Ai_027 - Is AI Breaking the Web?

[00:00:00] I noticed something recently as AI takes over my experience in searching the web, in finding answers to problems, in troubleshooting issues that I'm having. I noticed that all of the websites that I used to visit, even videos and things that I used to watch to get the answers to my questions or to solve my problem, are no longer receiving me as a visitor. The AI is acting as a middleman to prevent the very traffic that is used to monetize the Internet itself. [00:00:38] When you begin to change the fundamentals of a monetization paradigm, nothing will be left untouched. A huge shift is coming to the web. What does it look like and what should we do to prepare? I am Guy Swan and this is AI AI Unchained. [00:01:06] What is up, guys? Welcome back to AI Unchained. I am your host, Guy Swan, and this is where we explore how AI is going to take us into a future of autonomy and decentralization and not systems of central control. A huge shout out to Swan, Bitcoin and Coinkite. Buy your Bitcoin at Swan Bitcoin and For the next $10,000 worth, you can get it with no fees whatsoever. And also you can plug your life into Bitcoin. You can set up an ira, you can set up a business and get it on your Treasury. You can learn everything you want to know about it, you name it, and then you want to get it behind your keys on a cold card hardware wallet. If it is not your keys, it does not belong to you. Check it out and get a discount with my discount code Bitcoin audible, which is right in the show notes. All right, so I wanted to go over a few things today. First, there's just been a bunch of general news. There's been a couple of things that I figured out in the last week that I'm going to be doing a video on and I'm super excited to show you guys. And then also I wanted to extend the conversation we had from the last episode in what I think how I think AI is changing the landscape and that it's going to happen a lot quicker than I had suspected and be a little bit different. [00:02:32] What do I mean by that? That's pretty vague. So one thing I had not quite realized in trying to take advantage of or thinking of what the value of AI is and specifically in agents and LLMs, being able to scrape the Internet to be able to sort through so much information on the web and then pull essentially the requisite answer or relevant information to you and then display it to you in a way that is relevant to your question. As text, as an image, as voice, like there are, obviously this is a huge benefit, right? And when you think about it, it's the purpose of surfing the web. It's the purpose of going to different websites, to checking on a forum, to signing up accounts at a bunch of different services, watching a bunch of YouTube videos. It's parsing through all of this information in order to get what you need in order to accomplish your task. And it's also important to remember that this could extend pretty quickly to podcast episodes, to YouTube videos, as these things get better at parsing out a transcript and more accurately and then being able to make the corrections, the error corrections necessary, depending on the content, depending on the context and the subject of the video or the podcast or whatever the content is, any sort of media, an image even, it's going to pull that relevant information out of that content for you without you having to listen to, engage, or go to the content itself. You know, a good example is Cascader, which, if you remember, we had them on the show very recently to talk about their app. And one of the things that they have that they had, I think they had just launched when we did that episode, was Cascade or Vision, which allows you to edit a YouTube video or pull clips from a YouTube video based on the actual text of the. Basically what Descript does. If you haven't used Descript, it's actually a really fantastic application. It's got some quirks, but if you're trying to do quick editing based on a text, based on like what you are saying in a video, it's a really, really handy service. But Cascade or Vision is actually very similar to this in the fact that you read through the transcript, you have the transcript on the side, and then you highlight the text of the comment or remark. So in this example, if I could literally take this comment, starting wherever the sentence started, and then just cut it out in the text file, and then it will essentially print or it will essentially export the video file of the clip that I grabbed, I selected from the text. And they have another one that. And this is, you know, pretty frequent or pretty regular. And this is what I did with my small app on Whisper, my little scribe drop app that I have on my computer, which you can see in the other AI Unchained video that I did the tutorial. I'm just easily and quickly doing a drag and drop transcription of a podcast episode of any audio or video formatted media. And then of course, there's another one of get it's titled the Personal Newsletter Assistant, but it gives you curated summaries from a YouTube channel or a video, any videos that are published. Now I can see this being really useful, right? Is I'm pulling information from these videos and stuff and I get a summary of a bunch of videos and maybe that even entices me to go watch the video, right? Maybe, but maybe I actually get the information that I would have. It would have taken me 15 minutes to watch the video by just reading the paragraph related to the information I was trying to find specifically. So, for example, BTC Sessions has fantastic tutorial videos on YouTube. I'll link to this if you don't know and have not followed BTC Sessions. But he has fantastic like hour long to deep dives on a bunch of different products, wallets and all the different things that you can do in the bitcoin space. But maybe I'm just looking for, you know, something specific to the cold card or just any hardware wallet or, you know, Nunchuck Wallet. And so rather than going and watching his video, I search his YouTube channel for, you know, what do I do if I'm using Nunchuck Wallet and I'm trying to set up a multi sig with cold card and tap signer or something and then the LLM parses through that video and then spits back out the answer. And this is something that's already happening. Like if you. Every single search engine, every single one of them, if you look now, they all have AI Google. If you just ask it a question, it will parse through the Internet, parse through everything that is that it's searching through, find the relevant information and then link the sources. Not all of them all the time, but it will link the sources that it got from. And in fact, a lot of the LLM might come up with a lot of the answer itself, which specifically means that it's not linking to any sources. Just kind of like what we talked about in the previous episode is, you know, the amount of art and like all of this content and stuff that's being used in a AI generation, a video or image generation or just an LLM in general to be able to answer some sort of a question that somebody else somewhere actually took the work to do the research and get the answer and obviously is not being attributed to that. [00:08:38] Where that information came from is not getting the attribution to the person who put it together. Same with the art, same with the video, the style, you know, whatever it is. Now obviously nothing is truly original, but all of this Stuff takes work. You can't. Art doesn't just come out of nowhere. Someone has to create it for an AI, for an AI model to be trained on it. Someone has to do the research and find the answers for an LLM to be able to spit back out the answer. And if I search Google right now, I've literally noticed this with the last couple of weeks of searching stuff is that I don't go to websites anymore. [00:09:21] I basically get, I try not to use Google anyway. I don't like to use Google, but if I'm using Google, Bing, any of the major search engines, they have AI in them now. And if I am looking for something, the AI will find the sources and I'll look. And I was like, okay, cool, it's somewhere in this thing. I think the only one that I went to actually recently was Stack Overflow because it didn't quite understand my question. And the rest of the answer in Stack Overflow actually had what I was looking for. So there are clearly few exceptions. But the key point is that I didn't go to the website in the context of the BT Sessions tutorial video. And trying to find that piece of information is I didn't go to the video. And what I hadn't realized is this completely breaks the monetization of the Internet. Google is disrupting the very thing that they make money off of. The reason people care about Google search is, isn't because Google then finds the answer on their website. They don't want a website scraper to deliver their information to the customer, to the user. They want the user to come to their website because that's the only way to actually monetize anything. So these search engines are literally taking all of this other information that these other people have created and avoiding getting around the need for the people to actually go to those websites or actually go to that person's blog or that person's Twitter page or the Reddit post or whatever it is that actually has the answer. Imagine if 99% of the traffic to Stack Overflow are. Bots are just AI agents pulling the answers from, from the Stack Overflow database. And nobody ever needs to actually go to the website anymore. Stack Overflow literally dies. [00:11:26] It can't make any money off of bots coming to the website because bots don't click ads, bots don't see, don't have product recognition, even in the face of people not clicking on those ads. Like, even when I go to a website, I might not click on the T Mobile ad, but I'LL think about T Mobile. Like it will pop into my head and it will actually be fresh the next time I think about getting cell service of some sort or changing my plan. In fact, a great example is Cricket Wireless. I actually watch those ads because Ryan Reynolds is absolutely hilarious. But if I don't have to go to the video to get the information that I'm looking for, I. I don't see the ad. And I never hear about Cricket Wireless. Had it not been for Ryan Reynolds commercials, I literally never would have watched. I wouldn't have known anything to speak of about Cricket Wireless. And this could happen really quick. Like really, really fast. That the entire monetization model of the web, especially when the big search engines are doing this already. Seriously, how much, how much traffic is gone? [00:12:38] Like I said in the last couple of weeks, and it's not because, like I went to Google and I said, let me use AI and I signed up and I put in some special thing. Google just does this now. Bing just does this. The whole way that they make money is by selling ads in their search engine results. And there's a whole economy, there's a whole industry basically geared around good SEO. How do you get good search engine optimization so that you are the one that shows up in the search results that may literally just have been obsoleted in a matter of weeks. You know, I'm going to try something really quick. [00:13:19] Let me, let me search something really quick. Because I had this problem earlier and I searched it on YouTube and I realized I didn't think about the fact that I could probably search it in Google and It's about a DaVinci resolve issue. I was trying to figure out how to change the anchor point of an animation so that rather than being in the center of an object, I could put it at the end of an object and it would, you know, pivot. Like it was like it was on a shoulder or an elbow or something, rather than pivoting in the middle where it would just spin like a propeller blade. So I'm going to edit this out, but I'm coming right back. [00:13:51] Dude. Holy crap. [00:13:53] So not only did the AI give me an answer for the keyboard shortcut, but it actually showed me a video. [00:14:05] A 14 second. I didn't even. I don't know if I noticed this or when this started happening, but it actually showed me. It did exactly. What I was just talking about is that it went to the video. Went to a video. That's ridiculous. That's so crazy. I literally didn't know it did this. Then it went to a video. And not only did I, I didn't watch the whole video, but I could click on the video and it plays the exact point in the video in which they are talking about making changes to the anchor point of the video frame. That's so crazy. It's already parsing through the video and then finding the exact point that is relevant to my question and then leading me to it. That's where it starts the video when I click play. God, that's so crazy. [00:14:57] But the kicker here is that this is not compatible with literally the entire monetization model of the web. [00:15:09] Websites have already kind of been on the sunk cost side of the equation for a long time. You know, an enormous amount of activity has moved to, you know, particular apps and of course social media platforms and like tons of traffic have centralized around social platforms as opposed to actually going to websites. You know, I see I have interactions with people and or companies and you know, accounts that represent companies far more often on social media than on those companies. Websites save for sites that I actually have to use a service on their site but then typically they have an app and that's more often how I engage with it. Like I don't go to the Swann website very often. [00:16:01] They have the Swann app. And you've noticed that the monetization has moved towards social media. And it's also interesting that because that's essentially where people remain engaged and are interacting with others. And it's funny, X has even tried to. They've started to literally pay people for engagement on social media. But I think a lot of this is just kind of going to show how far we've come from the old monetization model and how quickly it might kind of reach its end of days. [00:16:35] And importantly, what I think is going to replace it. Like, you know, how do we solve this? What is the solution to this? And I actually want to hit on a few other things first before we come back to that specifically because I think they are related. This episode is brought to you by Swan. Bitcoin. Do you want to get bitcoin in your ira? You want to set that up in just a few minutes? Stack sats, pre taxes. What about a multi sig vault? What about getting your business treasury and even an employee plan so you can have a benefits plan and you can have your employees get bitcoin every month as part of their benefits of your business or you just want to stack, you want to stack bitcoin, you want to start your savings account or continue your savings account easily automatically and have it automatically send to your cold storage. This is the way to stack Bitcoin. Just do it. Just set it up and it will run all the time. It's a subscription. It's buying a subscription to a better future. Check it out at My Link Swan Bitcoin.com Guy Swan Bitcoin is the place to get all of your bitcoin services, your entire financial suite, retirement business, white glove service, a team to ask any questions to automatic savings account. And for the next $10,000 you buy, there are no fees at all. Check them out. The link is right in the Show Notes Swan Bitcoin.com Guy now anybody who knows listens to my shows, has heard about the pair stack and you know a lot of the different things that are going on with new peer to peer technologies, with new decentralized alternatives and protocols and all of these things. Nostr being a huge one. And if anybody has not signed up with Nostr, I urge you to check it out, please. Super sue actually, who is the developer for Wholesale, which was the previous episode, excuse me, of the Pear report that we just did, he actually just set up on whole on Nostr and got a few zaps and that was like his first time actually making money directly from other people in a social media environment. And he immediately like posted in the heat room or whatever where a bunch of us are hanging out and he was like, I gotta say this thing's actually really interesting. That's it's kind of neat to have just randomly gotten zaps have gotten money from people on a social media environment. But anyway, I've been trying to figure out after stumbling upon Wholesale, which was a tool that greatly simplifies, it's kind of like tailscale without account setups and without a central server or anything. It's just direct connections between two different machines. So any service that you can run locally, you can now run remotely, you can run over the Internet. So not only can you get that service in your house, including just like an API, but you can also just from wherever you are. I can go to a coffee shop, I can go to the airport, I can go to any, I can go to an Airbnb somewhere and I can use any service that I have running on my home Linux machine remotely using wholesale. And there are a lot of AI services that let me do this. Pinocchio in particular, which I have talked about on the show, which has a number of hiccups on Linux or at least on POP OS as far as I've seen, and I suspect that a lot of different Linux distros have some sort of an issue here or there with installing a handful of the tools on Pinocchio. But the two best AI gen like video and image generation tools are Comfy ui. At least in my opinion are Comfy UI and automatic 1111. And both of those work on my POP OS really well. And I can remote log into them and just use them over the browser with my MacBook at any time that I want. This has been a huge help for the handful of things that I will regularly do with this. When we're, you know, upscaling something or we're doing something with the show or making a social media post and we're compositing stuff. And I like to use it from time to time just purely for ideation on story projects and stuff. Like I don't actually use any of the imagery. [00:21:02] I literally just use it to, you know, zone out, listen to music, generate a whole bunch of stuff based on some sort of context or prompt, and literally use it to try to come up with ideas for stories or for character elements and that sort of thing. Now one thing that I've had trouble with, which has kind of bugged me, was doing this with LLMs. [00:21:28] And it was not until very recently that I realized Olama is one of my favorite interfaces. It's actually just, it's actually just like straight in the Terminal with Mac os. But I installed Olama and I installed it on my Linux machine and my Mac. And of course I can run much bigger models on my Linux machine. And it's like right in the terminal and everything. But I did not know how to. I thought I had to do some like crazy settings to get, to get it to use remotely and all of this stuff. But it is actually serving an API automatically. Like I didn't. I don't have to put in a flag or anything of like, you know, dash, dash, listen and put in a port number or anything like that. It literally just automatically grabs a port number and is listening for a. For API commands. Which means with an interface you can just give it that information and connect to it. And there is an interface called open web UI. For anybody who's used ChatGPT, this is going to feel extremely familiar. Or honestly, anybody. Anthropic or any of the tools. Anthropic, ChatGPT, any of the major AIs, it is the chat interface, right? Let's chat. Unleashed or Unleashed chat does the same thing. Venice AI does the same thing. It's that conversations over on the left, you pick your model up at the top, you ask it whatever question. It's a Q and A sort of design and setup. It has memory and importantly and really fascinatingly, it also lets you a set up accounts with your with the web UI so that you can give access to others from or with limited access to your ui. But it also lets you upload files and use them as context use them in the actual conversation. So I can literally put a file, a transcript for my podcast and ask llama3 rather than chat GPT to give me the summary of the podcast, to give me the, the, you know, the things I mentioned in the show that to link to which I have not been able to do locally yet. I have not solved that problem until just now. Until I just found this and I posted this in the AI rabbit hole in Keat and I think I dropped it in audio notes as well. But this was like two days ago, maybe yesterday actually, I don't even know. But this is so cool. And the thing is is I can get this interface because of wholesale, because this is running on a local host port. I can just direct that localhost port to through wholesale, just grab the port number, spin up wholesale and then with the key that generates, I can use this from anywhere on any of my machines and I could just as easily host this so that somebody else could use it and you could have an account that required my approval and you could pay a couple of sats or whatever to generate answers to generate like to basically use the resources on my machine to work with your data, to ask questions, to get summaries, to do whatever. And even crazier is that open web UI actually lets you plug in a video or an AI generation actually. And I'm not sure about video. I haven't, I haven't gone that far but. And I haven't done this one yet either. I've just only discovered that you are able to do this, but you can plug in a image generation model as well so that you can connect it to automatic 1111 and have it so that you can make the request and use a prompt right inside of the window of the open Web UI chat, just like in Anthropic or chatgpt and you can have it generate an image from a prompt. And with llama 3 and a lot of the bigger models, if you have enough hash power. And one of the other really cool things about Ollama in particular is that it will automatically use multiple GPUs so if you have like 2, 3, 4 like monster GPUs on you in your computer that will grant you enough VRAM to run run some of the really serious models out there. Ollama will automatically grab and spread those jobs out among all of the different GPUs and it will run like a breeze. Guys, this is everything. This is the all of the pieces in just a couple of open source pieces of software. This is basically the entire stack of being able to host a service for anybody anywhere in the world to connect to you and use or be able to share resources directly from your machine or to do it specifically from yours. So if like the value of my home AI machine, of my GLaDOS which somebody actually won, they donated sats to name my AI machine and I haven't yet put it, I was going to do a 3D printed thing for the name and everything on the glass case on the outside of this thing but I haven't done it yet. But they got to name it so I got to remember to call this Gladys. But the value of this thing for me just went way up because I'm actually going to be able to build this into my workflows so that something that I've had to manually do with chat GPT. If I can get answers that are good enough from llama3 that can do those jobs, I can actually build that into my little micro apps of just dragging and dropping. And now after it gives me the transcript I can have it automatically put in a prompt for the episode and then give me out the requisite links and the descriptions and all of those details which have required. Like I might not really need ChatGPT anymore. Like I could potentially just cancel my membership. That is a huge deal in my opinion. Like that's, that is a really big deal. And this is an expensive machine, but it's not a ridiculously expensive machine, you know. And suddenly the value of being able to host and accomplish these sorts of services and using these sorts of tools on my own in my own home, it just feels like getting to that self hosted future is so much closer. [00:28:20] And what's funny is this is still just the very beginning of the pear stack. Stuff like that's really the piece of the puzzle that put all of this together was being able to do this remotely and being able to just easily and quickly have an encrypted connection directly to my machine and wholesale is brand new. Like all of these things pair runtime, the platform for all of this literally dropped in February of this Year. So all of this stuff is super fresh. There's not even a UI for wholesale yet. And also another thing actually to point out with Ollama is that you can run the lava. [00:28:58] It's Llava, the lava model. And I think Lava is on version three now for that model. And that is a multimodal model that can actually look at images and describe them. And one of the things that I did just to see is I got it to read a. [00:29:20] I gave it a picture of a book on I had sitting around here somewhere. [00:29:26] Where is it? Theft of Fire by Devin Erickson. And I took a picture of it with my phone and then gave it to the lava model to describe it. And it was funny. Lava 2 got the title wrong and the subtitle wrong. Like, it was like close, but it was basically, like kind of goofy. And then it misinterpreted something. There was like a quote of. There's like a tagline of it. And it misinterpreted kind of the context and took it in a weird way. I can't remember exactly the details, but I remember reading it thinking, like, it's like one of those things that, like, oh, yeah, that's sort of accurate, but not even. But not at all what it means. And I thought that was kind of funny. But version three actually did a fantastic job. [00:30:15] And the only problem was that I think it misspelled slightly a couple of different things and it had like one or two words off. But for all intents and purposes, it got the context. And I'm taking a picture of, you know, an artistic cover of a book. It's not like white paper with black text on it. And it also would describe the COVID that there were, you know, three characters on it. And while it would get some things wrong, that's a very interesting. [00:30:46] Like, being able to get to the point where it can describe, it can see and start to tag or understand files is really fascinating because this will let me. This is where I start getting into the realm of being able to actually chat with my notes and with being able to automate tagging or, you know, writing certain descriptors about certain files or images and auto sorting them. Because one of the things was always like, in any of those things that I tried to do in the past is I would basically have to have a script flow for every single different type of file or thing that I usually would save and a format for it. Whereas this, I would still sort of need that in a certain context, but I could get it to look at and discern what type of image it is or what type of content it is to essentially do most of that job for me. And then importantly, it could add tags to the actual file so that I could find the important details of those files later and what I'm actually looking for. Which this might be more of a vector embedding thing. And I haven't quite figured out the secret sauce to this. I have learned recently that embedding models, I guess it makes a lot of very intuitive sense and I should have thought that this was the case. But embedding models are actually. You can use. You can kind of start out with a very generic embedding model, but that you're actually probably going to aim at a very targeted embedding model because the semantic relevance is actually based on what semantics matter to you. This show is brought to you by Coinkite. And I gotta say, every single time I do it, I still like, I just have this little, like this get giddy. It's just like this is so much fun having the NFC in the cold card and in my tap signer because I have a multi sig for nunchuck and one of them is with a tap signer and my mobile key and then the other one is with my cold card. Both of these are Coin Kite products with nfc. So I can just tap to pay on these. And the security, the comfort that I get knowing that I can have all of the capital that I need in Bitcoin for my business, for my savings, even my cold storage, accessible and viewable in, in my nunchuck wallet, in my mobile wallet, on my phone, still secured by a hardware wallet and still able to access by creating the transaction, tapping it to my cold card, signing it and tapping it back. I can do this on the fly. I do this with my tap signer while I'm traveling and with the tap signer it's literally one motion and so I can carry this around in my wallet. I am a massive fan of, of Coin Kite and their products have been for a really long time. And I have a discount code, Bitcoin Audible, which is the name of my other show. And if you haven't gotten a cold card or you haven't checked out the cold card Q which even has a password manager built in. It is so cool. You're missing out. Check them out, go to coinkite.com and the link will be right in the show notes. And don't forget my discount code Bitcoin Audible. So just for example, imagine there's A there's two sentences and it says a dog walks into a store and a dog walks out of a store. [00:34:23] Well, in certain context, in the context of like, what's the subject of this? When you're doing a semantic search, if you're searching for things that are related to dogs or things that have a store in them, well then both of them are essentially 100% relevant. But depending on what and why you are looking for that information, they also could be considered opposites. And the embeddings are based on what semantic relevance is trained into the model. So if you have one that's trained on just kind of like the presence of nouns or, you know, a dog or something like that, well, then it's going to have those two things as semantically right next to each other. But if you want your embedding to be related to an action to take in a certain event, well then the model, you would want the model to treat it differently based on the action or how the situation has changed in the context of what is being described. So you would want all of the things related to entering the store to basically garner one context or one set of event or one set of actions and everything exiting a store to be completely different. So that's just to say that based on what the vector embedding model is or how it's trained is going to be relevant to what information or what, what way you are probably recalling your information or your notes or your documents or your images or something at a later date. So you might actually want to dig through and find different embedding models based on what they are actually trained to do, like what they are created for in a certain. Whether you work with a lot of spreadsheets, whether you work with a lot of just written notes and journals, or whether you work with a lot of image files, like this sort of thing, you might need a completely different model based on what you are trying to do with, with it. And then this will let you create a vector database with or against all of your files, which I still haven't figured out quite how to. Like where do you put the database and how does it connect the files? You know, like one of the things that I've had constant trouble with and it's like the, the never ending awful situation of like DaVinci Resolve and like video editing software. I can't believe that they actually haven't solved this problem yet of doing a kind of universal way to find files and connect them to DaVinci resolve every single time I'm constantly organizing my VFX files and the stuff that I download offline or download online. And then, you know, like, for instance, I'm doing a 2sats video right now and I downloaded a video of a little kid climbing through a pile of money. And it's Venezuelan dollar bills or whatever, pesos, and that's Argentina, what's Finland Bolivar's. But it's literally like a giant pile of cash and he's like jumping around in it like it's leaves on the ground. But I downloaded it, stuck it in my stock video folder, then dragged it to DaVinci Resolve and started editing and cutting it and putting everywhere that I needed it. And then a little bit later I downloaded and put up a couple other things in the timeline. And then I went back and like renamed and corrected a couple of things because, you know, it's not based on my format at all. You know, it's just kind of like randomly stuck in a folder with whatever the random hash and bunch of numbers and nonsense it was. And I want to be able to find it. I want to be able to search it for pile of dollar bills, you know, money inflation, money printing, whatever. And so I put in a bunch of these tags and immediately half of these, half of the stupid things, God forbid I changed the name of a folder. Holy crap. And then I have to go back and I have to reconnect all my media because I go through the timeline in the video that I'm editing and like everything's just like black with like, like it was just red with the text and like media not found, media not found. So in that same. I have this, I have nightmares about this. And so like my other thinking is like when I create a vector database based on a bunch of different files, I create a bunch of embeddings from the files. Is how do I make sure these things are attached to the files and that if I then like, am I stuck? I can't move anything on my computer. I can't change the. I can't rename the file because that would be useless. But there's a really cool tool and I don't know how it works, but there's a really cool tool called Hook Mark or I think the took the tool. It specifically is called Hook Productivity and it's on Mac only, which frustrates me because I want to figure out how this creates this file system tracker, whatever the heck it does, and how it finds everything. And it's a little bit like a SIM link, I think, but it Essentially creates a. What's called a hook mark, kind of like a bookmark. [00:39:32] And it's a link that, while hook is running, you can click on it and it will find or reach out and get that file from anywhere on the computer. And that includes if you move it to a different folder, if you move it to a mounted drive, if you change the name of it, if you change the description, all of it. You can edit that file however you want. And as long as it's the same file fundamentally and the drive is mounted. That's one of the crazy things is that, like, I'll have a drive that I'll reboot and I'll forget to remount the drive, and then it won't find it. And then I boot it all up, plug the drive back in, and then I click on the link, and then boom, it just pulls the file up again. I don't know how it does this, and I wish there was an easy. Like, there needs to be a universal linking system to files, and especially whether or not it's like a do vector databases that update for every single folder. Like, I don't even know the degree to which it would take, like. Like, how much computation. How long does it take to, you know, create vector embeddings for an entire computer's worth of data? I. I still don't know. I. I haven't even gotten that far yet. But I can see. I can. I can begin to see the roadmap of being able to look through all of this stuff. And the really cool thing is because that's essentially what it's doing when I'm moving, going back to the open Web ui is being able to drop a file into it. [00:41:01] And there's a. There's another. There's a plugin. I'm not sure if anybody else uses Obsidian. I have started to use two different ones that are basically built on the same thing. I'm using the same notes, and they're both in markdown, but LogSec, L, O, G, S, E, Q. And then also Obsidian. Obsidian is my main. But Obsidian has plugins, and one of them will let you plug this open Web UI or no. Is it a llama? I can't remember exactly how it connects. It might just be olama, but it lets me talk to. This is also one thing that. One of the things that I haven't set up, by the way, I've only just discovered this as I went down the rabbit hole of realizing what I can do after I set up Web UI and realizing that There's a lot of other offshoots of this that I can do that can have my own local LLM connect to my Obsidian nodes and then I can make requests, I can highlight and you know, make commands and summarize, finish a paragraph for me right inside of my notes app, right inside of Obsidian. I can have the chat bot or the chat box right over in the, in the sidebar and I can work directly with my notes. [00:42:14] That, that is huge deal and I am very, very excited. Like I said, do not forget to subscribe to the YouTube channel at the Guy Swan and I'll remember to have links to Logsec and Obsidian and all these things and open web UI as well and Ollama if you want to start trying around trying this out as well. But I'm going to do a video very, very soon. I have to finish up my two sats and the little wholesale video that I'm doing so. Because that's kind of a prerequisite to getting this to work remotely. But that's one of the, just the coolest things about this is that I can do all of this remotely from anywhere. So keep an eye out for that because I'm going to have a video on it, I'm going to break all of this stuff down and show you my setup. And I'm really, really excited. But like I said, I've just, I've just scratched the surface of it. I probably should have waited to talk about on the show until I had more concrete examples and uses of it. But I'm. I'm too excited. I wouldn't be able to, I wouldn't be able to talk about anything without bringing it up. But all of this leads me back to where we started this conversation. [00:43:22] Talking about what happens when, what is the solution when the monetization model of the web begins to break down or is in the process of breaking down. [00:43:36] And I think when you put all of these tools together, when you realize what is possible and that we may actually be fundamentally changing the client server model, the idea of sovereign self hosted, easy to connect to each other services, sharing resources, connecting and utilizing, you know, services and computation peer to peer or even among a group in a social graph and being able to judge or allow connections and even go back to combining Bitcoin and the Lightning network and ecash and being able to pay might do micro payments for these services and to be able to pay directly for an API from somebody else to share resources and to be able to limit traffic based on your trust network in something Like a social graph from nostr. We're talking about an age where middlemen start to fall away completely. [00:44:40] Protocols are going to replace the platforms. And the thing is, is I don't think we have a choice. [00:44:48] That's what's crazy about this, is that because of micropayments with lightning, like I don't think we have a quote unquote economy in the sense of everyone participating and providing value if we don't actually shift this model to direct peer to peer. But this is also why I think for better or for worse, the open web is going to start closing down a lot. And I think there's a lot of different reinforcing elements here that will contribute to this. So it's not only the KYC problem, it's not only the fact that these tools are leading to the inability of proving who is human online and that AI is increasingly just scraping everybody's Data. But because AIs can interact, because these agents can interact with that data, people are going to be putting up walls so that they can API access. Especially people who have massive pools of information already available to them, like Stack Overflow because of their massive reach already and their massive trove of data and Q and A and all of this stuff. I could see them easily, if they start losing ad revenue or whatever, that they put in a block for AI agents and then start charging API access, which, which Stack Overflow might not be the best example actually, just because they are a great example in the context that they have a trove of data and conversations and forums and Q and A and all of that stuff. But they also have like a bunch of other products. So they, it may literally just be a loss leader to sell some other service. And I know they do a bunch of AI stuff already, so don't get me wrong, I know that this might not even be relevant to Stack Overflow, but I just use them as an example because they're, they're a great example of a place, a website that because of something like the Google Search, being able to crawl that website and get the answer that I am, that I'm looking for and then feed it to me as the output of the LLM rather than me, rather than me ever having to visit it, I can 100% see my answers from Stack Overflow staying high or even getting higher. The number of times that they are a source in some sort of a search that I do to find the, you know, find the relevant information that I'm looking for, but that my actual visiting to the Website plummets. But it's funny, without bitcoin and lightning, I, I would have a really hard time. Like this would clearly get grossly centralized. Like I would only see centralization as the path to actually get out of this because you need a monetary base on the platform. But you would not be able to create a robust ecosystem. You wouldn't really be able to create a market that way because there would be way, way, way too many, too much incentive, too high of an incentive to basically lock down and control it. And governments are not going to want to allow like, like a vibrant distributed market is to them dangerous. It's awful. And everything needs to be controlled and every need, everybody needs to be locked down. It all needs to be permissioned. I mean look at the level of restriction and the friction that is in payments and stuff today. Like even I'm just trying to use stuff like cash app or whatever to be on a bitcoin standard. And I can't tell you how many times I run into limits on stuff that I've used forever and then I'm fully KYC'd on and I can only move $2000 a week through a business, through stuff where I'm paying people. [00:48:46] And at every single turn there's another level of kyc, there's another increasing levels of, you know, travel rule. Like do you have, why don't you, you have to explain and write a little report on what you're doing with your money because it's too much or you're moving it off this platform. These are all disgustingly anti market. These are horrible, horrible frictions that disallow any sort of evolution in the market to actually, actually come about. And this is exactly why it leads to such awful centralization. But then at the exact same time, that's where you get stuck on platforms that take 60% of the payment, 70% of the payment. And we are in a market, we're in an environment, a technological environment where we need evolution at an insane pace. Like not at a normal pace. We need it at an even more aggressive pace than we would normally need it. Everything is accelerating and the monetization models that we have been used to are breaking down. If we are tightening the regulatory noose on every single thing that we do and how money is transferred and how everybody has to KYC for every damn thing under the sun, that's just a disaster. That just means that it's going to increase poverty by doing nothing, by simply not letting things, things loose. Evolution occurs through interaction, through testing and experimenting and at a time where everything that we are used to, every, every model that we have depended on and the platforms and tools that we've depended on and the institutions are breaking down and no longer serving their purpose, where we need aggressive experimentation, the cost to experiment is exploding. From a regulatory and government position, from a political environment perspective, and that is simply put, a recipe for poverty and dependence. [00:50:45] We need experimentation. We need people to not have to fill out any pieces of paper or do anything or ask anybody's permission, you know, basic common freedom in order to test something out, in order to explore an idea, in order to find customers and provide a service. [00:51:05] And what's funny is the technology is enabling that. It's literally being censored and attacked by the political apparatus, by the political, by, by the institutions that are supposed to be serving us and help to make sure that we actually have a prosperous society. Which is a joke for anybody who like, I know that's not anything to do with their actual purpose, but that's at least what is claimed the purpose of the government. But this is exactly why we need open payment standards. We need direct, we need peer to peer ecash and lightning and nostr and the pear stack self hosted AI. I don't think this is a nice to have. [00:51:50] I think this is a we have it or our future is really bleak. And you know, most things don't change until there is an enormous pressure to do so. Most people procrastinate to learn a thing or to do something until it literally hurts enough that it is worth it, that it hurts more to not do it than it does to bite the bullet and just get it done. But luckily for the people who are listening to this show, you are probably explorers like myself and love doing this crazy ridiculous stuff. So I'm happy that you join me and I hope that you learn and figure out how to run your own LLMs and do all of this stuff as well with me. And yeah, we will wrap this one up. Thank you guys. Thank you guys. I think I'm, you know, not to be a pessimist. I'm not, I'm not a pessimist. I am an optimist through and through. And I truly believe that the technology is on our side. It's just that humans tend to wait until things get really, really bad to step up to the plate and get the job done. But I think we can be one step ahead. I think we can learn the tools, we can start playing with the tools before we get there so that we know where we are headed so that we have some insight into where things are going and what we need to do to prepare for the future. So with that, I didn't even get to any of the news items and I've got quite a bit of stuff, so we'll just leave this for the next episode. I got a couple things to do as far as short videos. Do not forget to subscribe to the YouTube and Rumble channels. Don't forget to check me out on Keat. I've got the Guy Swan network room. I'll have the link to that in the show notes as well. A shout out to Swan, Bitcoin and Coinkite for supporting this show and making this possible. And thank you all. This is AI Unchained. I am Guy Swan. And until next time, everybody. Take it easy guys. [00:53:58] Disruptors don't have to discover something new, they just have to discover a practical use for new discoveries. [00:54:10] Jay Sameit.