then we have to take the good with the bad and if we don’t then we have to figure out another way of dealing with the whole issue of automated bot attacks potentially large scale propaganda campaigns or just you know people creating tons of spam content those automated phone calls right all of that stuff how do we get around it how do we improve it well a Sam Alin back company presents a potential solution it’s called worldcoin now the worldcoin is going to have a lot of things that are like kind of dystopian there’s going to be a lot of people that there are a lot of people that just do not like this thing and I’m sure some people in the comments are just going to be very unhappy that I even show this thing but it’s important that you are aware of this idea not so that you sign up for it or whatever but these are the things that will be suggested in the future the things that will maybe be mandatory for people and so on this website here they’re talking about AI funded Ubi so sort of everybody in the world getting a uh Equitable global distribution right so everybody getting paid once uh the age of AI begins but we’re going to need proof of personhood so as AI advances it’s becoming increasingly important to differentiate between human and AI generated content online what if we had a digital passport that can be used to privately prove uniqueness and humanness online so for example let’s say you open up a bank account they need to know who you are they need your driver license they need your name they want to see what you look like Etc but if you’re writing something online or you’re you’re opening up some social media profile they don’t necessarily need to know who you are they just need to know are you human and are you a unique human meaning that do you already have 10 million profiles with us or is this your first one let’s say there’s a social network where you can only have one profile or maybe two profiles one personal one for work but no more than that they don’t need to know your name your address or where you live or your height or your date of birth or anything like that they’re just like who’s on other side of the monitor are you a real unique human being if yes then here’s your account if no then well we can’t give it to you so in other words you can have something like a email verification but anyone can get email then you can have like a phone verification where like an SMS text message is sent to you to verify that you know you know you have a phone number but again I mean you can have a lot of phone numbers it’s just it’s a little bit more difficult than an email but you can still game the system you can pay you know to verify that with a credit card so a credit card you can pay for something but you know not everyone has a credit card not everyone necessarily wants to transact online it costs money not only to the person for paying but also to the merchant or you can have like for example a driver license but that has all your information on there that has your name address date of birth right so maybe you don’t want to be giving this to every single place that you interact with online right that’s a little bit scary so if the question is how do we prove not identity but you know humanness without all the extras they’re calling that proof of personhood so your person no other information is given this this protects against Cil attacks it will prevents you know prevent spreading disinformation at scale so something like a proof of personhood would be very useful now let me just stop here for one second be super duper clear I think the idea of a proof of personhood is a good idea I think we’re going to need something like this at some point like it would be useful to have something like this so I I approve of the idea but that doesn’t mean I’m necessarily on board with just everything here right so I mean they’re talking about some sort of a world ID you know Global Ubi and they kind of harvest your biometric data basically they scan your iris to to kind of use it as a fingerprint like a unique identification and then that’s stored in their database so like right there’s a lot of things that kind of go into it that’s kind of like wo so also it’s a cryptocurrency on the ethereum blockchain right so people are trading that coin like Bitcoin and Dogecoin Etc so so I just want to separate that so pop proof of personhood so this idea how can you verify that you’re a real person without a governmen issued ID without some centralized thing right something like this could be useful like if I go on Twitter SLX right if I could just look at the accounts that have verification of this that that’s a unique individual with just one account that’s posting on there like I’ll certainly probably I mean I check that box and say yes I only want to see content from proven humans I’d rather not see all the automated content so I like this idea but and I don’t think this idea was necessarily Sam Alman and opening eye I don’t think they were necessarily the first people that kind of do this to come up with that idea so this idea probably existed before but they took the idea and they made worldcoin wld World coin and you know the whole world passport 2.0 and all that stuff so I like the idea behind behind it but harvesting your biometric data putting that into worldcoin having that be a passport somehow having it be linked to Ubi and eventually AI governance I mean obviously there’s a lot to unpack there right but I think the point is this solution will be proposed there will probably be Government Solutions that will be proposed so this is yet another thing to think about where do you fall kind of on the tradeoffs versus opportunity Spectrum right this is the World coin orb so you go in and you look into the orb right there’s Sam Alman doing it so you look into it it scans your iris gives you a unique ID and then later if you go and do it again you’re not going to be able to get a different ID basically and the other thing to think about is this right now people are very worried and unwilling to do this I mean for the most part there’s they do have 4 million people signed up but if they started giving out Ubi some sort of a payments through this thing like it would be a whole different story wouldn’t it so if anybody that is you know has this proof of personhood starts getting 2,000 a month for for nothing just for having this sort of proof of personhood now of course then I think a lot more people would start signing up anyway so the point here isn’t for me to tell you to go out there and do something or buy something like no I’m not suggesting any action I’m just letting you know that this is what is being talked about as a potential path forward for proof of person Hood plus Ubi the people that are concerned with you know World governments of course hate this the people that are concerned with cryptocurrency hate this so it’s running on the ethereum blockchain so vitalic butterin is the guy behind the ethereum blockchain vitalic I believe was part of the teal Fellowship so Peter teal backs certain very promising individuals to go and do stuff certainly vitalic certainly was one of the notable people to come out of that ethereum is what half a trillion looks like approaching half a trillion market cap but the point is that they’re proposing something like this to you know help with Finance with transferring money worldwide keeping the Bots out governance of for example AI right so for everyone with one of those identities to be able to vote for how AI is governed and they’re talking about incentive alignment equal distribution of scarce resources so this is one of the proposed Solutions whether or not it takes hold is hard to know certainly I mean as I read through it you can see why people are nervous I mean imagine something like this falling into the wrong hands that’s of course absolute nightmare fuel but if not this sort of solution what other Solutions do we have for dealing with some of the coming changes with sort of like the rise of automation including on the internet and my solution in the workplace now if like me you have some concerns about these sort of centralized World IDs and these various proposed Solutions I do have another solution to this whole problem in mind it’s easy foolproof it does not require a any agencies to to come in and help gather data or anything like that here’s how it works imagine it’s 10 years in the future you’re on an online dating site and you find a girl that is perfect she’s smart funny she likes the same things you do you guys hop on a video chat and you like her even more she’s beautiful and seems to be interested in you there’s a tugging thought at the back of your head going this can’t be real she’s a bot you just can’t shake the feeling that this is too good to be true so you ask her can you prove prove that you are not in fact a robot that you’re not just running on Chad GPT or gemini or something like that she responds by saying the most horrible things imaginable calling you the worst slurs that you’ve ever heard in your life the stuff that she says is the most foul most offensive and the most hateful of speech once she’s finished you feel relieved because you know there’s no way a large language model said that no but seriously I mean I if you remember the Wall Street bets subreddit at the height of the the craziness that was going on I mean how they talked that’s how we all will talk online to prove we’re human that will literally be the new online lingo if you’re not constantly just throwing out various slurs and offensive content no one will know whether you’re human or Not by the way the movie they just made about that was actually pretty good dumb money it’s called it’s it’s it was pretty good not bad I I enjoyed it