Legitimacy Is the Final Scaling Constraint

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

1. 0:10

   All right. Well, they have now heard way too much from me. So, so I'd love for you gentlemen to share a bit about uh what you're focused on in the space. Richard, kick us off. Intro yourself. Yeah, great to be here. I was listening backstage. I thought that was a a fantastic um realistic assessment of of where we are and where we've been as [snorts] we've all been kind of proponents of the capabilities of this this space. Um so Fabric is a venture firm that's been in existence for more

2. 0:38

   than a decade now investing in around the space and was uh doing AI before it was doing kind of you know web 3. Um, and I guess you know we've been investing in everything from robotics that include gamifi gamified approaches to collecting data for you know embodied AI obviously [snorts] through to kind of defi and prediction markets and all all of the typical stuff. But but I think what's interesting my um my background

3. 1:02

   was in building a couple of companies uh successfully sort of scaling across the the world and speaking to the the suits as you put it in in the enterprise um and trying to get adoption. And when I was thinking about this topic of um legitimacy, I I thought of it in a sense as a refinement of of the of the importance of being able to uh get customers to reference what you're doing um to get adoption. Um people here may not have kn thought much about the kind

4. 1:32

   of like the technology adoption, you know, life cycle, but when you're trying to cross the chasm, you really need to focus in and get that referenceability to drive adoption. And and when it comes to what we're trying to pitch with agents that can operate autonomously, you know, the bar is even higher to get people comfortable that this is legit.

5. 1:52

   They because you're giving, you know, that autonomy over not just automation like software used to do. You're actually giving autonomy uh to technology, the the capability to act on its own. So I think that's why the bar is ever higher on referenceability and it comes down to that legitimacy that we need to establish. I'm sure we're going to dig in. I think most folks don't even realize how high that bar will get with agents. Totally agree. John, how about yourself?

6. 2:15

   Sure. Uh I'm John Furelli. I'm a partner with Kinetic. Uh we're a multif family office based in Hong Kong and Puerto Rico. We have been around also 10 years. Has it been 10 years? Um time goes fast. We've uh done several hundred investments most recently into human humanoid robotics um AI but especially uh infrastructure layers that provide privacy trustless verification were early supporters of near um you know kind of dovetailing to what Richard was

7. 2:42

   saying I think we're kind of at an important inflection point where legitimacy means who are we trying to inspire into what action right um I think it's fair to say that the physics is pushing us into decisions that maybe ween't aren't prepared for on a weekly basis the updates. How many of you daily basis? What time is it basis?

8. 3:03

   Super conservative. Yes. How many of you are using openclaw, ironclaw or some form of an agentic? Okay. Yeah. Pretty significant. Now we are the early adopter group right? So when you start a you know talking with uh next time you get to Thanksgiving which is really too far in the future and you talk to your brother and sister your mom and dad about agentic commerce we're going to need to reach a level

9. 3:28

   where there's user level guard rails that protect them uh because some of the hacks that we hear on hourly on X with openclaw is not acceptable just as we've we have been very accepting of things like crypto hacks for a large number of years and that level of risk is not acceptable. That's a great point about maybe crypto hacks aren't looking so bad anymore by contrast like hey you know it

10. 3:51

    it didn't uh your entire system of your computer wasn't melted down so there's that um let's interrogate though the the premise a little more uh Richard to what extent would you say let's just matters like why let me back up why does it matter right is it like if a tech just play devil's advocate if a tech is good enough if the products are good enough will that be enough why [snorts] does it

11. 4:15

    matter that the powers that be and we can kind of double click on what those powers are but why does it matter there is trust and it can't just be a tech solution yeah well we do need that trust stack and and I'll explain you know why I think that's that's certainly the case first of all you know we should celebrate what has been achieved which I think you alluding to whether it's kind of like you know 1 million transactions per second that NRA has achieved um or

12. 4:38

    100% uptown time uh and you know I think that's incredible and obviously the crosschain operation and just a whole suite of launches that happened yesterday of different products I think that are really [snorts] coming together in a super cohesive way. So that's fantastic but that that is necessary but not sufficient for for hitting kind of prime time. I think there is another architectural component which you know near I think can deliver a lot around

13. 5:04

    which is to make sure that agents have a proper notion of identity a proper notion of reputation built on their kind of performance kind of on chain and could be other elements maybe even relating to who who built them obviously that's the whole point that can be uh you know attested on on chain and then used for may it can be making decisions about credit but it's also about making decisions about whether or not I want to use that particular agent and we see in the marketplace that near has um

14. 5:28

    launched um and then on the back of that some kind of um liability or accountability. Those are three kind of like quasi architectural components that I think are needed for us to move towards legitimacy because as you know the shift we're we're talking about as we as I said before go from not just automation but to kind of autonomous operation is a societal level shift. Um and so you know we we need to get people truly comfortable with it sort of metaphorically what it reminded me of is um and I was mentioning this to John

15. 6:00

    earlier is a kind of a comparison with Whimo. Um so I mean I've been really enjoying I'm I'm Whimo addict now in the last sort of couple of days and and they're already on the streets of of London. and I look forward to kind of taking them. But in in that case, it's not good enough that they're 10 times safer um than a you know human driven car or even maybe a 100 times safer. Um

16. 6:24

    you need the legitimacy comes also from a one way death everyone freaks out. Meanwhile, the time we're speaking there's been probably five human car accidents. Totally agree. Uh John of the the three in my mind there may be three different buckets of different types of institutions that legitimacy matters uh governments enterprise and let's call it kind of the public trust how would you assess how AI is doing in those three buckets right now where is it doing well where not so

17. 6:56

    much and where do we have work to do I think we're consuming the concept of AI from two different angles one is how does it help me do what I think I'm good at what I'm doing today and do it better and then the other is well how will it disrupt or destabilize what it is I'm good at and what I'm doing today right so there's these two different angles and one has very much an effectiveness and safety aspect and the other has what

18. 7:21

    is my position in the marketplace right and and I think every one of us is evaluating that right now there are few things you know we can certainly talk about things like government and regulation but I think there are a few hallmarks of we're going to make that we'll see. First one is the ability for the market to have some form of third party, dare I call it certification, especially as as you're using OpenClaw right now. Uh you know,

19. 7:47

    you can't really pull skills uh with PRs where you're wondering if there's malware, you know, like uh you may you may even use the best security practices for containerizing agents, for having contextual permissions. Um, but it's still really tough and even then you'll probably limit the scope of what you're willing to do with it up to the point where you're willing to lose it all.

20. 8:09

    Right? If what whether you're doing yield optimization or trading strategies or things like that. But the second level is when we get to that point where certification and security is that good, we'll start to see a market that begins to price that risk to the extent where we might even see insurance products that allow us to use these applications and insurance feels like this is we're going to need them. we're gonna need them. We're gonna have to have them, right?

21. 8:33

    So, I, you know, the certification leading into pricing and offloading risk is kind of where we start to see it. And, you know, to the extent that government uh wants to regulate it, I I think those are the things that are going to tell me whether we're there or not.

22. 8:48

    Well, I think one thing that one data point or anecdote that I think is relevant here is kind of concerns or calls for boycotting or blocking data center construction, right? Without delving too deep in the politics of it. It strikes me that like when that's a question of legitimacy, right, of of politicians saying, "Nope, we we want to put a stop for this. We don't like it." Um, and that in theory could really choke scaling, right? So I guess I guess

23. 9:19

    what does what needs to happen to avoid outcomes like that where there where there are um kind of blockages beyond what you would think of the normal course of business. Richard, any thoughts? I'm glad you took us directly into politics while saying let's not get into politics. But I mean I think the the only thing that will uh ever trump politics or you know sort of hand it to the electorate you just still

24. 9:47

    doing it is um is you know or patterning to the electorate is to is is of course economics or you know whatever this the the Clinton phrase it's the economy stupid. So, so if you have either you know interstate or international competition on an e economic basis um that is driven by the adoption of this these you know fully autonomous you know technologies that gives an edge over you

25. 10:12

    know one nation over another you know that will force people to to uh get off their kind of party politics and and and adopt things but only frankly if again if it's you know it's legitimately ready for kind of for mainstream where where I think the bar is is is super I um and you know to to your point I think that you know John we need the whole package there because nobody's going to come

26. 10:35

    along at the moment you know we're all hobbyists and early adopters really in this in in this room as you also mentioned nobody's going to come along and kind of like you know follow the guideline on the YouTube for how to set up open claw even ironclaw and and you know these Thanksgiving family is not doing that you're you're saying next year.

27. 10:51

    Yeah. Well I mean we we may but you know but on the average one is is not going to do that. It's [snorts] just going to be a kind of like, you know, let me click it and and and and select or install it or get it from a a service provider and know it comes with insurance and know that it sort of everybody says it works sort of super well and then we're going to, you know, adopt it. So, there's a lot of um work to be done to get to that level of kind of shrink wrapped, you

28. 11:16

    know, preparation for the product. I think let's uh drill into Yeah, go ahead. John, could I grab that real quick? I I agree with Richard that kind of the economics uh is is a big player in this and I I would take it another level and say the physics is really what's going to drive this because you know there are hyperscalers working on behalf of major governments doing you know sovereign implementations. Uh what's the limiting factor here is really energy. So the extent to which you know municipalities,

29. 11:44

    states or even a f federal level will in any way and not at the government level but just communities pushing back on development. You know we've we've been playing this game in the crypto space for you know BTC mining uh you know searching for the the the cheapest power to provide compute to something. I think at the end of the day physics is going to drive a lot of this and it's you know right now we have a lot of different say

30. 12:09

    nation states who are competing to show they have the best physics. So at some point you know rubber's going to meet road here. I don't I don't I don't know right now how much can really be done to stop data centers uh in in the this is way out of my lane here but that seems unlikely. I agree. I mean sort of there's a um direction versus acceleration debate that Vitalik has prompted in the whole kind of deact you

31. 12:34

    know space um and you know do we want to go know hell for leather and allow um a race well investment in physics to just drive the delivery of you know AGIS whatever we want to call it um or or actually should we embrace um maybe some of the brakes on kind of just rolling out data centers as as one natural way of kind of tempering um but the speed of things. So we can think a little bit more you know uh consciously about the the direction and go back to those points about you know liability and and you know reputation onchain

32. 13:08

    reputation uh and and the identity that these that we going to have to be comfortable giving uh to these agents. Let's drill in for a moment on decentralized AI which I'm guessing most of this crowd is is deeply um concerned about. What do you think that space, broadly speaking, needs to do more of to look legitimate in the eyes of enterprise and business? How how can these two worlds work better together?

33. 13:38

    John, I'm guessing I'm guessing some of your your PCs uh would love to have enterprise customers, right? So, what what more do they need to do to kind of gain mainstream adoption, acceptance? I think any consumer of AI at all is buying outcomes. And we've heard that a lot over the past two days, right? So, at the end of the day, whether it's a consumer or a business, uh it's hard to say whether people value and price

34. 14:04

    privacy, but they do price not having their data leaked for bad outcomes. They do price not being hacked and having losses. they do price uh and I'm I hate to say this but where in our community we've had hacks where data has been leaked and we've had physical threats right so those are outcomes that we're trying to prevent so the extent to which uh we take this black box that is AI and we provide trustlessly verifiable either positive outcomes or the proof of prevention of negative outcomes is the

35. 14:37

    core of any value proposition that's going to make decentralized AI a thing right because right now nobody's buying decentralization, but they would buy those outcomes. Yeah, I mean, I 100% agree. It's kind of outcome oriented, I think. But it probably starts by, if you possibly can not using the word decentralized in in the pitch. I think you need to lean into well, you know, what is the the benefit arguably of of decentralization. Um it's

36. 15:04

    that maybe you can have more models and agents uh existing in a contextrich environment provided by users who might be comfortable that their privacy is going to be respected and if that's the case those that menagery of a you know agents and and models is going to deliver a better outcome uh because of that context and and that's and ultimately if people see that and reference or whatever you're going to get adoption but we still you know as an industry we've been guilty of you know sometimes pitching ideologically

37. 15:36

    sometimes pitch never in this space it never happens a little bit technically but I mean now is really the perfect moment to start you know you know talking about the outcomes um yeah not to get too grim here but to what extent do you think it's possible things might need to get a bit worse before they get better and by that I mean there's a lot of concerns about job loss from AI we have Not, but there are anecdotes. There's there's isolated incidents of company X laid off

38. 16:07

    employees, likely AI, but unemployment is still relatively low, right? We have not yet seen, I think it's fair to say, a massive like scythe chopping down jobs thanks to AI. It's possible we might see a lot more material job loss in the future, right? So we might see there are some early bubblings of AI resistance.

39. 16:32

    Um and AI is unpopular amongst many different kind of sections of the economy. Um but could it get even more unpopular and become more of almost a cultural war dividing point and then what happens? Richard, what are your thoughts on that?

40. 16:50

    Well, first of all, I think job loss is I mean it seems to me inevitable. I mean we see it in small companies, mediumsized companies, large companies. It seems to be happening across the board. Um what do people do about that? I think there's going to be a frantic scramble to kind of you know reskill. I mean I think anybody who who pretends that they're kind of keeping a breast of

41. 17:10

    all of the developments and to be skilled enough to adopt all of this even amongst this audience I think is you know probably slightly kidding themselves. Um but so I I I just I don't know how we're going to cope with that. I think you know you can see I'm asking an impossible question. There's only no way to know.

42. 17:26

    You can read online you can forecast you know the fact that we'll have a fast reaction as a society and and and shift people and maybe deal with it in the same way we dealt with what was the the top and quantitive easing and everything that had to happen post financial crisis and and then obviously dealing with COVID. But that level of reaction might might be required. [snorts] Um but I look I think that the bits that are a

43. 17:49

    bit more tractable uh are are questions of what are the possible advantages of of the the technology set behind web 3 can bring to the world of AI and and I think you know we've mentioned privacy preservation I think the flip you I think also alluded to the question of intellectual property I think that is something that is people are have legitimate you know concerns about um and I think there's great legislation

44. 18:14

    going through in the UK for example on that topic. Um, but if you think about it as something that we could address through tokenization, we could address through attributing the kind of contributions that we make to the, you know, to the data sets that drive this world, that would be valuable. And and I think it's it's important not just from a kind of moral perspective like oh you know my stuff got stolen and I should

45. 18:36

    have you know my I think there was actually a paper that that I haven't read read properly yet but I scanned but the uh I think it was from Harvard talking about unsurprisingly what happens when you provide incentives to agents who are going to operate in some incentive and then and they and they can swarm in one direction or another. um that there's humanlike behaviors emerge that are not that arguably machavelian or even nefarious uh in their nature.

46. 19:03

    And so we're going to see that happen. And it just underlines the importance of uh making sure we get those incentive structures right, what there is to be gained, what there is to be lost. And that's got to relate to what the agents and what you know we are contributing to that world, the the data sets, the intellectual property we're doing. But that whole area I think is is is super critical and one that we can contribute to.

47. 19:25

    And to your point, we're already seeing I mean they're just pattern matching or even bold book. We're already seeing agents right in the way they want to see us think they write. So totally agree John the final word. Give us a prediction of how this might no pressure a prediction. We're not going to call you in a year from now. We're not going to say John you said this. uh how this will shake out whether AI gets more or less legitimate in the next time frame your choice.

48. 19:53

    Let me put it this way and you just kind of part and parcel to your last question. whenever there's uncertainty first you see hiring freezes right so I think we will experience that and so the question then becomes is how do you deal with if you want mobility if there isn't say labor or job liquidity how do you make it work um one of the conversations I have with my own family is how to think about organizations are going to get much flatter reporting lines are going

49. 20:23

    to get disintermediated anything that you thought created friction in your place in the economy is going going to get a lot more slippery and so your ability to think much more entrepreneurally. Most of us are going to be living entrepreneurial lives and if you I think you have to start accepting that change in mindset people in the crypto community have been thinking this way for a long time. So a

50. 20:46

    lot of us who are here today, I think this will feel very natural. But no matter what pace at which either jobs change or they're cut or there's freezes, we just have to have an entirely different way about thinking about how we add value to society. This is definitely a societal change for the role of people. Sound bite of a day. We'll all be living entrepreneur lives. Love it. Uh we're out of time sadly.

51. 21:10

    Please join me in thanking Richard and John. Thank you gentlemen. Thank you. Thank you.