209 - Why e/acc Is Right with Beff Jezos (Guillaume Verdon)

what we're advocating for with eak is sort of freedom of access to compute freedom of access to AI we don't want these centralized entities just like a central bank like inflating away your money to control access to Advanced AI we want it in the hands of many because otherwise the centralized parties are going to abuse their power welcome to bank list where we explore the frontier of Internet money and internet Finance this is Ryan Sean Adams I'm here with David Hoffman and we're here to help you become more

bankless or maybe more eak in today's episode right cist yeah we have Beth Jos he's the founding father of eak that's effective accelerationism this is a philosophy that thinks AI doomers are not only wrong they're taking Humanity in the worst possible Direction the emphasis here is growth acceleration progress these are the core pillars and instead of slowing down on AI Beth explains today why we should be speeding up they this is a direct answer to an

episode that we recorded over a year ago with eler owski on why we are all doomed and why AI is going to kill us all so it's a a debate I guess if you listen to those two episodes sequentially we talk about a few things on today's episode including why AI isn't going to kill us after all according to Beth and what is eak he explains the fundamentals and why eak is more than just AI for Beth Jos it's an entire Philosophy for life also we get into what in the world homo techno capital mimetic accelerationism

is did I say that right David I don't know finally we end with be's take on crypto one thing I enjoyed about this conversation is that I think you could end up at this conversation from a variety of different angles um if you are coming from crypto you will start to catch a Vibe for this it will sound familiar to you a little bit if you've been pay paying attention to the AI world for sure uh but also you can go really far back and just talk about philosophy like if you are a fan of uh Friedrich nche like you will find

yourself in a familiar place um I've been a big fan of the concept of optimism as a way to um achieve strong mental health like optimists tend to feel better they tend to be less depressed they tend to feel like they have more control and agency around their surroundings so if you come from it from like the the mental health and uh like psych psychology background you will find yourself I think uh resonating with some of the things that Beth jzos has to say that's one thing I really just enjoyed about this conversation is

it Beth starts from a conversation around base principles getting really close to the metal and then expands it extrapolates it outwards onto like a political philosophy that people could align with so it's worth listening to this episode for that reason alone David there's so much more to discuss with you in in the debrief I think I've got to take some time to uh really unpack this and that of course is the episode that we record right after the episode you're bankless citizen you can enjoy that now all right let's get right to the episode with Beth but before we do we want to thank the sponsors that made this

possible including bankless Nation I'm very excited to introduce our next guest this is uh one guest but two identities on the podcast today uh Beth Jos is a founding father of the effective accelerationism movement this is also known as eak and you've heard David and I talk about Yak before particularly in contrast to the AI safety or Dell movement as some would call it uh eak is a basic philosophy that advocates for the full acceleration into our AI

powered future this is all gas and no breaks Beth welcome to bankless well thanks for having me uh pleasure to meet you Ryan and David and excited for the conversation today now I said there were um actually two identities one guest but two identities so um Beth is your online alter ego but in meatspace your name is your your identity I should say is Gom forone this is your original identity your nation state assigned identity yeah your human name let's call it and uh you are a physicist Quantum

researcher and the founder of an AI startup called xtropicalovex why don't we actually start because um for some people that may have been a little bit like jarring right so let's start with the the decision to use a pseudonym to Kickstart the effective

accelerationism movement why did you use the P pseudonym of be Jos like why not just use your uh your regular old nation state Fiat name I think uh you know at the time I was uh you know in some secretive teams uh within a secretive unit of uh a big Tech uh company uh you know I was at as as was a Google X and their you know secrecy is Baseline I was in particularly secretive team can't really talk about what I was doing um people can try to piece it together for my

patents and have fun with that some Easter eggs in there um but you know secrecy was my sort of Baseline um I couldn't necessarily uh say too many you know have too many hot takes that you know wouldn't get back to me right and or I felt I don't know if that was like fully true I mean I've had some feedback that way but there was always that sort of you know thought that if I said something that would be over the line it's over I would jeopardize my career you know again this was like pre- Elon

buying Twitter it feels like a completely different vibe now which is great U Back Then though it felt like if you said something some sort of lat truth or something that you felt or you were even just trying to experiment with with your ideas and experiment with different point of views um uh you can you can lose your job and and get canceled and so on and so to me you felt muzzled yeah I did and and it doesn't have to do with any one particular employer to be fair I think it was just the general Vibe I think many people felt this way uh and I I I think that's

why Anon accounts were so potent you're kind of um uh removing the sort of uh Threat Vector of of reputational counterattacks if people try to shut you down which of course now that I am doxed I no longer have that that Shield but you know I still stand by my my values and that's why I'm I'm here on these on these podcasts but you know at the time it's funny how um giving myself an anonymous account kind of opened up a whole new space of ideas that I didn't even let myself experiment with because

I knew I wouldn't even be able to communicate them right like so so freedom of speech induced freedom of thought for me and and sort of Beth was a sort of experiment in in in applying my thinking that I've applied to sort of quantum Computing physics based Computing and AI applying that sort of thinking and background to sort of civilization and and and societal systems and and how we organize ourselves in culture um and really it was supposed to start as a low stakes

experiment you know have a couple followers and and just put some ideas out there see which one uh stick but it kind of grew and kept compounding and and now we're we're here I guess and so um but that that was really nice for me um because at the time I think I had grown I had grown my reputation in in my my field of study which was quantum Computing and it wasn't clear to me if if you know my my fans at the time would you know just wanted a job or something

like that it was kind of the the opposite of why most people do an an- on account right maybe they they just want well I I think like eventually people just want their ideas to be evaluated on their own right regardless of reputation regardless of your background for me it felt like I want it it's it's like a I I do the analogy of new game plus when you beat when you have a beat the video game you want to restart with like your knowledge base and maybe some of your gear but like you're restarting the game and so for me it was like can I rebuild a sort of a a following or or

reputation uh um you know from scratch right in an uncorrelated fashion so I would never talk about Quantum Computing I'd only talk about AI um and so started it uh while in Google X and then I ended up uh you know leaving that and starting my own company now of course you know I'm My Own Boss more or less so so you know I I I have a lot of more freedom of speech but I I think overall people feel Freer to say sort of things that were maybe taboo or off limits a couple years

ago and I think that's healthy for everyone uh embracing that sort of variance in in the space of thoughts and so um yeah I don't know well it it also feels to me I mean try this on too just for my observation Beth has a different uh communication style than than Gom I think so you know Beth is a bit more uh mtic Beth is a bit more like inyour face Beth is a bit more uh rhetoric uh I think and that communication style seems to have meme Market fit online at least

I think accounts for the growth of the uh Yak movement so far like I don't know if you were trying that uh personality on or that Persona on and it's just like a better communication style for the spread of ideas but that's one observation I have from from seeing Beth uh communicate yeah I mean like we say it you know we're on a mission to spread uh viral optimism a sort of [ __ ] you optimism like Unapologetic optimism about the future right uh to to fight

the sort of pervasive pessimism and and and dorismar mimetic vaccine that renders people immune to sort of uh feeling terrible about the future and being demoralized we have a responsibility to to make it spread right and so if in an era of algorithmic information propagation you can ab test what sort of

packaging of your message uh has highest uh uh mimetic Fitness and you know usually that's memes that's you know a tone that's sort of brash in your face uh that's adversarial um that's what gets propagated that's what gets seen and Argo you know that becomes you know that that shapes people's priors because they have an information diet that is that is fed to them by by these algorithms and so uh yeah I mean it is a sort of

certain communication style um I had sort of uh experimented with sort of similar Firebrand style uh in Quantum Computing back then um you know I was really trying to push Quantum Computing towards something like deep learning differentiable programming right so that's what I that's what I did at at Google with with some of my team there is is bringing sort of deep learning thinking uh you know let grading descent program for you uh uh to to Quantum Computing and and and that was a sort of

similar fight back then against let's say the complexity theorists that needed hard proofs uh about everything every statement rather than just you know uh letting letting letting go and letting the computer figure out how to program for you and uh it it feels very similar to this sort of uh battle against the rationalists and the EAS that want to have you know axioms and want you to prove that you know Humanity will be Eternal and you guarantee our safety uh

forever and if you can't prove that we got to shut everything down right like very very very very similar sort of adversary so I've been through this drill before so I bring that sort of sort of baggage to this this new fight so that's so so Beth um we're we're already talking about adversaries the rationalists in the EA and so I think we need to to te up this conversation now that we've established who we're talking to Beth Jos the the one of the the the founding fathers of effect effective accelerationism so let me just throw

some context for Bist listeners and for you Beth as we get into this episode of of what we're hoping to achieve in this and the reason for this episode so about a year ago like almost a year ago to the day that we are recording we did an episode with a gentleman by the name of owski uh I don't know if you know him at all or if run across them in your travels um so we thought David and I thought in our um more innocent days that we were going to have a nice conversation with eler owski about the interplay between Ai and crypto and uh

it turned out that was not the conversation Destiny had in store for us it turned out we got stuck yeah we got stuck somewhere and we were going to end that episode in a full-blown existential crisis about this thing I didn't know as much about as I feel like I know today but yet I'm still kind of like wondering how much I really know that is AI safety and El eer offered a compelling argument for basically why we titled the episode why AI is going to kill us all because that was the distinct impression he left

us with in fact the episode the entire episode was a dire warning like halt all AI progress good you know like right now you know uh go go be with your loved ones kiss your your children goodbye because uh we're not going toan went and and followed that advice actually never it's never bad advice to go you like tell the people uh in your life you love him and and so we're having you on partially uh Beth because you stand on pretty much the opposite side of that argument I think and not only do you not think that AI is going

to kill us you think it's going to be great like maybe Usher us into a Utopia okay so this is not like H he was right and some you I think are saying eler is is dead wrong in fact it's the complete opposite and we should accelerate progress on AI completely across the board and so uh bankless is a uh technology podcast but but very specific in in in the crypto community and I want to give you the chance to um explain your side of the argument I want to give

you a chance to pitch effective accelerationism EAC to the crypto audience and just like pill us that straight up pill us like give us the gospel of e eak Beth getting right into it yeah I mean um you know yeah that that that yusy podcast really really was uh I think the start for you all to to get into this sort of uh this sort of area uh did you did it someone across your feed it definitely did it definitely did um I I I I think like the

sort of uh yeah Doom mongering or fear-mongering has been really nefarious like it it's really affected people psychologically right I mean but yet and yet like you know they thought gpt2 was going to kill us all gpt3 where GPT 4 4.5 soon yeah we're still here right the economy hasn't collapsed everything everything's fine the the the crypto bags are still you know they're doing all right they're doing all right these days uh but these are the last days maybe bef that's what L here says like

enjoy the last days no I I think that um that's the thing about having a system that is uh malleable and adaptive even if you have new technologies that are progressively rolled out it just morphs and absorbs that capability uh and and provide and creates utility out of those those new technological capabilities and and really eak is about trying to understand

where do we come from and where are we going right what is the process what is the process that gave rise to life what is the process that gave rise to civilization what is the sort of this weird you know clearly something's going on right like when when we were maybe you know CH children you know we had like massive computers that were exponentially worse than the ones we're using to converse today something's going on there's a sort of evolutionary process in the space of Technologies and the space of ideas uh you know what is

this machine what is this process that's always um uh morphing and adapting uh civilization and the Technologies around us and and really eok at fir at first and foremost is trying to understand this process process how does it work how did it get us here why is it good and and then how do we accelerate it right and you know that's what we call you know acceleration or or techn Capital acceleration or more generally homo techn Capal mimetic acceleration so

that's like all the things right um essentially we think that um this process of searching over the space of parameters or bits of information in how we organize ourselves culturally in terms of uh um genetics of humans in terms of the space of Technologies in terms of how to organize companies Nations Etc it's it's all one big search process and

competition induces a sort of evolutionarily evolutionary selective pressure on the space of of all these things and that sort of of uh competition breeds Fitness that then uh benefits us all right like in a sense like we've tried we've tried capitalism and freedom versus sort of authoritarianism uh uh or communism where everything top- down prescribed prices are essentially controlled so far free markets have been far more

successful at creating wonderful things such as the Technologies we use to to chat with today and the system we live today and overall um such systems where you have many more freedoms you find much better Optima uh in terms of uh again technology is ways to live your life uh uh um and and just about everything um but it's um I think I think it's too broad of a question like just pill me on eak um I think like

let's let's just let's go through what how you feel about let's just pick it apart really let's just pick it apart yeah I kind of want to do one more thing first just to kind of frame this conversation one of the um maybe like disarming things about the conversation that we had with elaser owski that maybe like me and Ryan just weren't ready for is that eleazar's arguments were extremely technical they went down to the basement on like how AI Works he was using phrases called like

gradient descent and uh you know reward Bo mechanisms and it all got like outside of our frame of expertise very quickly because he was like down at the basement level and he was making like this kind of logical argument that like me and Ryan were just like weren't ready to like fully unpack cuz he got really really Technical and this um this conversation this eak versus diesel people can approach this conversation at like varying uh like um Heights like some people I think uh Mark andreon

wrote the Techno Optimist Manifesto talking about this at a very high level and then we went and me and Ryan did a number of episodes about AI safety and each one of those conversations were like somewhere up and down in the very technical to very philosophical like conversation I'm wondering where you see your Innovation with this conversation you run an AI startup so you're pretty technical but you're also um just in your response just now is some pretty philosophical directional like approaches where would you say you you would like to inovate in this

conversation yeah I mean whatever you think is more appropriate for your audience for me it's always difficult to to adjust my level of of uh technicality I tend to converse pretty technically you know I was a theoretical physicist and then now I run an AI startup um usually I have very sort of mathematics first thinking trying to explain uh my thinking and I have to uh convert that into uh English to some extent um and so you know if if there's any words I say

or anything like that that you want to dig in like let's dig in but I I would say I would say yud is not very technical actually um and that if if you start digging into his technical knowledge it's it's it's quite it's actually lacking quite a bit right I mean this sort of um you know the sort of recur this notion of recursive self-improvement I guess that you know the runaway F maybe we can we can address right like um recursive self improvement is something we've tried in

in in machine learning for a very long time it's called metal learning right you you have a system that learns how to accelerate the learning and you have also architecture search algorithms the reality is that the the larger the space over which you search the exponentially harder things become it becomes to find the the the the true Optimum right and so what that means is that if we task an AI to improve AI it's it's it's exponentially more complicated every level of like an AI that improves an AI that improves an AI you go right

um and so it's going to take exponentially more compute uh and energy right to achieve that Optimum right and so in a sense there's a sort of there's already a soft cap on compute in terms of the energetic and capital costs of of compute and that keeps us safe right and it in a in a sense like everything everything in in in in our civilization even every life form is trying to f everything is trying to grow and that is the thesis of of eak and but that

because everything is trying to grow and competing with one another for resources we yield we we get uh better Optima but it also keeps everything in check there's not one sort of Singleton you know biological uh uh uh system or one single company or one single nation that takes over the whole system right uh and because they're they're they're competing with one another and and in general to find sort of uh Optima of high Fitness for example you know if

you're if you're building a company a company is almost like at least in a startup space you know at the seed round you're basically a search algorithm you're searching over the space of Technologies and you're burning capital and intellectual compute and and maybe uh actual GPU compute to find sort of to pinpoint exactly what product you got to build that's a couple bits of information but that's costly that's a costly thing to find so every Optimum that is highfitness is very costly to find and the more optimal it

is often it's exponentially costlier to find uh that Optimum and so that keeps us safe in the sense that I can pretty confidently say we're not just a small at a distance like this one weird trick where like now we have super AGI that fums and creates gry go and takes over the world like that's L that's like very exponentially unlikely uh that that that that that is the case you can't guarantee it right you can't guarantee that there's not one weird hack but from what we've seen so far it's taken

billions of years to get where we are now with uh life and intelligence as we know it and it's been a very long sort of process of of of improvement uh to this point um but there sh there won't be a sort of hyperbolic discontinuity in in the rate in which things improve and in general we we take the opposite camp like this sort of rate of self-improvement of everything is actually really important and it's what um uh uh creates uh the everything we

enjoy uh and we should uh seek to to to keep that process go growing and scale it right um really like systems in nature uh either get busy growing or they get busy dying right uh and so they either secure more resources SEC secure more free energy and figure out clever ways to utilize it in order to to to to grow or they try to stagnate they run out of fuel and then they die that is it

everything runs on some type of fuel nothing is forever no bit of information is forever to maintain its coherence it costs energy because everything wants to decohere naturally everything wants to sort of you know it's a it's a constant fight against entropy and so the the thesis of eoch is the one golden metric to some extent that measures the progress of the whole system is sort of free energy how much energy are we uh acquiring and consuming as a

civilization because that's a metric of our progress that can't be gained right if you if you and and this will resonate with you guys but if you measure it in US Dollars that's not that's not objective scale right uh you know uh that's that's not a scale that's a scale that you can you you can play with you can uh you know inflate things you can have inflation uh you can you can print money and and and then it seems like you have some progress but really really you

went the other the wrong way whereas something objective like energy is a good metric to measure progress and we think that um scaling up civilization and and and doing so with urgency is how we uh Ensure it's sort of long-term success and existence uh and I think that the one of the most dangerous things we could do is let this mindset of Doom and demoralization uh become

hyperstition you stop hoping for better things in the future it s sort of becomes a self-fulfilling prophecy and that's how civilizations and Empires uh die and and that will cause massive pain and so we're on a mission to spread sort of optimism that is

hyperstition nuclear energy massive Prosperity we have all this upside on the table and the longer we wait uh the longer we wait uh the lower likelihood we can achieve it and we we have urgency to make it happen and and and that's very similar a mindset you have in startups in Silicon Valley right the most successful startups are very optimistic and and they believe they can do something that seems like unimaginable at the beginning of the company but then

they do it and it it just keeps happening right and this this sort of hyperstition op optimism uh effect and that's the sort of mentality we're trying to scale to the world uh with eak and it's why we think that dumer ISM and pessimism is is really dangerous and needs to be fought somewhat aggressively right because because that is actually the source of Doom not some sort of fictitious uh artificial superintelligence from sci-fi it's not

backed by science right um you know I've worked on AI for material generation protein folding biochemistry chemistry all like AI for the physical world it's much harder than people think right the physical world is really hard and one of the best uh uh things at designing things in the physical world is life itself but you could see life itself as a big optimization algorithm that is trying to F every life

form is trying to F and yet nothing has fed right so that should give you a bit more peace and you you you can breathe um and but happy to go into any sort of argument like any sticking points you have about uh y's argument one thing you've mentioned a couple times is this word fum could you define that really quick for people yeah I I mean you know well H how how do you understand fume like from your interview with yud um how would you explain it yeah I I think you're using it in a very general sense

which I think um I which I appreciate because it kind of uh gets down to the bare metal of kind of how life works like life is interested in life like life is interested in propagating fum in the AI sense is main is mostly talking about that super intelligence explosion where AI just takes over everything and it's all it's just like this one single event and all of a sudden the whole world is AI and it's run by AI for AIS um trees want to uh plant more trees you know uh bugs want there to be more bugs ants wants more ants humans Want More

Humans uh and I think this is it's all trying to hit some sort of like um uh point in a curve which like once the ball starts rolling it starts rolling faster and faster and faster and all of a sudden we have like a population explosion uh it doesn't matter what kind of life form you are but like everything is trying to look for that like growth in population that is something that's fundamental about life and so you're using in a general sense where it's just like hey any system that uh any system whatsoever that's propagating is trying to find more energy because that's how it can propagate even more and this is

this is the playing field that just the universe exists on I think that this is how you're using it it it's every every system and subsystem whether it's a company a group of people a culture uh like you said bugs trees whatever um even nation states uh everything is self organizing self- adapting its inner workings in order to to to grow and by construction things that are not optimizing themselves to grow burn run out of fuel and they fade and

that's it right and so if you think about if you think in the future which culture which um like if if you think of like several nations in the future which nations will have survived and what culture would they have right well they would probably have an eak culture that is literally by construction trying to figure out what is the optimal way to organize ourselves in order to grow and the sort of pessimistic Doris uh cultures will have faded because they'll