The intersection of AI and crypto w Anand Iyer

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:00

   nothing we say here is investment advice legal advice life advice or representative of the views of the organization we are part of or affiliated with see chain abstract.com for more information hey everyone Al on hey nice to jump on man the first podcast post the big uh nuffle launch this is cool so if you could just give a little bit of your backstory of those you maybe not the whole full 25 years

2. 0:22

   less you want but where where where you did and where you are now type stuff yeah yeah um I'll run it by real quick but yeah I have a degree in computer engineer Le in from Purdue uh I spent a bunch of time big Tech from Microsoft um I was I've been a developer and engineer you know for as long as I can remember I started coding when I was really young you know I was I was running code in this language called basic back when I

3. 0:44

   was eight this was in the 80s but uh yeah you know spent a lot of time to developer platforms of Microsoft um i' I've been in silon Valley since 2001 so I moved here right after college and um you know I I I saw sort of the birth and and growth of everything uh Silicon Valley through the 90s and the 2000s and you know very much I again I think it was like forward manifesting all of this I wanted to be a Founder so for the first 10 years I was in big tech for the next decade of my professional career I was a a startup founder I was a

4. 1:15

   CEO founder and um without boring you folks with all the details it was kind of through an accident that I met um a bunch of folks were mining Bitcoin back in 2012 and that's kind of how I got into the world of Bitcoin and crypto it felt like this you know I'm I guess the analog that I like to use is that I'm a I'm an immigrant and you know like a lot of people like to find Solus in their communities that they feel comfortable or close to and crypto felt like this

5. 1:39

   very burgeoning early community of just like Rogue Tech enthusiasts and I was like okay this feels like a you know crowd that I can be a part of um but yeah you know stayed close to the ecosystem again you know my my my company went through we went through our roller coaster of ups and downs and fundraising and and growth and eventual acquisition and I decided I want to do this fulltime and and so here we are now I I run an early stage fund called canonical um where yeah we focus on

6. 2:06

   early stage investments in great companies like Alton's a great founder and a great example uh and yeah lately have been focusing a lot on on sort of the interu of AI and crypto yeah it's a it's a great story I mean getting that handle now it should have been if you were to try to get that handle now there have been a lot of uh bidding on that one um so I love to I love to basically break this down I mean we can start with um a bit more Broad

7. 2:35

   and get a bit more in detail so like now are you sitting in the market and say Hey you know I'm I'm basically very interested in new section of um crypto and AI what does chain abstraction mean in that world like is there something that you know can we do you see an intersection of crypto Ai and train abstraction um and yeah we can just Deep dive start deeping that from there yeah

8. 3:00

   you know I I was thinking analoges right so um and one sort of constant has been that the learning for me has been that history doesn't always repeat but it Rhymes and so like I use that as sort of the backdrop and I think about like what does chain abstraction mean to us and this feels like a topic that's come up a lot lately uh but it's almost NE necessary right because if you look at the progression of blockchains I don't

9. 3:25

   appreciate the choir here you folks are the experts but as an outsider I like to look at why this is essential today and and what what's brought us here but you know for me like the biggest sort of unlock back in the day was something like kubernetes um I'm dating myself a little bit but kubernetes which served to sort of an abstraction layer between various Cloud platforms and host of applications right it helped you deploy and manage and scale um across various

10. 3:49

    different Cloud environments um and so this is sort of analogous to that now you look at either developers or you think about users and you think about the necessity for abstract action because maybe you want to take advantage of capabilities on one chain versus using accounts on another chain this is where abstraction becomes Paramount so just like the backdrop now maybe to answer your question more succinctly I

11. 4:13

    think when it comes to AI there's there maybe a few different touch points right so there's the user there's the developer POV which is one demographic there's the user demographic which is the other um from a developer standpoint I think Alton you touched on this a little bit which is can you know know will a developer need to learn the minutia of a specific chains programming

12. 4:37

    paradigms um wasm versus rust or something else and what can chain abstraction do to sort of ease that or can you use tooling that's llm power to be able to deploy smart contracts more efficiently so that's like one way to think about it um the other is sort of user interactions right and this this can sort of transcend into agentic interactions as well right like uh I I think the velocity with which users have started using blockchains which is now morphed into I mean obviously there's

13. 5:07

    like some amount of Bing farming that happens that that we all are aware of and then now we go sort to an exponential level of like usage which is I say we lean into this right like I don't think we should run away from this Bing I think it is inherently what blockchains have to offer uh and I think this is where agentic platforms come into play right like they will need to interact and they will need to do things on behalf of users primarily and then eventually Maybe for themselves uh to

14. 5:31

    take advantage of what different chains have to offer so um that's where I think chain abstraction becomes even more Paramount I we are kind of graduating from you know interoperability at as sort of like one Lego block to abstraction as at another so that um whatever whoever the end users developers users agents can can leverage various capabilities various chains very very easily and succinctly so um and I

15. 5:54

    think to me there's you know just to to wind this back and and put a bow on it I'd say there's two ways to think about it the first is um you know who are the users are the agents are they sort of you know AI Bots that need to be able to use these platforms the other is how do we make this tooling more easy uh and and usable and customizable and I think that's another sort of demographic to to tap into I think great response but I

16. 6:17

    I'm picking up on this AGI agents you're actually using the AI agent as a user too and I think that's like this is where it starts getting really interesting right so will it maybe be like like so blockchain is a good way of proof of source like proof of truth let's call it you know like a moment that is like immutable on a blockchain then you have ai which is almost like Infinite Creation and I think do you think they

17. 6:40

    need to work together in order to sort of like re each other in in a way uh yes and no right like I I think that raining again that's got like such a strong connotation when it comes to how we think about The Ether of blockchains um and so I I think there are guardrails there are sort of modes of operation that Mak sense but but at the same time and I think this is up to us to Define but let me maybe even take a step back right which is you know knowing how these are

18. 7:12

    constructed and how these agents are constructed how they're operating and the fact that we want to be able to audit and verify them is like maybe most Paramount and today like we have very little transparency into close Source systems and how they operate so like let's just set that precedent that we want to operate in a more open transparent verifiable auditable way like that should be sort of like the you know the basics and building blocks on

19. 7:36

    top of which you build anything and then the second is guardrails right like what are the parameters under which these need to operate so that's like the reing in to some extent um so I I don't think we want to re in The Innovation part of it I think we want to re in the operational auditability part of it right so I think that's what they trying to say jar so yeah I think they they go

20. 7:52

    hand in hand for sure um and uh you know yeah I I think I can imagine a world in which today there's a lot of lot of thing I mean you know this is the beauty of like machines and the progression of like dating back to like in I mean I've been you know I I don't want to go off on this tangent here necessarily per se but I've been really digging into the sort of the philosophy and growth of intelligence over time and you know because we've been talking about you

21. 8:15

    know machine oriented intelligence like the need for replacing limbs with prosthetics and things like that for forever like before before these even became you know before medical advances even started to happen right so we've been thinking about like this concept of Robotics and and things like that for a very long time and philosophizing theorizing on it for a long time uh but these have all been to Aid human capabilities right so that's that's one

22. 8:41

    where we see and you know again we've been theorizing about this maybe for a more brief period of time maybe in the last three to four years maybe even Ilia has been talking about it which is you know there's a world in which agents think and operate somewhat autonomously and how do we give them the rails to to operate and and behave there and that becomes fascinating to think about but that's where maybe you know again you

23. 9:05

    build The Ether of transparency openness verifiability and then you give them guardrails that becomes a you know a good way to get started yeah I just think like in the future it's going to be me with my team of AI agents right like I do a lot of content so I'm gonna have like my content agent team right and they're gonna actually get to know me better than myself like I'm going to be talking like I'm already every morning talking to Ai and help me kind of like organize my day and stuff like

24. 9:31

    that's just going to increase so then a question is is it important to have like you know n's doing a lot of user owned AI vers maybe corporate owned AI do you think a lot about in that space like I kind of think like these things are going to get to know me really well I kind of hope it's user owned Ai and not something else but I don't know I guess from a professional like you what do you think absolutely it's it feels

25. 9:56

    inevitable in some ways right so like I mean just to keep this a little bit more it's funny how some of these things sounds so science fictiony but just like you said like you start your day and everything was starting to realize the potential of AI to an extent that we hadn't really we had only theorized about for a long time right like this is you know Neil Stevenson's the diamond age where they talk about these you know there's this uh character and she grows

26. 10:22

    up with an AI right like it's just it's sort of like the companion that uh that Nel's been Tethered to so uh but this going to become natural and that is you know there's that's great that's our steady state experience like we want this to happen like you know agents that help you perform various functions agents that maybe can uh like you know classic case in point I had someone messaged me and telegram last night and

27. 10:46

    I was like I would love to you know what's what's the what's the right answer there right like how you know of course I don't want it to be immediate but the same time I wanted action on it if it's going to change my day for some reason so how do these things act when not able to because that's what machines have been you know designed to do to help augment human behavior to a large extent uh but also then how much responsibility do we hand over so that they can start really working for us so

28. 11:13

    these are some of the things yes user own a in evitably where things are going because you want to have agency you know no no pun intended over how these things operate because they are going to mimic you your your concept of a digital twin for example like things like that getting more and more popular today but uh to be able to do that yeah you really need to understand who you are and you need to they need to operate in a privacy Centric fashion as well which is Paramount and again all these sound

29. 11:38

    exactly like what we've been preaching in the world of blockchains for for over a decade now so I I feel like these these two technologies on a collision course yeah I have a I have a question there so I guess like imagine a world where everything is done by agents right I mean that as you said they have their own agency they do things um and we we we talk about this a lot in the space that I'm probably at this point an expert in which is rollups right how does the security come from if you're doing something how do we how do

30. 12:11

    we translate that security to do something on on a on a offchain and improve that onchain um how do you think about that for agents like is this is this a world where we we drive security from um from cryptoeconomic values or we do this in a more of a Zone knowledge fashion where we we do this cryptographically that's a good question um you know maybe there some I think in these sorts of situations use cases are like you know ways to decide like what

31. 12:42

    the what the flowchart is going to look like in some ways um the the reality is that there are a lot of situations for which we can see agents operating autonomously using cryptoeconomics because that's just going to come more naturally like it's just more native the need I think we know this right like anytime you feel like you need to use something that has to go offchain or use

32. 13:07

    extraneous environments to operate on chain that just adds a level of complexity that I mean to we've solved this using oracles we've solved this using you know provs and offchain compute like we're we're getting better at this at this um but it adds some extra complexity so I think crypto economics seem more natively the way to get started but there may use cases for which like you know we we may have to think about offchain

33. 13:32

    processes as well and maybe I'll just throw in another sort of wrench into this equation right which is there may be situations where agents again have to operate um and again there's there's the user a agents there is the operational modality of these right and those need to be codified and verified and I think that's where blockchains come in um and then they need to transact how are they going to transact with one another right

34. 13:57

    like what does that look like and I think the question that's been asked many many times now is even think of these agents as like humans are on earth um what do their bank accounts look like right so and what they don't know what a bank is right so that this is like wallets and like you know crypto is going to be first nature to them um and uh there may be situations where like this is where the rench comes in right

35. 14:22

    like they these agents will need to operate without enough information and therefore will need to lean on humans to supplement information they don't have right so this is where offchain compute becomes even more real so like I think Al said this very Sly remember September 2023 um at a side event we hosted where he was like we're going to be working for AI in in the foreseeable future I think he projected this year by the way

36. 14:45

    uh I don't think I can hold him to that but I think it'll start to happen right like you know agents are going to put out uh bounties to say humans like go get this information for me because I don't have it and then they're going to have to pay us so because they don't have enough information to work with so it's going to be pretty fascinating MH whoa I just like I I really like talking about Ai and crypto because then we get into this level of

37. 15:08

    conversation where it's like an AI might eventually reach out to us and then it's going to be like yeah it's wild it's pretty it's pretty nutty to think about um and it feels like again it feels very Science Fiction it's like you know sounds feels uh dystopian to some extent but I the rate rate of progress of technology is is unlike anything that I've ever seen uh and um and I I won't I

38. 15:33

    won't spend too much time I don't want to sound like the uh you know the old person in the room but you know I I I think I was siging alt on this I took a there's this program language called list lisp uh Lis programming which is uh you know it's been sort of like the Cornerstone of everything AI for many many decades I think it was invented in the 80s but I took a programming class on Lis in 1999 and that was like it was like marketed as like the AI course

39. 15:56

    right this is 1999 by the way and like okay yeah is imminent it's going to happen robot's going to take over the world we're going to live in like a Jetson's future where there be flying cars and like that's the world this was in 1999 and so like because the internet had just started up and this is where we thought things were headed as quickly as then um what we realized was missing was you know obviously like the the growth and progress of AI you know there's a

40. 16:21

    pre-transformers era there's a PO there's a post Transformers era but data was really missing data needed to exist to the point where these models need to get trained adequately to get to this point um I I I I bring all this up because it it is really fascinating to theorize and it's always felt like science fiction and it's always felt like it's eons away but now more than ever it feels like it's right around the corner like you we could be talking about this and and it's going to be very

41. 16:45

    real very very quickly uh so yeah I'm I'm super excited I think it's a um it's scary to a lot it's exciting to me as a a pure technologist and and maybe also maybe very naive operational modality but yeah it feels like we're going to be uh we're going to be hyper productive and uh working for working on things that we really really naturally care about because everything else is going

42. 17:08

    to be you know going to be done for us hopefully by by these clever user own models and agents yeah yeah the other the other thing is like okay so we have obviously have all these like centralized companies you know if you look at data collection um and and curation it's it's mainly done by scale um scale I think is company and then you know models are trained by pretty much open AI Google and all of them they're very centralized um if you look at this pipeline I mean gpus are owned by one

43. 17:41

    company pretty much that's Nvidia um how do you kind of how do you win against this much capital and um this much sort of power right again and how do you decentralize that I think that's kind of um yeah how what what can some do I mean Elia is a good example of someone who's trying to do that uh from our community who understands obviously both uh but but how can how can we win that and what can someone

44. 18:09

    do yeah it's um I think this is the question right and I I feel like we need to spend a lot of time thinking about this so the way I I would position it is like again we've learned a ton from the last couple of Decades of of living and growing up on the internet right so um there are like that I mean you know the Google for example is a great you know is like one cornerstone of like if I need to access information on the internet like you don't think about anything else you think about Google right like I need to go Google it

45. 18:41

    right that's it um and of course that's been the incumbent model but Google is an interesting business model right like they although you would search something on there it would redirect you to a page or content that would get you that information this level of intelligence we're seeing right now is trying to make it such that like you can access this information right without having to leave that particular experience right

46. 19:05

    uh chat GPT and I don't know if you saw last week uh it's it's the week of the 17th of June 2024 right now but I think it was last week when uh a few different reporters started to call out perplexity because perplexity was creating summaries of news articles um wordsmithing what was being written on other articles they're like wait a second so this is interesting this is not just like backdating information this isn't just Wikipedia esque but this

47. 19:31

    is realtime information right so there's examples of complexity or or Gro or like you know which is Twitter or X's platform where you know they can give you it's kind of like you know rag happening on top of uh you know existing data modalities which is really fascinating so um yeah I I I would say that it is yes powers aggregating to the centralized entities but you know I'm trying to break down the problems that exist which is a we're moving into a centralized world there's no transparency here you know again there's

48. 20:02

    many group chats that we're all probably a part of where there questions that are arising about the level of privacy associated with these centralized Services right so I saw this Reddit post uh the other day which was there was a user who had plugged in his nest camera into his house in his apartment and he was collecting these video Snippets running them by GPT uh I think it was

49. 20:25

    GPT 4 and it was saying hey by the way like where are my keys or where did I leave this thing that I in my house right so and that was happening almost in real time um and I was like this is this is cool but yeah but it's so it's wacky to think about because a there's like you're using again a centralized server for like what's happening inside your house um you're one hack Away by the way from like you know your data getting

50. 20:50

    leaked which happens I mean like what do we hear about the doy sign leak which was like do I think I apologize if I'm not if I've got that mistaken but you know some of these companies have one job right which is like you want to maintain your users's information without it getting leaked and like when that happens all hell breaks loose so imagine a world in which again you can tap into like hackers are breaking into video cameras as placed inside the house and that has happened by the way you

51. 21:13

    know this is this is fact um then you combine that with running it through a centralized you know service and again you know by default by the way privacy is not turned on right so like you might opt in providing some information back to fine-tuning those models without even you even know you're doing this so these are all sort of existing problems that exist so let's come back to like swinging the pendulum from centralized services to user owned or you know like

52. 21:39

    a a collaborative that helps you manage these models and operation functions um and yeah let me let me put it this way right instead of answering the question for you do you want to see a world like 20 30 years from now where you grew up with this Jared you said it well right like I have agents working for me or like I want to have these bunch of things happening for me how much of this do you want controlled by a large centralized entity that's

53. 22:03

    trying to monetize your data versus you feel like you have ownership over it um and agency over it yet you still get to realize all the technological gains so that's the question that you know sorry for the longwinded Preamble but I think it just comes down to that ownership data privacy and agency yeah so I mean yeah uh one one question there is like one thought is you know we data privacy existed for a

54. 22:29

    while right I mean like Google tried to do this Federated learning 2017 where they just said hey I don't want to extract your data I just want to I just actually want to train the data on your device and send it back to a central server where I don't see your data anymore um yes but again right like my grandma or my mom doesn't care about this stuff so I think like do you see a diversion of like because if you if you read a lot of these sort of thing about

55. 22:55

    I read this book called homad do where people are going to be very you know like in terms of how um people who are thinking about it or not thinking about it are we going to diverge as a society to say some people they won't care about it People Like Us who are pretty techsavvy I would say right like anybody that works in crypto is this is this the future of how this is going to look like and

56. 23:19

    yeah it's a it's a valid point I I would say that users the average user doesn't care about it absolutely 100% like you don't you know I mean forget you know I mean I with all due respect to all you know all our elders and everyone I think there's people are immediate circles even the tech circles who don't care about this today enough um but I feel like it is we're going through this wave of Technology Revolution Renaissance in some way where it feels responsible for Developers for us to think about this um

57. 23:49

    because this is we we lead the way here so I think alter goes back to your question right which is like we didn't start to think this way but do we want to continue building on the same Lego blocks of just like you know being able to trust centralized entities with power aggregating towards these centralized monopolies that monetize our data for their own good and their own use or do

58. 24:13

    we want a sense of ownership when it comes down to it at the end of the day right so classic I mean more more pressent example is that like we talked about these data Dows or like you know this this is becoming a popular concept as AAL Vanner for example that's like thinking about these these data DS um but you know wouldn't it be awesome if like this exact federa training modality that you mentioned if I felt some level

59. 24:36

    of agency over it like I would feel comfortable like you know I think Fred Wilson from Union SC Ventures talked about this maybe a year and a half ago like hey we've we have we have these Primitives like wallets for example using which you can sign a transaction and or like you know permit it to do something on my behalf right sign a transaction and something happens on chain what if we could use this analog to that to be able to hand over my data and then when that happens that's the

60. 25:00

    beginning of like some process that needs to to happen right so like you know again attribution or like I contribute my data or I can encrypt my data or or you know maybe I can you know cleanse it such that it removes any sort of pii and then it's being used for training um so these sorts of things aren you know we just need to operate from those levels of first principles and then yeah my ideal steady state and I I'll I promise I'll stop rambling

61. 25:26

    after this which is I think you know open AI as we all know is not open I think we we really really and you know any other other models out there I would say same thing with llama and others I guess the question always comes up for me is that like this is awesome they're open weights but where did the data come from how are these trained how does someone get compensated for the data they contributed to these ecosystems should they be compensated do they even opt in um and then where are we at today

62. 25:51

    and like we should we should all be proud of using a system that we're exceptionally proud of right and that we have transparency and agency inter um and but I it doesn't start with the user the user doesn't care um and I I don't think they will for the foreseeable future although it's like I think current cohorts probably care about this stuff a lot more than maybe our our older Generations do right like I think you know my daughter and like the generations of today care a lot more about let's say climate and things that nature than maybe we did growing up

63. 26:18

    because it's it's impacting them and their future right so I think the current Generations will uh the older Generations may not but let's also lead with that precedent that we just want to see these princip you know build on these sorts of principles uh and and create a better future for uh for ourselves dude there's so much to unpack there but I really like so here's what's awesome so okay so maybe privacy data might not be the thing that gets people to do it but I do think that you said something there like value and what

64. 26:45

    blockchain is really awesome at is like creating value or or like making value appear in a way or catching value that wasn't possible before in a centralized way and then I take that we've got this web to internet for the last literally since Myspace is just sort of like farming humans data and making profit that's sort of like the business model and it's we just all accepted it but now we've got sort of this revolution Renaissance I'm kind of just re like

65. 27:16

    like paraphrasing what you just said just to make sure I'm kind of on the same way but like now that now we can actually create a way that people's data or whatever they want to do they can have the choice hey keep it encrypted don't mess with me cuz I've got my a agent team maybe I don't want to share that info or there's a way to leveraging blockchain because it's got that wonderful way to transfer value for low fees and lots of hexad decimals and all that that it could now you could

66. 27:41

    actually like opt into being part of the the the AI system and getting actual value for it and it's possible with web 3 and AI I'm thinking that's kind of what I'm just coming at so there's no real question here I'm just like my mind is being blown that's I think the question here cuz it's pretty wild to think about because I've just given away my data for 20 years yeah we we all have right we've

67. 28:03

    taken it for granted and it it feels like there's I'm getting more and more conscious of this now than ever before right like I I don't use GPT models and this is just me n of one here so take with the grain of salt but uh I think I was saying alt on this maybe last time we chatted which was I'm doing all my infs locally now on my machine and have you know have an M2 so it's capable of

68. 28:27

    doing some heavy duty uh you know inference and I think maybe this stems from a little bit of paranoia but also um you know I I remember my mentor told me this in the 2000s it's like when you're a developer you need to have something called healthy paranoia right so this is like I always operate with that so I feel like it's somewhat healthy paranoia but it's also something that I want to see out there in the world right how do we give agency of to to users to be able to do this more on

69. 28:51

    their own it's a company you know like these guys called EXO for example I know Alton's pretty close to this team but they're a good example of this right like enabling inference at the edge right or enabling the ability to do to really like manage and operate with your own data in a more codifiable way of course you know we heard about you know Apple loves to Brand everything as apple right so now they're calling it Apple intelligence which is you know a lot of things will happen on the edge and and you know they'll they'll back they'll

70. 29:16

    default to chat GPT or GPT or other models you know if they can't get it done on on on the device but uh you know it's just a toggle away from from becoming you know from being able from being shared with centralized services so yeah I I I react and I I 100% you know agree and empathize with with what you're saying and yeah you know I I go back to Alton's question like do we need do we need this do we need to think

71. 29:40

    about this and I would say it's a uh it's worthwhile Endeavor to to manifest our way into this you know because it'll be worth it at the end of the day um and whereas like yes you know will incumbent I I think both will exist 100% I I I don't think this is a replacement this is like the open-source argument against Clos Source Windows this is the you know very this is going to be a very cheesy example vot to say this is like the Android to iOS right so we we will have

72. 30:11

    these comps that that just need to suit the modern day uh requirements that that are emerging because of intelligence yeah it's not that cheesy example but um there's a lot of folks that came into the space at think crypto AI space just because AI was just hype and they just you know they just wanted to extract value out of it um how do we so now that looking into from an investor perspective um leaving that aside like what do you see that's the in the short

73. 30:44

    term that's very valuable for for crypto Ai and uh and the long term as well that you know I think leaving the short time maybe it's not really um it's not really relevant for folks but it's something that can equ value and do something you know have a have a dent in the in the broader AI space Not only in crypto AI yeah um I I would say that you know maybe just react to your first comment which is the the the history of like in the brief history of crypto has been that

74. 31:20

    like it's rarely Winner Takes alls Winner Takes most you know that's been sort of like the precedent that's been set we have many layer ones that are great great we have many layer twos that are great we have a lot of layer threes emerging we have chain abstraction which is hopefully going to solve all you know of like the user experience issues which I know is like the fundamental theme of this of this podcast um and yeah it was

75. 31:42

    probably not unnatural to think about you know many examples of solving the you know the same problem in a differentiated way with token and cised uh mechanics so it started with compute and the need for comput sort of bare metal access to compute um and akos for example has been doing this longer than most right and they are uh by and large I think one of the more reliable systems out there when it comes to uh sort of access to bareal compute it can be run for training can be use for inference whatever you whatever you want

76. 32:15

    to use it for uh but then like you start to move up stack or you start to use more software Centric modes right like very API Centric modes not like you know again the loose analog is something like you know Heroku versus is AWS right um but yeah you want to be able to deploy something and up get up and running quickly so we're going to have to start to move in that direction so that's kind of what's Happening Now um so maybe more specifically to answer your question right I think we're starting to see a

77. 32:40

    lot of lwh hanging fruit get picked off which is access to compute and let's just put this out there let's get this done very quickly and there'll be a few winners there and maybe you know like a long tail I don't think it's going to be a torso it's gonna be a head and a long there be a very very long tail of things that that will exist um so I'll answer your question somewhat indirectly all time by saying that I'm

78. 33:04

    very excited about differentiation defensibility like the these are like the most important things today and I I'll also you know I'll also credit a little bit of like pedigree so if you've been spending time doing more time doing AI than crypto and you find something that's like highly defensible and differentiated to bring to the crypto ecosystem or use crypto economics or crypto Primitives to to substantiate building decentralized AI those are like

79. 33:28

    super valuable um so I I don't think there is a specific call to action here about short-term versus longterm I and I know this is you know Al you and I have known each other for a little while where it is for me it's really about the people and bring and bringing credible backgrounds into this world that actually matter to me the most and it may be shortterm it may be a Band-Aid today but as long as it's symptomatically fixing something that's

80. 33:51

    going to exist in the Long Haul like that's going to matter you know more so long way saying that I'm not I'm not I don't have short-term lens it's all a long-term lens and would love partnering with just great smart people who have a very opinion point of view and building something that's like very very very hard to do um and and that makes it that makes for a differentiated product in gr

81. 34:13

    Market man I I don't want to just wax philosophically we'll make sure to cut that part but my question for you man I always like asking people this so in your industry your day-to-day what is one area that you think more attention should be on that you're like I see it but you don't think think as many people see it in the AI and it could be AI crypto either way yeah um I know we touched on this briefly I would say it's data for me

82. 34:37

    it's a it feels like it's the um it's the oil in some ways right like it's it's necessary for functioning of large scale systems and we're taking it for granted to to some extent um it's going to continue to be a problem because I think Alton touched on this right like these incumbent centralized systems have sort of run away with large models sitting large models today um and so if I had to

83. 35:01

    get a if I had to get you know if I had to get up to par with where they are it's hard because this has been you know 30 years of data in the making with smart sophisticated systems that have scraped the crap out of the internet and can have have trained themselves exceptionally well so data is something I think about a lot so there's several offshoots of that user contributor data

84. 35:23

    user provenance like hey Jared's data what does that mean um and then how how do I like from again there there's the user POV there's a developer point of view which is like I think like a terrible example again is that like you know I would love to run I I this is this is like hypothetically you know impossible but like not even hyp was like practically impossible but you know if I run a model I would love for it to

85. 35:48

    if if I run an inference against model I would love for it to say like what are the differ sort of data sources that contributed to this inference it's kind of like saying I want to run this run this Dash Dash debug you know like in debug mode where I want to kind of figure out like what is the source of this uh what is the source of inference like why did why did you tell me what you just told me you know like because it's not like you're just pointing me to a website there was something that contributed his intelligence and I want

86. 36:12

    to understand what that is because that will the and there's so much that needs to happen underneath for that to to become the so long W saying data is one thing and then that's like here it's it's like now here and now the preent maybe and I'm not trying to you know sound eal gives myself too much credit by by any means but like in zuck's interview with dwares for example the one thing that got teased out a lot was energy the need for energy um and how

87. 36:37

    you know we're we're just taking this for granted right like centralized services are just like running away with this and like using like you know gigawatts gigawatts of energy just to be able to train these models like because all these gpus are running at at Full Throttle to be able to do this kind of work and that is just training and then we have inference which is a whole other

88. 36:54

    you know ball game so that's another one that I think about and I don't know where like there there is an there's probably an answer here somewhere about like where that meets meets sort of like blockchain infrastructure um but uh I I feel like this is the beginning of like unlocking that piece and so I'll just I'll just throw that grenade over and and let someone maybe catch it or run

89. 37:17

    away from it awesome man well Alton any other questions or should we wrap it up I mean I think we should if I if I ask another one it's going to get um um well so longwinded sorry no not all super U um what are your last thoughts I guess if if somebody were to listen this podcast um what what they should take away um about about anything about you about chain abstraction I about AI about your

90. 37:48

    Twitter Twitter handle um yeah uh that is an excellent excellent question I would say you know well just maybe two things one build for the future you know think have a longer long enough time Horizon because I think things are going to change a lot in the next few years but the the coolest smartest greatest Founders and I'm guessing your demographic is a technically technical audience maybe you know uh somewhat thinking about starting companies or

91. 38:21

    trying to follow you know full footsteps Alon but um you know put yourself out into the future and think about what the world will need in in the shortest term it could be I I don't know what the ter for is six months two two years things are going to change a lot because those people who kind of live in the future are uh you know are have an exemplary view of what to build and how to get

92. 38:44

    there and the second piece is um many and I think Mark and Reon said this very well in a more succinct way than than what I'm about to say which is yeah the best Founders tend to manifest their perception into reality so you it's your job to try to convince others to some extent like you don't don't make this a full-time job but like you know when you're able to evangelize and convince people that like

93. 39:07

    look you know I see a world in which there needs to be like a decentralized social network you know something like you know an analog to Twitter is what forecasters become right so and Dan has been steadfast about that for for the last couple years and so I think there are Visionary Founders like that who have a lens of the future and can turn your you know turn perception to reality and and evangelize that to others and so

94. 39:32

    if you're that if you're that archetype um you know obviously I would love to talk to you and and and learn about you um but also you know kudos to you for for being one step ahead and thinking about building something for our future so that's that would be my ask and hopefully my takeaway that's great man well Al tell I mean this is what a cool like interesting episode but uh thanks so much for your time and sharing your thoughts and being so thoughtful it's been really interesting thank you so

95. 40:06

    much thank you both if you get a chance to work with that on I recommend it's it's a um but yeah uh thanks for thanks for being with us I I hope we didn't um I hope we had fun I mean we it was just really fun episode uh to to talk about something completely different cuz we've been chatting about chain abstraction chain abstraction this is more like chain abstraction ction AI everything bundled into one so that's that's fun

96. 40:31

    definitely all right cool thank you so much bye thanks all