TFTC: A Bitcoin Podcast - #463: Orange Pilling AI & Bitcoin Times Energy Edition with Aleks Svetski

Episode Date: November 30, 2023

Marty sits down with Aleks Svetski to discuss Spirit of Satoshi, the Bitcoin-focused AI project, as well as the upcoming edition of the Bitcoin Times. Svetski on Twitter: https://twitter.com/SvetskiWrites Spirit of Satoshi: https://www.spiritofsatoshi.ai/ Bitcoin Times: https://bitcointimes.io/ 0:00 - Intro 6:52 - Riffing 9:53 - What is Spirit of Satoshi? 18:12 - AGI 24:04 - Closed/open source AI 33:38 - Building your own language model 41:04 - AI’s opinions are based on its master 45:02 - Data cleanliness and hallucination 52:58 - 1000 minds vs the Borg 58:59 - A wealth of data 1:03:54 - Biden’s AI executive order 1:05:26 - AI agents and Bitcoin 1:12:23 - ASICs 1:16:36 - Energy 1:21:09 - Bitcoin Times Energy Edition, ESG 1:32:16 - Bitcoin is energy 1:39:29 - Conquering mindset 1:49:21 - Wrapping up Shoutout to our sponsors: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠River⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Unchained⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠CrowdHealth⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Bitcoin Talent Co⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ TFTC Merch is Available: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Shop Now⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Join the TFTC Movement: Main ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YT Channel⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Clips ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YT Channel⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Website⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Twitter⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Instagram⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Follow Marty Bent: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Twitter⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Newsletter⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Podcast⁠⁠⁠⁠

what's up freaks it's your boy marty ben here to introduce this episode of tftc i sat down with alex svetsky talk about the spirit of satoshi and the bitcoin times energy edition which i have the immense honor of having written a piece for esg gotta ridicule it talk about that on the back half a lot of ai discussion in this agi it's a red herring to control speech i think that's the lesson i took from this conversation it's a very important one don't let them bamboozle you with agi they want to create a regulatory moat and then create the models with the context that they provide to then project propaganda on you beware alex has ways to work around that and he's building that in spirit of satoshi

this was brought to you by good friends at river do you have an android phone do you like stacking sats you want to do so on your android phone river now has an android app go to river.com slash tftc set up an account today you get five dollars worth of bitcoin after your first hundred dollar purchase river is the best place to stack sats you can dca you don't pay any fees on your dca i think they just added the ability to dollar cost hour average every hour if you want to it's a beautiful thing they don't have any third-party dependencies they're not trusting prime trust fortress trust somebody to handle

their custody they do all that in-house you can sleep easy knowing that your bitcoin is one-to-one backed on river and stored in multi-sig cold storage they have river lightning services if you want to build in a lightning network you can leverage your api to do that so if you're stacking sats whether you're an individual an institution a business go to river.com slash tftc set up an account today making it as easy as possible this was also brought to you by a good friend down the hall what a point down the hall unchained they're down there working hard, building the financial platform of the future, which leverages Bitcoin's native multi-sig properties, which enable them to do cool things like bring you financial products

that actually work and make it so you can verify that your Bitcoin is where you think it is. The lending product is a two or three lending product. You put a Bitcoin in this collateral and you get a, you get us dollar liquidity in return. That Bitcoin collateral is held in a or three multi-sig wallet in which you hold one key unchained holds one king kingdom trust holds another key this is multi-institution custody that gives you visibility into your escrow account so you know your bitcoin's not being re-hypothecated blockfi is not taking it give it three hours capital celsius isn't taking it and doing whatever they did with it genesis is taking it and giving it to three hours too it stays in the wallet you can audit that on chain because you have a key in

in that quorum multi-institution custody is something that unchained pioneered and will be the standard moving forward unchained also has a trading desk they have an ria product which allows you to hold your own keys as well they have an inheritance protocol they just launched an ria and sound advisory go check all this out go to unchained.com slash consultation set up a consultation today learn about what they're building why they're building it and how they're doing it. Tell them that TFTC sent you. Unchained.com slash consultation. This report was also brought to you by good friends at CrowdHealth. CrowdHealth is here to help you opt out of the incumbent health insurance industry. In that model, it is healthcare enrollment season

here in the United States. And I highly recommend as a CrowdHealth community member myself for almost two years now to consider CrowdHealth as an option and take care of your healthcare cost. health insurance is notoriously opaque expensive and impersonal crowd health is the exact opposite it's very personal it's much cheaper and it's very transparent they connect you directly with the doctor they would walk over how they're paying your bill how they're negotiating lower what you do is you pay a monthly fee builds up an account if you ever have a health event you go to crowd health say hey i'm going to the doctor they say get the bill bring it back to us we'll negotiate the price lower you pay the first 500 of that bill and then the crowd health

community pays the rest i've used it to pay for some health events in our family and i've helped others pay for their health events it's a beautiful thing it's how you opt out take control of your health care cost it's much cheaper much better so it should be done me my family are very happy go to joincrowdhealth.com slash tftc sign up today and you'll get 99 for the first six months of your subscription or just use the code tftc at checkout joincrowdhealth.com slash tftc last but not least this report was brought to you by bitcoin talent co bitcoin talent co is a recruiting firm built by bitcoiners for bitcoiners if you're a company in the space looking to bring the best talent poach them from other industries go hit up bitcoin talent co get set up with them they are

bitcoiners that understand what's going on at the protocol level what's going on lightning what's going on a mining what you need they know how to go find it it will work with you to understand your needs your process and they'll go get you the best talent if you don't want a full-time employee you can tap into their flex product which allows you to tap into a roster of part-time contractors can help you on a three-month sprint and six-month sprint maybe need some design work done some growth marketing strategy uh maybe you need an engineering sprint the talent co flex product can help you there so go to bitcointalent.co tell them the tftc sent you get on board as a business as an individual if you're looking to get into the industry get set up with them as well

get your resume and your profile in the mix bitcointalent.co tell them the tftc sent you and enjoy this episode with alex svetsky you've had a dynamic where money's become freer than free if you talk about a fed just gone nuts all all the central banks going nuts so it's all acting like safe haven i believe that in a world where central bankers are tripping over themselves to devalue their currency bitcoin wins in the world of fiat currencies bitcoin is the victor i mean that's part of the bull case for bitcoin if you're not paying attention you probably should be probably should be

whoo alex svetsky how the hell are you mr bent good i am about as tired as you are and uh and just a couple years older so is that what happens when you get older you get more tired pretty much bro it's fucked that's what i actually i got eight and a half hours of sleep last night woke up feeling like a bus hit you or no like i've had the most energy i have oh okay had in a while waking up well that's good that's good i um i got a long sleep and i feel like a bus hit me um i think i've been do you know what does that to you man it's international travel i i figure like you get bombarded with fucking all sorts of weird radiation you know your sleep circadian rhythms go out the window and i swear to god like for the last month uh in brazil it's been raining like

consistently i don't mean like you know rain and then some stop like the rain and stops rain like i mean like fucking rain non-stop like 24 7 rain nothing but rain and um and then like finally the day that it gets sunny i leave brazil to go to europe for bitcoin amsterdam and what happens when i land in amsterdam rain um finish that get to prague rain get from fucking prague to lugano rain and from fucking lugano i fly back and then the fucking rain starts again in brazil i was like you gotta be kidding me man so i can't be good like i it can't be good like i literally haven't seen sunlight in over a month so uh goes to show like if you want energy you need some sun you know you need some light you need to quit fucking flying around and uh you need to ground

yourself like all of these things that we talk about right yeah now i mean even just traveling here in the u.s over the last few weeks it knocks me out my ears all clogged my sleep's all fucked up till last night and then yeah it reminds me of my friend your story reminds me of my freshman year of college where i went to college in chicago and there was a 90-day period where the sun didn't pierce the clouds and it was the first time i'd ever experienced that i was like this is not good for your mental health it really isn't i guess that's why all the scandies like fucking depressed want to kill themselves over winter but anyway that's not they got other things to be depressed about now it's like you know literally they have no sunshine and now they have no energy either

really genius idea energy which is going to be a focal point of this conversation it's very important whether it's the sun creating photosynthesis for the plants that feed the animals that the humans eat or the high higher density energy that creates the power that runs our world today the world has gone insane it's really faded energy to a level of detriment that is material at this point and we're going to talk about that on the back half of this conversation but the first half is going to focus on an energy intensive industry which you've been diving head first into with the spirit of satoshi which is artificial intelligence obviously i think it's been exactly a year since the first what many would deem consumer friendly ai applications hit

the market we've seen a crazy expansion of these applications and people using them as individuals or as businesses a lot of noise out there a lot of posturing particularly by the incumbents in the space the open ais of the world the google's uh uh anthropic it seems like they're dead set on creating this regulatory moat and saying you got to regulate this industry daddy um we should be the only ones able to get these tools in the hands of individuals and businesses at the end of the day and as you've come to find in many others this probably isn't the best thing for humanity and i think this is part of the reason why you started spirits of satoshi and so i guess we start there

what is spirit of satoshi why did you decide to build it and what does it do sweet um thank you for that intro and that uh that layup so in in really simple terms the spirit of satoshi is an attempt to see if we can embed a bitcoin paradigm into a language model uh it's like the the long and short of it right so i did some talks in amsterdam and in lugano over the last couple weeks and in there i kind of presented this idea of look these these new language models they're not um they're not like general intelligence by any stretch of the imagination they're not they're not coherent they don't rationalize they're not alive you know they don't have any agency of their own despite all the talks about agency and this and that and like you know you

have all the nerds frothing at the mouth like oh my god we're gonna have like intelligent machines in the next year or two um i think you know first of all that is an unfounded fear um and then second of all um there is a a potential concern here with these tools becoming uh ubiquitous as a means of uh knowledge sourcing basically you know i think over the next couple of years we will we will use the internet in a different way we'll instead of searching for things on Google, we'll probably just ask our local agent, you know, a question and whatever it tells us, you know, if it's ChatGPT, you know, that'll be what we construe as the truth. So I want to kind of like tease out two bits here, and then I'll talk about why we're doing Spiritus Associates.

So, you know, the first one I already alluded to is this idea of AI, like many other things, just gets blown out of proportion right like people we we have such a narrow understanding of what intelligence is anyway like we still as human beings can't define it and you know when you look at you know so-called artificial intelligence um you know it sort of lives in the realm of you know what i'd say like this narrow sliver of computational intelligence right it can compute very well and a lot of this compute stuff is basically revolves around the of large numbers um you know it revolves around like running things you kind of like what asics do right like you know they are they are guessing things right um you know just to to a different

end um and to me when i think about like what general intelligence uh implies is you get like uh breadth instead of depth um so you know general intelligence includes computation but includes you know cognition uh i mean you know particularly with humans it includes emotional intelligence you've got hormonal intelligence you've got digestive intelligence you've got muscular intelligence you've got like you know the intelligence of like you know your wife knows like if you uh you know looked at a chick sideways by accident you know three weeks ago and you know she wasn't even there with you she's got this like crazy fucking intuitive intelligence like intelligence you know with respect to humans is like so broad and you know people are very

quick to take something like you know language because we haven't we haven't had anything other than human beings you know do language in the past you know maybe a parrot kind of came close to it but you know if a parrot said i'm gonna kill all humans we'd be like okay it's a fucking parrot right good luck when yeah you know when chat gpt says it you know people are like oh my god we're all going to die so um do you need a moment no no we're good okay um so you know i think the the first sort of myth that i try and dispel for people you know when we're talking about this ai stuff is that um agi i don't think is a thing honestly um i don't think it's coming anytime soon if it is a thing at all um and when you hear people talking about regulating ai

specifically what you look when you when you sort of dig under the skin and you alluded to this is it generally talks about what language or what you know it talks about safety you know the same kind of words that we heard during lockdowns and you know all of that sort of shenanigans it's like for your safety you know responsible wording you know blah blah blah these kind of things so it's important to sort of separate like the you know the real fears from the fake fears and you know agi as a concept i think is moot and to be honest i actually think the the joke is on us and the nerds is that what they are finding so you know we're building a fucking language model we're right at the at the forefront of this stuff is what we're finding in the language model space

is that the more general you try and make something um the less depth you get in any specific topic so you get less specialization um and you know the way uh some research labs are trying to solve this is by creating like a governor agent who manages multiple uh so think almost like a Merkle tree of some sort. So this governor agent manages multiple language models, which each specialize a little bit more. And this is how they, what is said to be how GPT-4 works actually. And you build that sort of a structure. So you have more general intelligence, but obviously you can't, there's a limit to how many specialized agents one governor can operate. so then you create a second kind of function like that

and then you have a, you know, a governor above them too and then you kind of like keep building this pyramid but what happens is you end up getting a situation where you use exponentially more energy for less and less and less like reward and you end up in a situation where the depth of specialization diminishes, the latency increases

And my whole theory is that by the time we figure out artificial general intelligence that is energy efficient will come all the way back full circle to human beings because we are the ultimate general intelligence and all we need to do to think, to run, to do math, to, like, build things with our hands, to do all of the things that you would – you know, that implies general intelligence. we just need to eat a fucking banana whereas like for a machine to be able to do all this sort of stuff it would need like all the nuclear plants in the world so just to sort of give you a sense of um the irony of the pursuit of general intelligence i think it's it's it's another poorly understood eschatological threat that is designed to free people out into giving a bunch of bureaucrats a bunch of power into controlling uh language more than anything else which leads into where i think the the fake risk is but before i get into that curious if you've got any thoughts on that as a as an opener why are the sam altman's of the world and others and then

you have the the ai doomsdayers so you have like the sam altman crew which is like yes we're going to bring agi to the market by 2030 it's a scary thing we think we need to do it so it's not as scary and uh what's sort of looking for apocalyptic as many people act make it out to be and then you have like the doomsayers who are like no we need to stop this now uh if we get agi ai is going to kill kill all of humanity like very curious to dive deeper into why you don't believe agi is possible is it purely like an energy thing like the amount of computing energy it's it's an energy and generalization thing so it's like it's a it's a mixture of both like so to get like um i mean sorry it's an energy and generalization thing but it's also a um an agency thing like there's

there's mechanics and there's agency um you know we we like all language models anything that we know in ai at this point is we give it direction we give it we tell it what to do it you know it performs based on something that we've we've given it an end for machines to have that that starts to imply things like consciousness which i went down the consciousness rabbit hole and basically here's the the sum total of what humanity knows about consciousness we don't know shit basically like there's been philosophers arguing about the thing for thousands of years like the only progress we've made is that nobody actually knows fucking anything and we know less than what we started with respect to what consciousness is as a concept and maybe you could argue that you know what we

found with consciousness is that you know there's something to do there with like free will or agency um but we have no idea where it comes from how it exists why it exists what it is or any of that sort of stuff so i think there's that the agency problem is a big one or the consciousness problem and i think it's very easy to conflate consciousness uh or agency as a concept with intelligence and i don't think they are correlated um or sorry i think they're correlated but i don't think they're the same thing and then the um the the the compute and generalization problem is that you know once again human beings are the in my mind the apex of general intelligence because we can do um you know we can remember we can love we can eat we can run we can build tools we can

compute we can do all of these things and we're pretty good at a lot of them but we can do all of that in this you know this incredible machine that we sort of live in um and our energy source is like literally some fucking bananas and some meat like that that is so much more sophisticated than you know what is possible with um with circuitry and metals and electronics and all that sort of stuff it's it's in my mind it's um you know human beings together um you know in a in a free market are far more powerful than um than you know some sort of machine uh alone now why do i think the same altmans of the world and all this sort of stuff you know go for that it's it's the same reason why you know the climate hysterics think that that you know genuinely

think that they're going to save the world by you know building fucking windmills and solar power plants and blocking out the sun like these people are deranged and you know they're sort of so out of touch with uh humanity with life with reality that they think the answer to all their problems is um is in some sort of circuitry and what they fail to realize is that there's downsides or there's uh there's not downsides is the wrong word but um there's trade-offs with everything there is absolutely fundamentally trade-offs with everything this is why like uh chat gpt sorry open ar just shut down arrakis which was their um their their second uh chat gpt project they thought that they could get the same kind of um uh performance uh with chat gpt for i

think it was i can't remember the numbers off the top of my head but it was like 10 or 20 percent of the energy uh output or whatever and it completely failed a year and a half of work down the toilet and they just shut it down because they're finding that they're reaching these um maximums because all these things are doing is they're just doing probabilities like it's no different like chat gpt is fundamentally not very different to how uber matches you with riders like matches riders and drivers like or how Airbnb matches you with the you know with places that you might be interested in these things are all very similar they're all probabilities they're all large laws of large numbers whereas I think human intelligence general intelligence

is something else and I don't think we are anywhere near figuring that out and I just think that the nerds and the doomsdayers probably sit in the same bucket as they overestimate stuff um you know if you listen to every single one of them every single one of them were fucking convinced that you know gpt five six seven you know by the time you know it comes out like you should like basically their projections placed agi coming out like roughly now next year you know the more sensible ones 2030 but we're already seeing these you know these um limits of computation being hit um and very little usage um coming out of these other than code assistance or you know what i like to think of as midwit obsolescence technologies like you know

being able to you know write a um you know uh a report that some cnn intern would write like outside of that there's not a lot of stuff here so anyway i'll shut up there for a minute um and we can dig into a bit of that yeah i mean with running with the assumption that agi is completely overblown in terms of if it's even possible and even if it is that you don't think it is like is it that detrimental to humans being able to survive on this planet next to ai i guess we can jump into again like this big debate that is beginning to surface particularly after sam altman zuckerberg all the big players in the closed source ai space went on capitol hill late this summer and essentially lobbied the government to put in regulations that ensure

that they have this regulatory moat and they are the gatekeepers to artificial intelligence applications like i think it's becoming very clear that we're at a very important inflection point early on in this technology's proliferation where we need to get out in front of the narrative that sam altman and his crew are pushing and really drive home the fact that there are two sort of models pun intended of the ai future this closed sourced model where you have these centrally controlled companies running everything in the black box and then this open source world where you can basically leverage these open source models train them yourself and put out your own artificial intelligence applications and if sam altman and his crew get their way

it's not going to stop people from being able to do that but it will likely render it illegal and push it into this gray or black market what correct how would you frame this battle between open source ai and closed source i think yeah i think you touch on the most important point here and this is why i believe agi is a red herring is if you look at the wording in these calls for regulation what they all center around is responsible language safety filters um and you know things like toxicity filters and uh but basically the the words that come up a lot is responsible ai uh safety um filtering toxicity all this sort of stuff and here's where i believe we've we've had a paradigm shift is i don't think we've gone anywhere fucking closer to agi

In fact, I think all we've just proven is that 95% of language is just patterns. That's it. That doesn't mean anything. I don't think that has anything to do with AGI or consciousness or the larger general intelligence narrative. But what we have opened the way for, and this is where I think there is a fundamental paradigm shift, is the age of the language user interface. And what I think about is, you know, we went from obviously like command line, you know, then we had the GUI come up, you know, and people sort of used the, you know, the graphical user interface on computers and that changed the way they used and spoke to technology.

You know, we then moved into mobile, you know, and mobile has kind of taken over as the primary way we use an interface with screens. And I believe the next phase shift is this idea of you won't use your thumbs and, you know fingers to tap and click and do all that sort of stuff anymore you just talk to the thing and the language user interface will interact with you and respond to you and what i kind of said earlier is like you've got a question about something you know today we all go to google and we you know search for something and you know most people not only do they not go past uh the the first page of you know the google results but they don't go past like the first or second

result anymore right so it's like whatever google tells them that is the truth now insert language user interface into where google sits today that is where i believe the risk lies is that if people's entire model of the world what's good for my health what's happening with the environment a solar panel is good what about energy um you know should i live in a pod like all of these questions think about it like all of imagine everything you know and learn coming from chat gpt and chat gpt upstream is basically regulated to be safe and responsible um and you know the the thought police or the language police uh basically yuval harari sam altman and and klaus schwab right that's essentially where the danger is is that every single application

every single product is in some way going to have some sort of language or conversational feature embedded into it and everything you know and do the way you interact the way you communicate with every single product and service and particularly search online and information gathering is all going to be informed by a set of safe and responsible language models and if you think about like what that means like that that's the ultimate form of inception like if you want to entrain an idea into somebody like you answer the questions that they have and that is the paradigm that you're feeding into them it's like plato's cave uh but you know using ai to make it uh you know to make it the thing that they uh absorb from their screens so i think this is where um

this is where the real risk lies and when they wave the the red herring of our agi it's going to kill us all ai if we do not if we do not regulate if we do not make this safe if we do not apply toxicity filters if we do not you know make responsible ai the world's gonna end so we need to do this and this and this um you know even these days like if you want to train your own language model if you want to fine train all of the tools out there you know come with toxicity filters they come with like you know language filters the libraries that you train the base models on they all come with libraries that have been filtered and made sure that they're safe and hygienic and all this sort of stuff like all of this sort of stuff is pointing to basically a

homogenous paradigm of the world and this i think is where the real risk is and this is why i think it's fundamentally important to build a model whose you know pun intended model of the world is is different you know that represents a bitcoin model of the world that you know energy is good energy usage is good uh you know inflation is fucking bad like you go ask chat gbt about inflation it'll tell you that inflation is a sign of a healthy economy like we know that that is wrong if you ask alarm it doesn't matter if you ask an open source or closed source model it doesn't matter every single model out there says inflation is a sign of a healthy economy because that is the paradigm that has been entrained in all the data sets that these models have been

trained on. And that's what we need to, I think, offer an alternative on. And, you know, I guess that sort of brings us to why the hell we decided to build Spirit of Satoshis. We want to train it from scratch so that by default, it represents like it answers from a different paradigm. It answers from a different model of the world. Yeah, I'll stop there for a moment. But I think that is where that's like the difference between the fake risk of like this eschatological AGI is going to kill us all we don't understand it but because nobody understands it let's freak everyone the fuck out kind of like you know the the sun is going to kill us all tomorrow right um or climate climate change global warming right you can't quite touch you can't quite prove it but it's

scary enough and we're all going to die so that way we need to regulate it agi is the same thing whereas the real risk is right here it's fucking facing us is all language being regulated by a small group um and if that's what happens then you know you you have a whole world being trained on these what i believe a false set of ideas yeah it's funny how the language of it all large language models trying to dictate language it's all right in front of you um it really fucking is when this shit hit me like i remember reading the um there's There's a great article by a guy called Mark Bison, which is called The Trillion Dollar Tin Man or something like that.

I'll see if I can find the link and send it to you, but this guy is like a fantastic writer. And he's like a kind of like old school libertarian. Like he's kind of in his, I guess he'd be in his 40s or something like that. But dude, like this guy writes like a scalpel. Like when you read it, like each paragraph is like a punch in the face and he talks about how you know this this sort of call for regulation is like a complete like red herring i remember reading that and i was like holy shit this guy's a fucking legend and i went down the rabbit hole and started looking at what all these regulatory pushes were and every single one of them was to

do with regulating language literally that's all it is nobody knows because they don't know how to define agi they don't know what the fuck it is nobody knows what it is but it's just like it's just out of touch it sounds just sci-fi enough that people will be like oh shit you know i don't understand it's kind of like when vitalik talks about ethereum right people just lap that shit up because they don't understand what he's fucking talking about yeah people don't want to feel people don't want to say that they're dumb that they don't understand it so they're like oh yeah you know it's an existential risk we have to trust the experts cover me harder daddy i guess this is this is a good opportunity to dive into like building your own language model

that isn't sort of filtered using the filters of the incumbent libraries or the closed source models like OpenAI? How do you even go about this? Dude, it's been a journey. And I must admit, it's been more difficult than what I thought. So the initial way we went about it, and there's a couple of Bitcoin models out there like ChatBTC and PlebAI and all this sort of stuff that do something similar,

is you take an existing language model, and whether that's ChatGPT or open source or whatever, it doesn't matter, you take something and you simultaneously build what's called a vector store where you take a body of knowledge. It could be a book, it could be a series of books, it could be some transcribed podcasts, all this sort of stuff. And what you do is you basically break it up into chunks

and each chunk should be as semantically coherent as possible, meaning that like if you read it you kind of know what it's you know going on about and then you store those in what's called a vector database and that's merely a database that doesn't store the the words it stores a represent a a numerical representation of what the words mean right so like it lives in this like multi-dimensional vector space so like this piece of context this chunk of text means something. And what that does is it makes it readable by the language model. So, then what you do is you put a wrapper on the front of that. And when somebody asks a question to this model, it checks whether the words in your question relate to something in the vector store.

And if it does, if there's what's called a similarity score, you inject that piece of context into the response from the model and the model will respond with that context and you will get a more relevant result or a more relevant response and you know this this was how we built the first model we built the first spirit of satoshi beginning of the year like this play by eye works like that chat btc works like that you know there was another one i think it was ask satoshi or talk to satoshi or whatever so you know that's like the low-hanging fruit way of doing it and a lot of in fact whenever you hear online you know people talking about train your own model this is exactly what they're doing and the the problem with calling that training your

model is it's actually it's not training it's got nothing to do with training you're not fine tuning you're not training you're not doing anything all you're doing is you're referencing uh data in a vector store um now that's useful insofar as you know you want to use an underlying model to get a little bit more accurate or more context specific or context relevant responses but then you know as soon as you ask a question that is not in the vector store basically you know when there's no similarity score there between your question and what's stored then the model just you know goes and does what it normally does the other problem that you also get is no matter what you have in the vector store it will always be the the final response

will always be influenced by the underlying model. So in our case, ChatGPT was, you know, OpenAI was the underlying model. We also did one with the Lama underlying model, but honestly, ChatGPT performed much better. And, you know, the problem we were getting, as soon as you start getting into the thorny questions around, like, energy, around inflation, you know, around, like, you know, deflation as well as another one that ChatGPT really fails at,

you you would just like touch apt would transform the answer into something it's like no that that's like that is fucking wrong that is not like the bitcoin model of the world so the alternative that you can do is you can actually go and try and take a pre-existing open source uh model something like llama something like mosaic something like falcon you know something like Mistral. There's a bunch of them now. And you can try and change the weights and biases, which are basically the parameters of the model. And what these parameters do, they are basically the governing components as to the probabilities of what word comes after what word, right? So like to give people some context of what that means, like language models are

basically sophisticated autocomplete. If you've ever used Google and, you know, Gmail kind of like finishes your sentence or, you know, when you type in Google and you type in San Francisco and you get like the rest of, you know, the search term. Like that's what these models are. They use probabilities to guess the next word and they are sophisticated enough. They can produce a whole sentence, a whole paragraph, a whole stream of text that is semantically coherent because of the relationship between words in their neural net. So when you are going and you're retraining a model when you're fine tuning a model you're changing the parameters you're changing the weights and biases applied to the different words or to the relationship between

the words so that you know instead of like let's say well I'll use our case as an example when asking about something like inflation instead of the probability being skewed towards you know the output being inflation is a sign of a healthy economy so long as it's around two percent the probability with our model will be something like inflation is bad because it's a decrease in people's purchasing power right that is because the content that we are feeding it and using to retrain the parameters has more of those words in that order talking about inflation bad inflation is you know uh stealing purchasing power inflation is there so all that sort of stuff it's got more of that than it does inflation is normal inflation is healthy inflation is you know hitting the two

percent mark or wherever the fuck else is in mainstream content sources so that's like a proper way to to fine-tune but even that then has its limitations and this is what we're finding out now like our model already by default like we did a funny um one the other day we asked it like i'm gonna have to read this out to you you're gonna fucking laugh at this uh satoshi on biden so we asked him you know who is the president and you're gonna laugh at this shit it said satoshi goes the president of the united states is joe biden he is a puppet of the globalist elite and a tool for the federal reserve to implement policies that will eventually lead to the collapse of the u.s dollar and the rise of bitcoin

I mean that doesn't really make sense like it makes some sense obviously like but what it is you can see like it's taking the probabilities of the kind of text of the kind of content we're feeding and it like spat out that we're fucking pissing ourselves laughing when we saw this because like this is what we're doing we're building something that like represents a different paradigm or a different model of the world um and like i'll stop there for a minute and because there's one there's one more step um but that's a that's a bigger step that we're not ready for yet but here's the well i was just gonna say when you at this point when you put it that way it makes it pretty clear that something like agi seems ludicrous because at the end of the day

the answers to these models are wholly dependent on the paradigm view of the individual building the model or the individuals so could these like similarly to how certain individuals trust certain publications whether it be cnn or something like zero hedge is that how we're going to have to view approaching and using models is like hey i i sort of agree with the worldview of the people who built this model so i'm going to trust there is an element of trust it's not true i mean there can be knowledge gleaned from it but it's not the quote-unquote truth at the end of the day it's the it's the product of the data fed and you're essentially trusting that the people feeding the data like agree with your worldview

propaganda tools at the end of the day exactly exactly that and that was exactly my talk in amsterdam i said we're gonna have a propaganda war over the next 10 years and basically all these models are just going to be spewing out propaganda like that that's and it's not because they're evil or this or that or whatever like we can even take away that component of it it's just that if you take wikipedia which is a big part of the base of all closed source and open source language models is basically the the wikipedia index and what the fuck does the wikipedia index say about energy inflation solar panels like all of the things that we rail on about it says that So what are the models going to do? The models are just going to mirror that. You don't even have to do the filtering. Now, what these large scale companies like ChatGPT, et cetera, do like OpenAI, they also add a filter at the end and they try and make sure that it doesn't say anything toxic or this and that.

because if you remember like in the early days chat gpt was a little bit more wild and that was just because like you know it was trained on such a large corpus of english language and somewhere in the english language in fiction books and in like you know literature and all this like there was go fuck yourself and this and that and all this other stuff right so like you know the it's incredible what you'll get out of the probabilities from word associations and you know relationships like which is why like you know the the biden one that we just pulled out like you know his globalist elite you know tool for the federal reserve i was like where the fuck did that come from but like you know this is sort of what these things produce and it's very easy for us to

anthropomorphize like oh you know i thought of that well it's like it didn't didn't actually think of anything you know it just this is what you know the the it's like a lottery this is what the lottery produced when you you know gave it a particular input and you know it ties back to like if you if you don't have an alternative um you know you you the world's going to end up just being fed shit um which is representative of the old paradigm the old model of the world and you know i think if if anyone it's incumbent on bitcoiners as a as a community as a as a group of people who have kind of um i guess we've transcended the old form of thinking right we we operate we think we behave we want to live on a different trajectory or something else and it

reminds me of like uh einstein's sort of saying right it's like the the the kind of thinking that brought you the problems of today is not going to be the kind of thinking that you know solves them or something like that that's essentially what this uh this new approach for building language models is it's like we want to think differently worship we need to entrain the model to to operate from a different uh set of principles so so that's kind of why we're doing this yeah and diving deeper into the mechanics of it particularly how you feed it data and when you know when to stop feeding it data because we were at a 1031 event a couple weeks ago and we had AI open Socratic discussion, it became very clear that data, quote unquote, cleanliness is very

important. And there's a point of diminishing return of value of data at some point when you're feeding these models. Very much so. And then this is something we've like, so I'm going to give you a couple of examples. So when we first started fine tuning the model, we took, you know, we didn't know shit about what we're doing we took the whole bitcoin standard we broke it up into i think whatever it is it's like 15 chapters or something like that and then we just fed the whole chapters to the model like full chapters then we took you know 21 lessons and like basically all the bitcoin books all the big articles and we just like fed these long chunks of uh data and we're like okay this should be sweet you know train and you know when what we got back was the model would uh you

ask it a simple question and it would just stream these long fucking responses that would just like you know loop on themselves and basically say the same thing over and over again but just like really long and what we realized was we actually didn't train the model on the data we trained it on the length of the data and the style of the data so this once again and whenever i talk about this stuff you know people sort of uh realize that language models are a little bit less magical than what they originally thought is you know when you're doing this training you're you're training it more stylistically like language models don't actually know anything it's just that like and this is another epiphany that hit me and i had to i'm going to do a tweet about this

but you know how they say language models hallucinate like 20 of the time well here's the truth language models hallucinate 100 of the time but 80 of the time the hallucinations are accurate and that's like the big like when you make that shift you start to realize that everything that llama spirit associate chat gpt anthropic everything that comes out of these things is all made up it's just that the the the degree to which you train them you know the the degree to which the the data is um as clean as possible um determines you know the degree of the the good relationships you know the accuracy of the relationships between the words and the and the parameters and that will help influence the accuracy of the hallucinations that come out of

the model and you know this fundamentally ties into what you said about like data quality is you know we did that first round with these long texts then we shortened it big time and we just did short paragraphs and then when we trained you know the second version of the model it wasn't very good at answering questions so we would ask it a question and it would try and complete sentences instead of like answering the question what the fuck so then we realized no no if you want to train a model to answer questions you actually have to give it examples of questions and answers and then it starts to learn how to answer questions so then what we had to do is instead of like taking just the bitcoin standard and just breaking

each individual each individual chapter into individual paragraphs we actually took every single individual paragraph and we converted it into a question and answer pair and then we fed the question and answer pairs to the model and then we started getting better results the model actually started producing responses to questions and the response lengths were pretty good but then we ran into another problem which is all it then did was just answer questions we asked it to finish sentences or to do other things it was really bad at that so you know what you end up doing is you change the weights and biases too much with too much of a skewed kind of data then And you end up, like they call it forgetfulness, which helps to conceptualize it,

but it's not like that it remembers or forgets anything. It's just you've changed the parameters to such a degree that this is what it's good at doing now. So what you then realize is you actually need a good data blend. You need question and answer pairs. You need short chunks. You need long chunks. You basically need this whole well-rounded set of data variations so that the um the model can do multiple things and then that's when you start to realize you

also need a little bit more size because the more parameters your model has um the more of these things uh it is able to do um and then there's also things about like um you know if you have i think it's um the number of tokens of data um that you feed a model so there's all these like you know variables and to be honest i was listening to a podcast the other day about like the guy who um basically invented the the state of the art which is you pre-train a model then you fine-tune a model then you do reinforcement learning he kind of created that um that flow and basically all language models out there sort of follow that flow he did a podcast about a week ago kind of saying like you know what i retract my original statement and yeah sure the whole

industry is kind of built around this idea but i actually no longer believe that um we know what we're doing um because um it seems like we're not like we've reached the point of diminishing returns now the models aren't getting significantly better and we're all following the same thing i believe now that the process is continual pre-training um you know less about this fine tuning stuff is like it's almost semantically the same shit it's like continual pre-training and then you you do some final reinforcement learning at the end when you want to kind of adapt your model to a specific use case so so there's all this sort of stuff and like what what it reminds me of is that you know this whole industry is fundamentally it's it's as much art as

it is science in fact it's probably even more art than it is science because you you know you take this whole data blend you know you you you put it into train you run eight epochs and you come out of the other end you're like i don't know what the fuck this thing's gonna do on the other end and it either comes out good or it comes out bad like we we actually our model when it was operating in bitcoin amsterdam was fucking fantastic like it really ran well we did one more fine tune before Lugano a week later and the Lugano one was shit like it wasn't like you know not answering anything but just the the answers for example like lost the humor that we had embedded in it uh in the first one and we're like fucking hell like you can't win with this thing we thought we were

gonna make it you know incrementally better for the um for the event so anyway this is a long way of saying like data is like critically critically critically critically important and then the type data the data like or let me rephrase this data is the most important thing and then what matters when i say data is not just the data quantity but more important the quality the blend and what that blend is made up of and what you want the model to actually do um is going to determine what that blend looks like if that makes sense yeah adding a bit of context into it which is like here's how you should probably answer these questions and this gets into a broader question too so another thing that came up during that

socratic discussion at our 1031 retreat is again juxtaposing open source ai versus the open ais of the world and what you just described it seems like the big large language models are running into this problem where they're hitting a ceiling and not really improving as much as they have over the last 12 months and that gets into the discussion of the world of i believe it's a thousand minds or ten thousand minds versus um the world of the borg which is open ai running everything and like how do you view uh the future of ai either built by the borg or a thousand minds which would be thousands of individual models like Spirit of Satoshi that do one thing very well

and then you combine them with agents to actually produce complex answers. Totally, and this is exactly where we're pivoting our business. So, you know, we're looking at a name now that kind of exemplifies this idea of specialized AIs because we've gone through this whole process of trying to build a specialized AI and we've had to build all these, you know,

like taking the Bitcoin standard and transforming it into 2,000 Q&A pairs like that has to be done programmatically so we had to build like all of these fucking pipelines which involve multiple models like manipulating the data and then we have this final piece where human beings check random bits of data and they get paid in sats for validating so we've got like a trust don't trust verify section in our reinforcement learning tool so we have this whole pipeline which takes data in one end and transforms it out the other into a format that you can actually train a model on this shit doesn't exist yet like we had to fucking figure

all this crap out ourselves and we're like man we could actually commercialize this pipeline as a process and help any industry company body content creator whatever actually train their own specialized models because i do believe that we are going to have a world of many many many many ais as opposed to general ais which actually ties back into what we said at the beginning why i don't think agi is a is a risk or a threat or a problem is that the more general you get as i said the less specialized you can get um you know having a marketplace like this is why there isn't like you know the the world is made up of seven or eight billion human beings all with their own will you know all with their own praxis and needs and wants and the market is how we all interact

right and that is the most efficient and effective way to you know for example build a pencil right we all know this as bitcoin it's like you know the pencil is made because there's one person chopping trees there's another person mining graphite there's another person fucking doing paint like all this sort of stuff and we get this pencil ais will function in exactly the same way because it is far more efficient and effective to do that and having you know trying to build like a Borg, like you said, that manages many AIs is going to run into its own bureaucratic problems in the same way as a communist state runs into its own bureaucratic problems because it can't adapt it. You know, it's not a market force. It is a centralized force. So I think

fundamentally that's the way it's going to go. I think general models, like a good way to almost think about this is generalized models might still function like Google, right? Like if you want to get like a general direction of something um you know a general feedback or just something like good enough you want to ask a question you probably go to a chat gpt like this is not to say chat gpt is going to disappear like you still go to google to get some basic information but when you want to dig you will go to a forum you'll go to someone you trust you'll go to um you know you'll go to like a reddit thread or something like that and you'll go down the rabbit hole and you'll get way more richer information, but you're not going to get the breadth of

knowledge. So like specialized models will be like the forum that you would go like an online forum and, um, chat GPT will be like your basic Google search. And I think this is sort of how the, um, the industry is going to evolve. So it's pretty optimistic, right? Cause everybody's afraid that open AI. And so you think they're gonna fail to achieve what they're posturing they can achieve just because the limitations of totally man totally totally totally it's the it's the same fucking thing you know they they they like the the risk as i said that we face is that um but i just think the cat is too far out of the bag for that for them to stop it but like as i said the the risk is that um

you know companies have to use like licensed ai models and those licenses have been approved by government bodies for use in commercial applications right so if you are a business in europe for example that you know is doing you know uh selling some sort of health foods and you know your customers have questions about health foods and you know you might want your own language model that talks you know about those health foods um but you know you're not allowed to use your own language model or you're not allowed to use it because you know from xyz company that helps you build your own language model you have to have open ai or some sort of like woke subsidiary be the licensed entity that helps you build your language model because it is

considered safe and responsible so so that could be you know the risk um but uh in terms of like you know them being able to functionally build like the ultimate ai that runs and controls everything i think it's a fucking it's a nerd's pipe dream man yeah no and it's very important too speaking of the nerd's pipe dream i mean mentioned him a lot but i think he's somebody that should have a lot of focus on and people should be very skeptical of what he's trying to do it's just sam altman because yeah he's got open ai apparently he doesn't own any of open ai he's doing it altruistically but he has world coin which you can see that he wants to probably leverage that to gain access to his models and pay and that'll be how he monetizes open ai if

he can successfully force world coin on the world and then a lesser known company that he's involved with that many people don't talk about was like a nuclear um company like the nuclear energy company where you can see he's beginning to position himself like all right i have the ai i have the digital panopticon cryptocurrency and then there's this other company i forget what it's called logan if you can look it up but it's a nuclear energy company where he can sort of just get to the base of everything and if you control those three things it's a really scary future yeah i mean you're 100 right uh i just think the you know always the irony of uh of these kind of people is that they um they i don't know they underestimate the the market they underestimate

how nimble you know smaller competitors uh can get and like i mean just just the the the best thing i guess about the the ai space is also the worst thing is that um you know i saw a stat you know people say ai was sorry chad gpt was trained on the whole internet and like the that is like so far from the truth it's not even funny i think the the scale is something like the scale is something like the amount of data that chat tpt was trained on versus how much data there is on the internet it's kind of like el salvador uh as a country and the sun like it's like that like the the the or it's like 20 orders of magnitude or 30 orders of magnitude difference like there is always going to be more data than there is able to train on and what that means is

it's impossible for one company to gather up all that data like you know just just think about like um i mean think of just like one fucking law firm like that specializes for example in like one niche thing i don't know like uh give me a topic like you know a law firm that gets people out of traffic traffic tickets right like most of the the data most of the crap that they have like is not going to be stored online anyway it's not going to be accessible to open ai it's not going to be accessible to anybody else like you know these guys if they wanted to automate some of their processes for example could get an an agency or or a um you know or a practitioner or someone like that to come in and help them build their own uh language model

or another good good example that i uh that i have it's going on in our new pitch deck is recruitment agencies what do recruiters do they spend most of their time sifting through fucking resumes then doing first interviews second interviews third interviews trying to figure all this stuff out they're sitting on a mountain of like valuable data which think you could almost think of it like oil that's locked inside the shale and we don't know how to get it out and you know they can't in its raw form train a model on but if they could somehow run that data run all those resumes run all that feedback run all those interviews run everything that they've got through some sort of process transform all of that and train like the ultimate you know technology

recruitment agent for their specific vertical or their industry like they could automate all of their stuff now open ai is not going to be able to do that like open ai like chat gt might be good at you know reading some resumes they can feed it a bunch of general stuff but it'll never get good at that specific vertical and which means that i think you know the the marketplace of many many ais many many players uh doing things you know building ml ops um which is essentially what we're going to try and do is um is going to just out compete like they just can't fight a you know a hundred thousand front wall which is essentially what this um this ai proliferation is going to be so that's why like i'm not i'm not too concerned the only thing that they can pull off is the

licensing um but i think as i said i think we've things are moving so fast that it'd be very hard to get like licensing in place now but you know i mean i guess i could be wrong let's see the executive order and i put in slack before we jumped up control f for license so biden came out with a executive order on safe secure and trustworthy there you go there's those words right no license all right no license in the executive order that's good but yeah i mean we can scroll through this i think i briefly skimmed it before we hopped on but um i think i'll pull it up on my computer too so i can see it but it seems pretty bland just got to make sure it's safe there's some woke

talk in here. Promoting innovation and competition. Setting up for consumers. Supporting workers. Advancing equity and civil rights. Making sure the government uses AI responsibly. So,

first executive order from the Biden administration came out this morning, October 30th, but no license and learning app, which is good. I didn't see that. Yeah. Okay. Okay. There you go. So, I mean, you know, this is, this is a perfect example. Like this is, um, you know, it always starts slowly, slowly, and then they kind of lean into it, but you'll, you'll hear it echoing these words all the time. Safe,

responsible, trustworthy, like, you know, all the words, like when, it's the classic Streisand effect, you know, when the government tells you they want to do something for your safety, you know, exactly. It's the opposite thing, bro. Yeah. Yeah. but diving into like the future of a thousand minds like how do you think this will work what will the end product look like like agents interacting with each other and then i guess we can tie bitcoin into this too yeah like does bitcoin over the lightning network using something like macro and permissions to give these agents allowances is that a use case that you see as

Possibly. I mean, I think machine to machine payments makes a lot of sense. I mean, I think we're quite away from that. I was just listening to, there's a really good podcast called Latent Space. And this is some guys, they do like some, they basically do a weekly podcast and it's pretty damn good. And they did one recently with this lady who was, she was was previously i think head of product at dropbox and um and she they just raised 200 million dollars for for their startup and they they're focusing on building agents and the whole podcast was around how like agents just don't work yet um and you know there is a lot like if you look at twitter for example and you look at you know ai agents you'll think that the future is already here we've

already got agents but you know you listen to the people actually at the very very very tip of the spear on this and they know that like we are so far from agents that like it's not even funny like we don't we don't even like these things are practically still macros like you know we haven't really gone too far from excel macros um but there is progress being made and you know the assumption is that at some point in the next five ten years you know maybe every brand like i kind of think of it this way it's like in the 60s every business kind of adopted the telephone right and you know in the 90s they adopted the you know the cell phone you know in the 2000s they started you know every website started having a sorry every business started having a website and then you know in the

2010s it was a social media page you know everyone had a facebook and an instagram and something like that and i think you know over the next decade like every brand every company every you know content creator influence or whatever is going to most likely have some sort of avatar or agent that represents them, that does some stuff for them that either filters noise or, you know, executes, you know, emails and deals with like the noisy stuff and the menial tasks. And maybe even like goes out there and performs economic functions on their behalf based on what you tell them to do. And I think, you know, there'll be probably some future where people who you know are really good at like you know researching and finding gaps in

the market and then able to direct a suite of different agents to go and perform different tasks and you know they might build like uh small one or two man startups you know that really have a lot of the base work done by agents like i i kind of think that that's where things are going ago that doesn't mean that these agents themselves have any agency you know we as human beings as the directors of the task have the agency and the agents are maybe just more like you know executors of specific tasks that you know have some sort of like semantic element where they you know can understand things they can read on the internet etc right and you know So I do believe something like that will happen.

I think to get there, there's a number of things that need to occur. First, compute needs to come down significantly. Computer is way too expensive at the moment for that kind of a future to exist today, which means simultaneously, not only do chips need to get better, I believe in the next year or two, we'll probably see AI A6 come out. But also energy needs to get cheaper. And there's something we'll talk about, obviously, in a minute. But then also, thirdly, is the frameworks and the pipelines to take all of the data out there and actually transform it into something that genuinely tunes and trains a model like around a particular topic or brand or persona or character to a high fidelity degree.

Those pipelines don't exist yet. And as I said, that's something we as a startup are trying to build based on what we've learned from trying to build Satoshi as a model. So all of these things sort of need to start to emerge before we can have that kind of future. And to kind of tie that all into Bitcoin in some sense is that transforming that data and doing all of that sort of stuff is going to require compute, is going to require multi-agents, it's going to require chains of language models and AI programs essentially helping transform and manipulate. it's going to require human feedback and across each of those uh components of the chain having an incentivization model uh having some sort of uh economic component i think is going to be essential and this is where something like bitcoin obviously comes into play and you know we've got a very rudimentary version of it like if people go to train.satoshi.com sorry train.satoshi

for fuck's sake train.spiritofsatoshi.ai um they will be able to see like how we use it like you can go in there you can do like the whole trust and verify you can um you can answer like satoshi you can rank uh satoshi's responses and all this sort of stuff and you get paid in sats for your contributions like this sort of stuff is gonna proliferate and i mean what what better money to use what better economic incentive to use than the you know the money of the internet you know money that is energetically you know sound and conservative like obviously bitcoin fits into this so well i just i just caution on people sort of like jumping the gun like oh yeah you know we're going to use bitcoin tomorrow for this like this there's a bunch of other things

that need to come into play before more of this happens um which is what i'm thinking on i think this is going to be a decade uh process not uh i think we're going to go you know we've had the gardener hype cycle right i think next year and the year after are going to be a bit of a trough for ai like there's going to be a bit of a uh you know all these massive expectations blah blah but you know in that trough is going to be where a bunch of the real value is built like these pipelines new chips all this sort of stuff and over the course of this decade i think we're to see some really interesting things uh be built out i was told we were one to two years away from all this yes it's going to take longer we were we were one or two years away from million dollar

bitcoin as well apparently that's actually exactly what i was going to go into because it is fascinating when you like i think bitcoiners have a really unique perspective on this particularly bitcoin miners um seeing the hype cycle and ai uh particularly around gpu compute and its energy intensity and how expensive the gpus have gotten miners are looking at that like hey we've seen this before um it's very similar to the trajectory of asics over the last five ten years even and i guess that's the big question in the space now like how much of this ai stuff is hype i think many people would say probably 99 in terms of commercial applications um and then yeah like the the gpu commodification maybe moving the basics that's one thing i did want to touch on since you

mentioned it is obviously asics and bitcoin uh proliferate because they do one thing one thing only which is hash cash shot 256. uh help me understand how this would work in ai and training models like because asics do one thing one thing only aren't ai models using gpus because they do multiple things or would you have specific asics for specific parts of model building and training yeah this is a little bit outside of my technical scope to really explain properly but conceptually i think i understand is that when you are when you're using um a gpu to do a training um as far as i understand and i don't understand much here

but it's like the the single task is um is something to do with like uh word relationships so so it's like it's it's still computing it's still like a relatively narrow task and you know the the limiting factors around like ram and how much um you know because like what these models live on is it's like it took me a while to wrap my head around it in the beginning but it's like when you have a 30 billion parameter model like it requires an amount of ram or vram from a gpu for it to operate or to run like it's not like traditional storage like it doesn't get stored on like a normal database for example like it's got to be operating on a gpu and i think what's going to happen is in the same way as we did with Bitcoin, we're going to figure out

what those, what those particular elements are. And I believe there'll be probably ASICs for training and there'll be probably separate ASICs for inference. And, you know, there's going to be, you know, whatever element of the GPU is most important for, you know, running those specific tasks, I think is going to be refined and we'll have application specific integrated circuits. will have ASICs for specific AI tasks. I think probably training will be the first one that they crack and then inference later because that, from my understanding, is a little bit more broad.

It'll be interesting to see how NVIDIA survives that because they don't create, they don't build ASICs, right? They're purely GPUs right now. I'm pretty sure, yeah. Yeah. The hot chick on the block may get left behind. No, going with the trend too, you can argue that we're reaching the point of ASIC modification

in bitcoin bitmame with their s21 series price that really low at 14 terahash where if you go back to like the s19xp when it first hit the market it was 40 terahash and quickly went up above 100 uh during the 2021 bull market who knows the 14 terahash for the s21 is just a product of where we are in the market and the fact that we're in the bear market but many people seem to believe we're getting closer to ASIC commodification in Bitcoin. And you have to imagine that you're going to see a similar trend play out in AI as well as the economic incentive is just to be more and more efficient. That's it.

That's what we do. And this is the thing. It's like, you know, to, I guess, to tie this conversation out to energy. It's like the energy is the one thing you just can't fake. Like you just can't print the shit. I'm sorry. You know, you can't like, you know, something uses energy. It's like there's no way around it.

and the the forcing function is to use the energy smarter better you know more efficiently more effectively that's it like you just can't there's no way around that no and it's my again i think bitcoiners particularly miners have an edge in this particular part because again the ai ai computation is extremely energy intensive there's a number of things that I'm really looking forward to seeing how they play out. Number one, narrative-wise, does AI get hit with the energy narrative as hard as Bitcoin has or does the fact that NVIDIA, OpenAI, Facebook are all in

and a lot of the heads of these companies are pretty close with the government. Do they sort of get priority over Bitcoin or do they get the same scrutiny that we have historically or do they become an ally? Do they help shift the narrative, which seems maybe not being led by in the ai industry but it seems like we are having a shift in the energy debate about like hey energy is pretty fucking important it's been made very

clear the last two years particularly do they become allies and say hey no we need to build out as much energy capacity as possible and then another trend to pay attention to is the competition for rack space um i know there's a ton of mining companies that are getting courted by uh ai companies looking to host gpus to to sort of prioritize gpu rack space over asic rack space at the moment bitcoin a6 interesting and there's uh i didn't know that okay yeah rack space i mean we learned this after the china ban it's very tight in the united states still is people are building out a lot more capacity but there are limitations and you have this i guess people could call it a black swan of ai really taking off particularly in the united states within the

last year and that competition for rackspace has gotten extremely more competitive um and so it'll be interesting to see how they play together if the ai industry becomes an ally or tries to box the mining industry out particularly in the united states and then the economics of it all and there's nuance to the different types like i believe that you're going to have co-located ai and bitcoin mining farms because if you want cheap energy we're finding particularly here at ercot and other parts of the country in the united states that you need to be engaged in demand response and from what i understand ai particularly if you're doing model training can cannot suffer downtime and so if they want cheap energy they're going to need to

find a partner that can participate in demand response which is bitcoin miners and so you'll have these sites that are co-locating gpus for the time being maybe a6 in the future for ai and then bitcoin a6 you get a lower cost of energy when you're asked to respond to demand the bitcoin operation spins down, sends electricity back to the grid and the AI operations allowed to operate continuously to train their models. I mean, that would be, that makes the most sense. It's just, you know, you would hope, you'd hope people aren't stupid and like the, because the alternative is like, oh, you know, we need to ban the Bitcoin mining because now we can't run our AI. And if we don't run the AI, China will, and China's going to beat us. So therefore,

uh we should turn off bitcoin mining and make uh save save the uh the limited resources we have for our ai compute instead like that could be the the bad version of it yes and that concept limited resources is the most frustrating thing because they don't need to they're artificially limited limited exactly yeah via via government mandate um and it's not even only government I mean, ESG is dying, but it has done a very good job over the last decade of disincentivizing the build out of reliable energy infrastructure, which is the big issue coming out for the Bitcoin times this fall. the energy issue the energy debate is raging it's very important to bitcoiners and so i guess we can begin diving into what we're trying to accomplish with with the uh the bitcoin times this fall yeah for sure man so um yeah let's uh let's segue into

that so um i think i mean for those who don't know i think you and i probably did a episode i think we did the austrian edition last year right when uh when that came out yeah so you know for those who aren't uh privy to what uh this is the bitcoin times is something i do once every year and i try and get uh six or seven uh essayists or thinkers brains uh in the space to write around a particular topic so last year obviously we had say fit in pierre goldstein raheem parker and um uh conrad graph right on basically bitcoin through an austrian lens and austrian austrian economics through a bitcoin lens and that was like it aligned with the first time i actually printed these things as a collectible and uh this year when i was thinking about what we were going

to do i was going to do like a you know i had options to do a philosophy edition a um a macro edition markets edition privacy editions cypherpunk edition like all these sort of ideas but it just felt right considering the the climate hysteria the esg stuff that's going on um you know the mining talk and and probably also like a personal interest of mine like i've been going down the nietzsche rabbit hole and sort of nietzsche's position on like energy and vitality being basically the source of life and he's you could almost sum up his entire argument about like the fall of man being um that we've uh we've traded vitality for a false sense of morality and um and in the process we've become basically uh slaves um and that's sort of you know something

i wanted to dig into so anyway the long story short is we we're doing the the energy edition this year and it's got obviously yourself uh headlining alongside brian gitt alongside gideon powell um harry suddick uh myself and uh drew armstrong from cathedra and andrew myers from satoshi energy so um yeah man i'm really looking forward to this and i i wanted to specifically like dig a little bit into your piece um because i know you were going to tackle the esg side of things yeah i mean this is something i've been beating the drum for many years now at least five or six yeah probably closer like four or five actually but i like to think i was early on in this trend and it feels like five or six right that doesn't like the last like sometimes i think

i left australia like fucking 10 years ago but i'm like 2019 like it's four years like we've seriously gone through a battle in the last four years yeah no but i mean esg is the perfect um sort of um what's the word i'm looking for it's the perfect manifestation of that nicha line of thinking you just described which is we've replaced vitality with the moralization of society we think we're moral beings and this moralizing has led to a complete rejection of reality and what actually makes the what we're doing now possible street you're on the other side of the world connecting being internet connection i've got a camera here i've got lights here you've got a laptop and like all this is made possible by energy

and esg particularly the e part of it we can even dive into the s and g part of it as well or just complete rejection of reality and actually really ties into the ai conversation we were having earlier in this push to sort of control the language it's a complete rejection of reality and attempting to centrally plan a moralized view of what the world should be on the markets. And energy is a big part of ESG, obviously. It's probably the most predominant part of the ESG movement over the last decade and a half. And we've really seen the repercussions of what rejecting reality and trying to moralize has led to, um which is less reliable energy infrastructure human suffering i mean i think one of the most

stunning facts uh of the last two years is the fact that more people died due to lack of access to energy in europe last year than were killed in the russian ukraine war which is well i didn't know that okay insane let me make sure i have i pulled it up before but uh the you know it's funny like the suffering as a concept is like something we all have to go through but it's like it's always so stupid when it's like unnecessary suffering it's like you know you can we have the capacity to make all the energy we need um and like you know say save the suffering for you know greater things it's like you know there's you know there's suffering when you take a stick and you put it into you know your um the front wheel on

the bicycle and you fucking wipe yourself out right and you're suffering because you like you literally did it to yourself versus like suffering to build a business or to you know start a family or to create something like this like one is like a suffering that is trying to reach for the stars and the other one is a suffering of like self-induced fucking stupidity and like it's like the contrast is so extreme it's not even funny and like as a as a civilization we are spending more time putting sticks into our own spokes than we are like reaching for something of meaning it's insane yeah i mean and yes the stat yeah during the winter of 2022 there were more deaths in europe caused by the shortage of energy uh which let i mean people weren't able to protect

themselves from cold temperatures than from covid and you in the ukraine war combined um which is insane and we saw the headlines coming out of germany and this is germany i think is perfect microcosm of the insanity of esg they started their transition to a renewable grid system early this century and have successfully decommissioned reliable coal and nuclear power plants over the last decade and replaced them with wind and solar and you look at germany you just look at the weather there it is not conducive for producing a lot of power from wind in solar it's very cloudy there um you have stretches of time where the wind doesn't blow as much as you like to think and then like they're doing this to save the world

to save people maybe you don't even say to save the world you can say we want to save humans from suffering and dying and it's very ironic because this transition led to human death and suffering uh during a time of war uh when the energy supplies particularly being provided by russia via nordstrom 2 uh were politicized and attacked physically uh really cutting off germany specifically but a lot of europe from reliable energy sources which led to higher energy prices lack of energy accessibility uh and deaths at the end of the day and so this whole rejection of reality the boogeyman that you described earlier climate change global warming which doesn't have seemed to have materialized as somebody's been following it

for some time you're saving people by killing people at the end of the day and it's just this complete inability to to recognize that we only have what we have today because we are able to utilize energy to our benefit i mean going back to the beginning of the conversation the sun provides us energy to wake up and go work food banana provides us energy to go toil and then you work up from there humans with our logic and our ability to to think creatively have figured out ways throughout millennia to harvest energy and get increased productivity and efficiency and leverage essentially on what we could do physically and really created machines to enhance that beautiful capability that we have.

And we've gotten to a point, and it's funny because you can point to the Henry Adams curve, which is energy production per capita globally, and when did it start to flatten out? It's been flat for 50 years in 1971. it's like this whole fiat world uh going disconnecting money um from any sense of reality as well it's just all debt paper money um has allowed us to sort of reject reality and i think it does stem from the money when you disconnect money from any hard truth in the physical world just make it this thing that you can print willy-nilly

um you're then able to completely begin rejecting reality in other parts of the world and energy is i think one of the most obvious sectors that has been neglected over the last 50 years yeah you can just you can lie to yourself on what you were saying reminds me of a little passage in the piece that i was trying to write um i want to uh damn it i'm not going to be able to find it enough. Yeah, I'm not going to find it now. But the, yeah, damn it. I had it highlighted, but then I refreshed the page. Fucking moron. There was a section about that. But yeah, mean the the relationship between um between energy money i mean you know the i think i don't know who i first saw write this like called calling bitcoin energy money do you

think that was a sailor meme or was that like sailor definitely popularized it um but even before then people were like tying it to henry ford's idea of energy money back in the day henry ford or buckminster filler both of them okay yeah okay yeah because yeah bucky was talking about like kilowatt hours uh per something right as like the unit of money that we could use yeah i think bit bane uh bit bane popularized he's a lesser known bitcoin twitter guy but he's been really focused on the energy thing i think we can safely say that sailor popularized it mm-hmm yeah i mean the you know the the point that i try and go into is that i say um like you know nietzsche talks about the uber mensch right like so this kind of new man that

emerges from the age of what he calls the last man and you know to him the last man was like the um basically the npc or the lemming as i you know like to call him is like you know he predicted a world that will because it oriented itself around average around equality um around you know lies and has um has a disdain for vitality and power and energy um will you know basically the world will transform into ugliness and we see that like you know did you see the um the thing on twitter today with that stupid uh fountain in europe like that fountain oh dude it's like a you know this new fountain of like and the sculptures look like basically turds like it basically looks like a fucking pack of kids from you know preschool got

together and like slapped a bunch of clay together and this is the new like they spent 1.8 million dollars on it um and like this this is the ugliness that the world is headed towards because you know we we lack beauty we lack vitality we lack energy we lack something uh to reach for and you know the the the piece that i wrote is called uber money and i say you know the the what needs to happen for the uber mensch to kind of uh emerge is a mixture of the right seeds uh to be planted and that is the um you know the things like bitcoin things like uh you know the warrior virtues etc and you know humanity has those seeds naturally but i think bitcoin is one of the most important seeds but then also you need a lot of fertilizer like you need a lot of basically shit

manure and that's what uh we've got in the world today and from that you know a better a better quality human being is going to emerge yeah there we go what the fuck is that no this reminds me of the statue i think it was like a martin luther king statue they put in boston just looks like somebody holding a cock it's uh yeah not it's it's disgusting like look at this what is this supposed to represent vienna's renaissance produced some of the greatest structure that fountain was just unveiled in vienna this is in vienna austria yeah vienna exactly like this was this was the cradle this was one of the cradles of civilization what the fuck is that well and this reminds me of part of something i said on stage last week at an event in missouri

which is i think i'm touching on in the piece as well which is this is the beauty of this particular inflection point like there's many people throughout the world who wholly depend on the government to fix these problems and what we're seeing in america particularly in And the rust belt is that Bitcoin miners being incentively, economically driven to get the lowest cost of power, essentially fixing the problem that Triffin's dilemma created in the first place, which is you build all this energy infrastructure and had a large manufacturing base. You then ripped every, you ripped the U.S. dollar off the gold standard completely. And to ensure that the dollar was a reserve currency of the world, you flooded global markets with dollars and bought their goods. instead of manufacturing your goods domestically all those manufacturing plants steel smelting plants moved offshore leaving behind this energy infrastructure that was being wholly underutilized

and still is to this day and bitcoin miners are essentially fixing the problem um that triffin's dilemma created in the globalization of the dollar created uh which is they're now saying hey i need cheap power middle tennessee middle kentucky middle of oklahoma I see these substations that haven't been utilizing their full capacity for some time will come in. You can buy that power in bulk from the people you buy power from, get lower pricing, and we'll use that energy. This is revitalizing in a small way these areas that have been completely neglected here in the United States for decades. And I think this really highlights, and it's beautiful with Bitcoin, and also AI too, if we don't get cucked by the licensing. like we don't need to ask the government to fix these problems they created this

these problems in the first place uh they broke the money which broke everything and you have the tools at your fingertips whether it's building software at the protocol level of bitcoin building software at the application layer of bitcoin uh to basically give people access to a different monetary system and then on the physical side of bitcoin the mining infrastructure like it's just the natural way of things miners being economically incentivized is going to these areas that have been neglected and breathing some life into them it's not going to save these cities and these towns wholly by itself but it's the spark that's needed to begin reversing the trend and you can do this today by yourself just using pure economic incentives you don't need mandates or spending

bills to be passed you can go do this yourself you can build spirit of satoshi and give people access to to good information about bitcoin you can go try to get cheap power to to mine bitcoin profitably and the positive externality of these things is better information and better revenues for small towns that hopefully can be reinvested to begin fixing the problems that ripping us off the gold standard started in the first place yeah this you you touch on something that i've been uh harping on about recently which is like a the the the conquest approach or like the the approach to like claim space and claim territory versus the one of uh defense and this is where i think like you know i've kind of had it with like libertarians and libertarians and all this sort of stuff like

you know everything they seem to do comes from a place of defense it's like okay we're gonna go and cease it you know oh you know don't don't tread on me like non-aggression principle this sort of stuff like and you know notwithstanding the the wisdom and things like the non-aggression principle etc but the the energy where they all come from is one of defense and one of like um like leave me alone whereas what you're sort of describing here is like coming back and claiming the fucking energy you know claiming the land building shit again not asking for permission it's a it's a conquest it's a it's a conquering energy it's like it's an attacking energy it's not a running away from energy it's a going toward energy and this is something for me particularly

um that i have uh at times it's it's annoyed me about the bitcoin space because you know we we fundamentally are um you know many of us are more libertarian than anything else but like it's not enough just to sit there and um you know stack your sets and like you know act like you don't exist and all this sort of stuff like there is there's there's an imperative to some degree to go out there and build shit to create shit to like you know if you don't like um the way you know like i guess not to blow smoke up my ass but i guess i'm an example of this i didn't like the way chat gpt worked so i went spent 2 000 hours figuring the shit out and then raised some money and went out and started building something and like i'm gonna build a competitor to it it's the

same with anything else like you don't like seed oils well go and fucking make a product that like competes with the seed oils you don't like polyester and like sleeping on shitty you know know uh bed sheets that are made of plastic well go and fucking build a cotton thing like you know you don't like eating the bugs well go and start a farm like you know produce meat like this is so like that energy of conquest i think is fundamentally important and like it's in the word it's like you require energy to go out and actually claim land back to claim territory to claim space and this is what made the west fundamentally great in the first place like what did we do we went out there and we claimed fucking space and now we're receding like oh we

can't use that much energy we need to conserve energy we need to like turn off our heaters at a particular time oh we can't do this we can't do that it's like it comes from such a like a place of weakness a place of uh like begging for like the crumbs that are like sort of uh spilt off the table and i don't know i just find that so pathetic whereas you know real real leaders real like a culture and a society and a civilization with some vitality wants to go out there and like you know it'll it will still argue there'll still be fights there'll still be conflict but it'll be about like the place in the stars versus uh groveling and bickering about you know your place in the dirt right like it's it's a complete different um paradigm shift and i think that's

just so fundamentally important it's you know we've been we've been playing defense for too long it's literally time to play attack now and i think this is is what this conversation reminds yeah and it's beautiful because you see it in the rust belt very like it's been completely neglected again for decades and again it's not gonna it's not a panacea for everything but it is a sort of what i've seen what we do what standard bitcoin does in tennessee in kentucky they go into these rural towns that have had manufacturing move out we go to see the substation we're like hey are you utilizing all this capacity they're like no i'm like all right how much can you get out of it like another 10 to 20 megawatts like all right

get that energy delivered we'll buy it from you contract sign build infrastructure plug-in miners and the direct sort of positive outcome from that relationship is with the utility so that's where it starts with us the utility is like oh we can actually make money and the way it works in this part of our country is it's utility co-ops and the co-ops have a mandate to keep prices low as low as possible for residential consumers and since we come in and soak up that excess capacity they can buy more energy in bulk which means they get lower pricing they can pass on to consumers so they're like all right we're following our mandate we're making more revenue um as a mining company in tennessee we're paying taxes on that so the state's getting more revenue um but that's

like the the direct positive outcome is between us and the utility company and again like you said just go out and conquest like these parts of the country have been completely neglected by the government for decades and almost forgotten here in the united states while the coastal elites reap the benefits of the cantillan effect and have seen an insane amount of wealth accumulation which has driven the the wealth gap here in the united states but what we've seen is that once we get in with the utility company they see the obvious benefits of the increased revenue that they're going and they tell their friends and other utility companies what's going we don't even have to do outbound sort of exploration of new sites people are coming to us like hey come

do this in our town come to us in our town so this is i think a direct example of just go out build the better world and then the people you interact with will see hey you're actually coming and helping me making my business better which allows me to serve my customers better at the end of the day um and then there will be small domino effects that lead to larger domino effects after that but it's literally getting out there and just doing it not asking permission okay you've got a substation you're not utilizing it let's let's utilize it we don't need permission yeah it's it's it's so simple it's like and you know i should say like people shouldn't confuse simplicity with um with ease right it's not easy like you gotta you gotta do the fucking work like

there's no there's no like uh magic wand that someone's gonna wave and like all of a sudden there's this all the problems go away like you've actually got to go and do the work but it's actually fucking simple like you you want something go and build it you know you want to you want to go somewhere you fucking go there um you want to create something you roll up your sleeves you figure it out and you create it but that's that's what you know human beings with agency do um you know this is what like ai will not do like it's not going to figure this shit out themselves itself like we are the ones who are going to direct these tools you know bitcoin ai asics miners energy power plants the grid all these things are fucking tools that we as the

masters of our world our universe like we're the ones who direct these things and like you know it's it's such a small shift and then all this stuff that we're saying it's like it's so obvious we all fucking know this but i think it's sometimes you need to be reminded of it it's you know this is literally like without sounding woo woo but like you you're fucking pulsing with energy like a human being is an energetic being and like you just need to direct that shit in a particular way and like the people who transformed history you know the alexander the greats of the world the julius caesars you know the napoleons the nikola teslas the isaac newtons like these people directed you know their energy to to toward an outcome toward a vision towards something and

they literally fucking change the course of human history and like that's the energy we need to take on if we're going to fix things because for damn sure the you know the monkeys that are pulling the strings at the moment are for damn sure not going to fix fucking anything no and that's where i like vote harder daddy's like this always seems so dumb to me it's like what is the most practical way to do this again i think bitcoiners recognize it like almost right away because they recognize the fact that bitcoin succeeded because it's been this grassroots movement like we're going to fix these larger structural problems throughout our economy and with this particular example the energy sector it's going out there and doing like grassroots movements like substation by substation

and then you get allies starts with utility companies and residents are like oh my bill's staying lower in this higher high inflationary environment why is that they ask questions like It's like, oh, actually, a Bitcoin miner came in and allowed us to buy energy in bulk, which allows us to keep prices lower for you. And then you get allies. Obviously, you have the green pieces of the world and others trying to besmirch Bitcoin miners for soaking up too much energy. But at the end of the day, I think people will recognize the benefit that these operations provide and have provided over the last five years particularly. And we'll get allies. And that's how you solve these big problems.

it's not by going to the government and saying hey fix the energy sector it's like going fixing it yourself having the people on the ground seeing the benefits of that and then when the bigger battles come like those are going to be the people that that stand up and say no you're not taking this from us yeah man totally oh so anyway we've gone over time by like 20 minutes i don't want to keep you too long no it's all good i think um i think we nailed it so we we hit the ai stuff um we hit energy um i'm all out of comments agi is a red herring uh to control speech uh opening eyes of the world are uh overselling what they can do reaching limitations totally world of a thousand minds and making sure that speech isn't controlled

It's probably important. Energy is important and we can fix it. You can fix it. That's the message I want to get through to everybody. And what I said on stage in Missouri last week is like, stop waiting for permission. You have the tools at your fingerprints. Just go do it.

Pick up the phone, figure out how to get introduced to somebody that builds ASICs that you can acquire, get on the phone, call a contractor that knows how to build a site next to a substation, can plug in transformers, can plug in wires, like like you said it's simple it's not easy it takes a lot of coordination and learning but

it's the blueprint is there it's just getting your ass off the couch and actually literally about to say that it's about getting your fucking ass off the couch that's it um yeah man um thank you once again for having me on brother um really appreciate it you know if people want to go and check out uh spirit of satoshi um i'm going to say the domain name right this time it's spirit of satoshi.ai for the love of christ and um and yeah the the energy edition uh is going to come out for christmas basically christmas uh shipping and then um by the time people hear this podcast um it should be up available for pre-sale which i think the pre-sale is going to come with 40 000 sets off the price um so i think probably do a code either marty or energy one of the two

and people can go and pick that up go pick it up freaks go check out spirit of satoshi it's really cool thank you for building it thank you thank you for getting your ass off the couch thank you sir my ass doesn't uh like to sit on the couch all right that's all we got today freaks peace and And love. Okay.