AI, Crypto, and the Future: Competing for Mindshare in an Exponential Age - TOKEN2049 Singapore 2024

welcome everyone um so we're going to be chatting about crypto um and Ai and what's the impact on crypto overall um so initially you know let's just start with very brief introductions and then we'll get into questions and hopefully make it interesting for you guys hi I'm Diego Monica I um co-founded a company called Anchorage digital and uh most recently I um became the GP at H Ventures hey guys so my name is D Wan and has been to token things that the first one and um yeah so running primitive Ventures and that's a not a VC firm but we write VC checks and it's a multi strategy investment firm and zero outside Capital so we do all the things that's weird for 10 years uh my name is Eric Wall uh I was a computer scientist maybe I still am um and um then I ran a crypto hedge fund for a while and uh now I'm the founder of tap Wizards which is uh making Bitcoin magical again that's it hi I'm Suna amaz I run F Capital we focus on crypto companies at the earliest stages across both infra and consumer before starting volt capital I started a company that is now on deck and before that I was actually in big data in AI during the Dark Ages at a company called Elian which is now a $2 billion company thank you for the introduction so with let's just get started with with the first question just to warm things up a little bit uh so crypto overall you know it's ultimately fight for attention a lot of crypto projects are fighting for attention of one another layer one Slayer 2s D projects it's ultimately how many users can we actually get how many people can we be interested interesting too now there's also Ai and obviously has been around for a while but over the last two years it's been the one of the main topics of discussion everywhere not just at crypto conference conferences but everywhere so how is that shift of attention on AI affecting crypto may we can kick off with Eric well I think it's changing it uh dramatically right you saw what was it s d braced $85 million in a seed round for an idea that he had with Stam that was kind of you know off the some random idea and 85 so the hot ball of money is kind of shifting a little bit from crypto into Ai and I mean it's understandable we're making ass coins and they're making super intelligence we're making ass coins and they're making super intelligence where would you rather be what do you want where do you want to put your money at so a lot of the crypto projects are either going to like a lot of them are going to become AI projects you see like near for example more heavily profiling themselves as an I know they were already like from the beginning always like AI landed but it's now it's like the the main thing with near right and you had uh what's the what's the the uh super intelligence Alliance the fcoin a bunch of coins went together like like were an AI coin so I think you're going to see more of that and I think that you're also going to see like the the the lucky thing that we have is that some AI project because like if you think we're in an in an in an intelligence explosion you're going to need that fast liquidity for your project so you might see some AI project that could stay just being AI projects but if they want to give liquidity back to their investors before like the whole world goes ape [ __ ] then they might have to go into crypto so that they can get that fast exit exit liquidity so the one thing that we can provide for this industry is fast exit liquidity that's what we can do soon I saw you disagreeing a little bit do you disagree oh um that is a that's the cynical interpretation and the more generous interpretation I mean there there are quite a few so one fundamentally you cannot achieve super intelligent or AGI if you cannot pay your engineers and in many countries uh like look at what what's had the strongest pmf so far obviously meme coins have had mind share but in terms of dominating market share it's stable coins and um going back to the basics of this fundamental let's fix the money fix the world which has dropped off a little bit recently and because of uh some of the distractions with meme coins is fundamentally the order of operations is be able to pay people to work on interesting projects like that's fundamentally the order of operations and so crypto is very important driver for that secondly Eric is right that it provides not even fast liquidity a liquidity Avenue like look to the most recent um exits in AI like look at inflections it wasn't even an m&a it was more of an LNA was like a licensing and acquisition while Microsoft had their shares in the company um the amount of money a lot of these model burn per year I mean you're talking about hundreds of millions the ability to have an exit that looks anything like we've seen over the last two decades is uh almost nonsensical to me and there is an argument where one of the very few Pathways to liquidity anymore is crypto I mean look at when Amazon was founded versus when they ipoed three years later look at when Netscape was founded versus when I ipoed 1994 to 1995 the only place you see that kind of timeline from origination to exit is crypto anymore in in general Tech it's now um anywhere between 10 to 16 years because regulations have made it so difficult to IPO because growth funds have supplanted um the public markets like you're seeing 20 billion private rounds because that's what the public markets used to do and instead you have growth funds carrying that burden and carrying that weight so um I I in terms of you know squawking and being loud sure meme coins are this ephemeral like temporary distraction whatever but you really really really need to refocus on the fundamentals this is so so [ __ ] important now I think picking up on the fundamentals at his most basic Primitives we can say that one of the things that we have that is working is General um um you know gen the ability to be doing generative Ai and creating content creating images and so it's most fundamental layer gen creates abundance and what we've created here in crypto is creating digital scarcity and so one creates digital abundance the other side creates digital scarcity so at most you know base level these are technologies that actually made uh for each other in the sense of like covering for each other's weaknesses and there's many such use cases in AI we have lots of problems with tracking who trained what on which model we have a great mechanism and an indelible ledger to track IP uh you have a shelling Point issue in terms of coordinating a lot of these demand and supply for gpus there's Market markets that you can create in crypto that not only have better payment mechanisms with micr and transactions but also have the ability of actually having incentives at coordination and voting and governance for for these markets to exist so it's actually really at my level the way that I see this is we have a lot of solutions for a lot of problems that AI is creating and even though right now most of it is narrative in terms of um the things that are working I think absolutely there will be many of these cases that will actually be solved by The Primitives that all of you are working on right now yeah doy have you seen young people still interested in crypto as much as they used to be 3 or four years ago or is AI taking away from that a little bit Yeah so I think for like any like new Founders or like any new like talents and then so they will evaluate like the so they will evaluate the industry from like three different like uh levels the intellectual leverage that they can have like the financial it's like intellectual Capital Financial capital and then also uh human capital right uh I think for our industry and um it's that every cycle like crypto can be just like like forcely combined with like any like new narrative and then so for like better so for better asset liquidity probably and then we are also investor in like cm and so I can so I can just like tell Eric like very very transparently because like when we are looking for Founders and then so we only looking like like crypto SL AI founder um so I have been joking about my mental model that like crypto and AI it's like arranged marriage like for like all the arranged marriage like it can be a little bit cringe like at the very at the very beginning but like once you have like two people in the room and then so they have to figure out like who do dishes and then so who do what and then so now we are at the face right so like that's what like Ai and like crypto is um so when we see how we can invest in this intersection like the founders that we are looking for is okay uh for so for founder Li and deep right no matter what he going to do like he can do darkcoin like whatever coin like he can like no matter what things that s going to do I going to invest right so it's just B founder and who knows that crypto so well and then who has the hiring capability to complement the intellectual capital on the AI that like don't I'm not saying it's Bad Bet saying it's a Bad Bet yeah right I'm just saying that's where the hot ball of money is that's and and and and sentient is basically a proof of that and what you're saying is also like just you know enfor reinforcing the same point yeah and then so when it comes to younger Talent right so I really think like crypto for like this cycle has been kill like first Kill by Ai and then so like first it's a first killed by SP SPF and like 3ac and like all the Fiasco and then so now we don't even have like Credit Recovery yet like there's not enough like Credit in the market right and like the second kill is by the AI um so because I quit San Francisco in like 2018 like because most of the crypto industry has been completely spread out like there they are not in San Francisco anymore when it comes to web 2 and then cicon Valley barrier they have unfair advantages when it comes to talents but when it comes to crypto barrier and like Valley has like disadvantage right uh and then I went back to SF like for the past two years and then so I live around noi Valley and it's every corner people are now talking about LM so I've been joking that this is a MLM for llm right this like a multi-level marketing for like LM it's that young people actually going back to the valley um so like young people looking for a few things like novelty diversity and then also wealth effect right um so I think like for like this cycle like like the wealth effect is like definitely not as good as like the other domain um so I think there definitely major distraction for like younger talent but for like veteran Founders and I think they're more like methodical and they're more seasoned and um so they think about things like more strategically and then so they would just say leverage like so they will basically leverage like the best thing that like each industry can like offer Eric you've been skeptical um at crypto projects that are focusing on AI or trying to maybe Spin The Narrative as as AI focused coins or tokens is there any you know use case that you can think of where the intersection between crypto and AI could actually make sense or are you just are do you just think it's it's never going to work it's it's always just going to be something that AI maybe use itself but not something as part of a crypto project yeah so I actually do want to agree with what she said that the stable coins and the payment rails infrastructure actually you know we're going to need something to transact with regardless of whether or not we're in an AGI arms race so I do think that she made some valid points there the criticism that I've been raising against the AI project is is that so the reason that Bitcoin works the reason that consensus mechanisms work is that because we can we can verify the proof of work right if you want to build like a decentralized AI project where like everyone's adding their gpus you cannot verify AI computation unless you're doing it with like a trusted execution environment on like Nvidia ag100 chips but you know grass Grassroots movements don't have like a bunch of Nvidia age1 100 chips there are nearly at the scale at the big Tech players so you're not going to get like foundational models out of Grassroots movement so you're not going to be able to compete uh in a decentralized way at the way that the world currently works with the how the supply chains for the AI chips work you're not going to get a Grassroots movement making an AI coin that trains a model that is like more powerful than than current models so like fundamentally you can't you can't really compete with the with the incumbents with the big Tech uh so you can't so but if you if you really wanted to do something with crypto and AI I think that there are some things that that we have that are powerful Beyond just stable coins we have prediction markets and prediction markets what what prediction markets are a Cornerstone to um futarchy so I don't know how many people in this audience that knows what futy is but it's a way to govern uh companies and even Nations by using prediction markets to sort of peer into the future of which different outcomes are more likely and if you if you for example have a liquid a liquid Market that says oh well the next uh GPT Model 5 is going to be like destructive then you can use that to inform the the board decisions at open AI well okay so maybe we won't release that model uh without more restrictions so futarchy is is a way that you can like in a forward looking how the word Works currently with the regulation is that someone does something bad and then we regulate that after the fact but in this case with where we're building something like super intelligence or artificial general intelligence we're going to have to have a more forward-looking uh type of uh regulation and self-regulation mechanism and prediction markets are fundamental to that so I think if you wanted to do some type of crypto AI project then maybe look into prediction markets that that's where I would look I would not look into okay we're GNA we're going to like compete with open AI on foundational models because you're just never going to get get access to the same level of uh AI superclusters as those uh companies will will have so I I just feel like the the criticism that you're you're bringing on is incredibly narrow I think it's very easy to agree that decentralized inference makes no sense and decentralized trending makes no sense in marketplaces that bring gpus together are flimsy uses of crypto in the first place there's just so much more that the technology itself can actually support the use of indelible lgers the ability to actually prove um I mean the coordination points itself crypto is the ability to coordinate agents that don't trust each other they're mutually distrusting and so that's what we've created and to a large extent what we're about to unleash on the world is agents that have their own independence and capability of operating and so it just feels like there are so many applications that are not decentralized INF decentralized training or even marketplaces which are the three ones that U that you said that don't make sense which I agreed they don't make much sense yeah and to be sure um most many projects these days aren't pitching that anyway because they concede as Eric said they aren't going to be as performant um largely because of physics and latency um issues that we just can't override yet um but some of the use cases that are being pitched and and fundamentally one of the most important ones that massive I think the majority of crypto space is missing at least a lot of investors I talk to is data being the bottleneck so in terms of data hygiene in terms of how much dat we can actually use that's uh publicly available I mean you have some teams working with Unreal Engine just trying to generate data synthetic data is probably one of the most slept on narratives in crypto not in web 2 where majority of them working in AI understand how important it is but within crypto um I can probably point to only one team maybe two U first batch Dria definitely working on this and it's this idea that um as we you know as our models get more sophisticated and as we have more use cases that we need to to apply this data this uh AI too we're going to need uh different types of data and we're going to need a lot more of it um so that was one where using crypto actually allows you to streamline um not only incentive for garnering the data but incentive for making sure that uh you are practicing good data hygiene through slashing and incentivization mechanisms built into the network um you also have interesting use cases around Federated learning which allows you to update um models at least share perams weights Etc while running them on local devices this is massive for um Healthcare use cases right where a lot of that data is pretty sensitive I think almost every single dollar that was invested into fhe plays this cycle should have been invested in Federated learning crypto plays because that is one of the uh most obvious and ambitious but practical um like grounded in reality of what we can actually deliver on in this cycle over the you know investing time Horizon that needs to be realized in um because you're able to get rid of that central point of failure which is the central pool that these weights are uploaded to so we've moved Beyond a lot of these use cases that we're trying to like we were feeling around in the dark trying to figure out what could crypto AI be and that's a lot of venture investing that's a lot of tinkering and iterating is opening up the adjacent possible as to what actually makes sense here and in terms of where the puck is going those two and I I I've been going on for a while so I'll I'll uh you know maybe open this for another discussion maybe later in the conversation but robotics which is the embodiment or substantiation of AI like we built the brain time to build the body there have been actually very interesting crypto plays around uh the intersection of crypto and Robotics but even just follow the stable coin mental model right like stable coin is not competing with like US dollar when it comes to as an so like as an asset like stable coin is just like competing with like US dollar for accessibility um I think so the thing for okay where we can see the best like Union Point when it comes to like crypto and AI um I think people in us and like they just don't have this at first experience but CH gbt is not accessible in Middle East and then so in many part of like Europe and not accessible in like China right so it's not even accessible in like Hong Kong um so I think like crypto has been making accessibility Global um so that's always the case right okay so I so can I just really quickly ask question so I similarly at Le at first was like okay you know if you want to use LMS are not CCP approved like this is a good way to be able to do that but they're all open source anyway so is crypto actually solving that I so I was about to say that because we have to seen two Trends and the first trend has been magically like drastically improved like when it comes to like llama 1 versus like gbt 3 right like gbt 3 basically beats like llama one in like all the dimensions like all like the like all the benchmarks and now we have like llama 3 right and llama 3 and gbt 4 and then so I so I think like gbd4 is still better but like not that like order of like order of like magnitude better right um okay so then where is that crypto coming into the place like given now we have this open source model that they can have like onart performance like um like afriy and uh with like local compute right so with like local computer and like as Compu and then so with like using like token as like a gross driver and then so you can actually bypass like all the censorship potentially so like that is our mental model when it comes to like how crypto can potentially solve this accessibility issue like just like stable coin solve that for like US dollar but do you need uh do you need a crypto to build a VPN like VPN has nothing to do with crypto if you wanted to be able to access things that you cannot because of your oh yeah like so like three choking points and I don't think crypto will ever solve right like app store right just like iOS like app store like that's not going to stop by crypto like CDN like content so content deliver uh so like CDN content delivery Network like not going to be S by crypto networking like not going to be sld by crypto yeah so we not talking about those like those are things are not going to be S like crypto any like anyway right I'm just saying like if you say that well China can't access chat GPT well and we need something crypto related to has to local comp you're competing with vpns no no no no no it's like a asge compute so I basically have my model like running on my own phone right so I'm not even using like like like so I'm not even using any like Cloud inference right yeah so like those are the things that and I I think is actually but you don't need crypto to like open source llms yeah so like crypto is not solving LM but like crypto can potentially solve like the adoption issue right so then then why people should ship from BYU to my local gbt that's right also um if you think about it Google's internal memo when they talk about uh open AI has no mo and neither do we um which was a very similar tune that Zuck was um writing in the manifesto when they shipped the most recent llama update it's this idea of like how really do you compete when they're all open source and they seem to be commoditized and if you think about it financialization of it like actual incentives behind what you use and what you contribute to I mean we've been playing in that sandbox figuring out how to monetize and financialized and incentivize open- Source projects for the past decade so crypto Builders have a huge leg up in terms of figuring out how to capture value and incentivize both users and contributors in this new era the open source so while I respect like the open source local model Paradigm and I think it's interesting I think what you're going to see is that all of these big Tech players are going to improve their algorithmic capabilities in how they build these models and yeah they may share their weights with you but when you interface with the model you're going to get a lot better uh cleaner experience when you actually communicate with like The Meta AI they're not going to open source all the different things that they're doing on the back end so yeah while they might Outsource like some of the model you're not going to get as powerful of an experience as you do when you interface with them directly because they're going to want to maintain like an economic hold of this technology so they're not going to give you the entire Kings to the to the the keys to the kingdom they will open source some stuff to sort of invite the uh open source Community to build stuff based on that but it's not like um they're they're going to open source it for all of us and it's going to be fine that way we're still going to probably an argument for for for crypto open source plays that's that's that's yeah but we're not going to be we're not going to be able to train open source models that's fine and it doesn't it doesn't have to be as exactly it doesn't have to be as performant enough if if you can't even if your problem is accessibility then you'll set you'll do you'll make the tradeoff for performance I think you'll just get a VPN and then you'll you'll you'll inter you know it's another tool in the tool of the cat and mouse game that is accessibility so maybe it will have utility maybe it won't but it's another tool around access the the other interesting side of this by the way is the places where AI will actually help crypto and there's lots of cool like little places where this is happening I just saw one yday that I thought was pretty cool so we have all these oracles that are injecting information on chain and then we have humans doing the resolution when something is weird and when we don't actually agree what happens for example did Maduro win the presidency or not things like that and so what they're doing I saw there's this company called chaos Labs um which we just fed our portfolio our portfolio shared portfolio right here uh great work and what they do is kind of an interesting one like they use llms and they have these data sources and the llms are ran in a deterministic fashion in a way that they actually decide the outcome of the thing that is injected on chain to do resolution of if the Oracle went one way or another so that's a cool use case there's like very simple utility of a concrete model that is ran in a specific way that gives us utility to crypto that came from AI so it's kind of fun because we're always talking about how crypto can enable something in AI when in fact there's lots of these School use cases in which AI can enable something in crypto I think like at the end of the day like crypto and AI going to be invisible right so they'll be just being completely abstracted out and and then so I think the fact that we are still debating about crypto and AI is like both of them are not being invisible enough right so if like they're completely invisible because I traditional ml has been used by all this email filtering and all that blah blah blah right like those are also AI as well like they're completely invisible and like used like every day and I think we're still here debating is just say we don't have um like the adoption is definitely not there like as like as aric said and then like the early pmf like might be there and then so because I every like you mentioned that okay lrm like when it comes to lrm so yes I agree with you foundational model like cannot compete with all the fan people but when it comes to MLM like median language model not that MLM and then also slm and I think it's actually possible uh and then also like crypto is really good for like financialization for imagine we have like initial coin offering now we have like initial model offering for all this like small model like like uh like Builder and like median model builder and then so that can actually just like take a chunk of like the economy um I think like that's also like a possible future as well uh because we internally and then we deploy local transcript model because like we don't want to upload any recording to the cloud right and then so like we want to tip that specific model like open source model author but like there's no way that we can tip because we use that every day right I'll flip the page a little bit we've been arguing here for for a bit now if we assume uh that AGI is is near and eventually there will be AGI era what do you guys think the impact on bitcoin etherium and some of the other major coins will be and more specifically is there some difference between for example proof of work that we see with Bitcoin and proof of stake that we see with ethereum if we assume that AGI exists and is functioning yeah I'll I'll take this one um so uh if we go into an AGI arms race an intelligence explosion we should assume that Nations this will become like a nation state core issue cyber security Warfare it will be a military in uh mil military Tool uh foremost and it'll be used by that for that uh for that purpose initially so and and if you know China and the US are like competing like who's reaching AGI or super intelligence first it's going to look like you know the US versus Germany during the World War and you're going to see a ramp up in investment into like creating more powerful models that might look like the race for nuclear power so I think that it's very likely if it if the if the whole world like goes into a full-blown AGI arms race you will see massive Fiat printing at the scale of um what we did in the second world war where you have countries like um UK US Germany and France borrowing uh 100% of their GDP and massively inflating their Fiat currencies also the trade relationships break down because they're not in competition with each other they can't trust each other anymore and in that world like are you going to start like using gold gold where like gold in a specific vault in one of those countries when they're not trusting each other no so Bitcoin might become the only instrument that you can actually trade during the with during the AGI arms race but then you got to question yourself well is actually Bitcoin a good currency for that if uh uh bitcoin's Power us usage is like 1% of an AI super cluster if Bitcoin is like this fractional energy uh consum consumer compared to what the AI clusters are using it's not even connected to the main arteries of energy on the world in the world then Bitcoin might start to look kind of weak because you could like easy easily overpower that and if you say well well you know it's not about the energy arteries it's about where the chips are located well then think about this so most chip Farms Asic chip Farms are centralized in specific regions in Texas and somewhere in China and so on and so forth but you have $500 drones destroying $5 million tanks in uh the Ukraine Russia War right now $500 drones destroying $5 million tanks so if you believe that Bitcoin becomes the new infrastructure for uh for Commerce for for trade those Asic Farm facilities are going to become targets for terrorism and if you can blow them up with $500 drones well maybe proof of work is not the consensus mechanism that you know guarantees safety for this decentralized net Network so I'm glad that I'm finally able to make like a bullish case for ethereum and proof of stake based currencies like someone has to do that someone has to like you know make a stand for ethereum and proof of proof of stake in this like if you look at the ebtc chart someone has to go in there and say maybe AGI is going to be the Saving Grace for for proof of sake based currency so maybe ethereum has a chance there that's what I think what probability do you assign to that scenario well so I I actually well I Let's do let's do like a raise of hands like who thinks that we're at the cusp of a of an intelligence explosion in this room like like Tom Horizon was 10 is that in like a decade in like two let's say let's say put AER let's say like who thinks that we're going to reach uh AGI within 10 years 10 years nonzero yeah non-zero chance okay just you know this is so hard because look it's way beyond our Event Horizon you cannot see let me just give you a counterargument why with AGI all blockchains are useless every 20 years we go through a dramatic change in cryptographic Primitives if there's actual AGI just needs to be 20 years ahead of current cryptographers to break every single hashing mechanism signing mechanism and cryptographic mechanism that all these blockchains uses they literally like can make it useless uh they can make if you can break it can create collisions of Shot 2 56 they can actually you know fake history and we can't actually govern ourselves anymore and so that's why it's so beyond the Event Horizon because if you truly believe it's AGI and it's much better than humans then our cryptographic me mechanisms are just screwed they they were made by humans computers are much better at attacking them so in that version of the world blockchains are useless but there are some moments going into the Event Horizon where we're still got to do something so there's a small moment there like between AGI and Asi before like quantum computers break all the cryptography that's where you got to you know why not plan for that for that moment like get your bunkers get get your so like that's why I think like many of us are going to just eventually become like longevity people right so it's like in between if we can fix like longevity and then so probably like we can fix like all the problem all together so why why don't we then ask the the panel itself do you guys think that AGI will happen in the next 20 years Way Beyond my event horizon I hate making predictions because I I hate being wrong that's my prediction it could but but even when it happens we we won't exactly know what it'll look like that's the problem is some people like even aert like right now we have AGI or right now we have ASI similar to Quantum it's just this idea of we also don't know if it will be malicious if it'll be just superpowers subservient making us more powerful it's not social like computers are not social so there may not be a drive for dominance or hierarchy in that way it could just be a massively massively powerful computer so um I think it'll it could happen absolutely the next 20 years but that's not so much the time frame as it is when it's here what does it look like um so so I was looking for so when been talking about like the actual emergence of like AGI so I was looking for only one thing um we invented plan it is not because of like how we made make like the birds anatomy and then just try to see their bone structure right we invented plain better can fly so like we invented plan like which can fly better than the birds because we discover like aerodynamics right uh at the mathematical level I don't think we have that thing yet for AI um so like that's also the reason why um like open eye like the way how they train model was not being very like appreciated by the academics because like they considered that as like a brute force and so I think if we can figure out that simple and beautiful formula for AGI whatever and then so like that's so like that's I will be thinking oh hey so we probably have to face that Singularity uh so for now like that's why uh we spend a lot of time just like digging into like research and at the math level um we haven't seen anything yet there um and then like also currently like all the input is so it's just like tax data right but the thing is that in order to create like AGI you need to have like a word model you need to have like spatial data right and so those are the things us still like far behind and so like that's what I'm not too worried about in like 20 years yet to me it sounds too bearish I think that basically you know AGI the test we used to refer to was the touring test I think that AI today crushes touring test yeah but like touring test was like 50 years ago right but that's when we sort of like if if it solves the touring test then we should really start worrying and I think we're way beyond that point I think arguably we already have something that looks kind of like AGI already today I think that the path to getting something that's really AGI like is could be less than three years I think that the timelines here are extremely compressed and we we know that neural Nets are uh mathematically they are uh Universal uh function approximators there's there there's mathematical theorem from the 80s from the '90s that prove that with a with a neural net you can approximate any continuous mathematical function only thing that you got to solve is like how do you train that and how do you get the the data for that and I think the word we're on the path of solving those problems and I think with the the most recent model uh the 01 preview model starting to generate that synthetic kind of data that helps uh helps create the synthetic data so that you can create gp5 so we're going into we're very closely getting to that path where you're going to get automated AI resar from a then you just like the inflection point just goes exponential so the premises is uh so the premises if we can like completely understand like neuron netork is that if we can decipher the brain so like that is at least my assumption um but we are far from decipher we don't understand brain we don't understand AI but somehow they still evolve and they still emerge this is a new life form that is emerging it won't understand how itself works we don't understand how we work so we don't need to have it all figured out in order for it to happen well so in order to make like cryster work and then you need to have like f know right so like that's like so you always need to have this like sequence of like scientific event like we know how to train AI models but don't know exactly how they work but still they do they go out and they solve problems in the world in the same way that you don't know you you couldn't decode your own genome but you could uh still go around and and kill mammoths with Spears so you don't need to know how your own intelligent structure works in order for you to be generally intelligent anyway I'd love to talk about this for a lot longer but we've run out of time uh appreciate so much the thought-provoking debate discussion thank you so much for the audience uh y that's it thank you sry