Lessons From Self-driving Applied to Web3 - Asta Li

Transcript

hi everybody today i'll be talking about boom and bust um cycles in self-driving in web 3 and the lessons and takeaways from those cycles so ginger reduced myself i'm i'm asta i'm a reformed self-driving engineer i worked on a variety of autonomous robots over the last decade what started my journey in self-driving when i was at uber in 2016 and then was one of the first engineers at aurora starting in 2017 spent a lot of time thinking about data infrastructure when i was working in self-driving and left over a year ago

to build data infrastructure for web3 so today i'm building privy which is a developer tool for managing user data off chain and i thought it would be fun to share some of the parallels between my former life in self-driving and what i three what i see in web 3 today so to give a little bit of structure for the talk i'm gonna go through a bit of a walk down memory lane and talk through um some of the shared history between self-driving and web three in terms of boom and bust cycles um and then talk through a couple of takeaways so yeah history as far as sort of expansion and consolidation is

concerned um history doesn't you know repeat itself but it sure does rhyme according to this supposed mark twain quote and i i think one particularly interesting lens to look at history through is uh through the lens of the gartner hype cycle where um we sort of see this trending in technology again and again where there's a an innovation trigger um there's a an expansion of companies and projects that rise with the peak of inflated expectation and then there's sort of this dip and consolidation phase as you enter the trough of disillusionment um and then a sort of slope of enlightenment beyond that so let's talk about that spark and in in the case of self-driving and

in web 3 it's not exactly a single spark but it's a combination of technology innovation as well as social coordination that sort of sparks this expansion phase so in terms of self-driving i think the single biggest innovation is around deep learning around the research that encouraged advances in computer vision that we'd never seen before and for web 3 a lot of this is around the the launch of ethereum and the the erc20 standard and i think in addition to technology innovation the sort of less talked about spark for expansion is actually coordination or human coordination and

so in the case of self-driving this is like the darpa grand challenge which is a government-funded initiative for a variety of projects across academia and industry to to try and build a self-driving car of a you know in a competitive environment this is regulation making self-driving street legal for the first time and and for web 3 this is standards that people agree on and coordinate around that allow for the interoperability promise of web3 in fact i think there's a parallel which is that the lack of regulation is one of the reasons we've been able to run interesting new experiments in the web three space and beyond that the hype and funding and news coverage and just public interest across self-driving

and web3 are major coordination points that lead to innovation and expansion so expansion is sparked not just by technology but also by human coordination so that's uh moving on to the expansion phase um in self-driving uh so i i got into self-driving in 2016 and over the course of 2016 to 2017 there was this massive explosion in self-driving companies i'd go to cvpr which is a computer vision conference and it was like every company was a self-driving company there were like 40 or 50 companies that are all trying to launch you know l4 l5 self-driving there's a huge capital influx and a ton of hype and uh and

headlines that were you know saying ah by the end of 2018 we're going to have fully driverless cars on the road and elon musk tweeting left and right about um cross-country autonomous summons working in in just a couple year time frame and so this is me at the end of 2017 sort of peak hype of self-driving waiting for my glorious self-driving robot filled future and uh meanwhile there's some interesting stuff happening in web 3 as well which is in 2017 you get this first massive mainstream awareness of web3 like you couldn't have paid attention to the news and not heard about crypto

with this came the ico boom um in particular there was block one's four billion dollar ico over the course of a single year and they were completely pre-product which is crazy to me and over the course of 2016 you had something like 800 icos raising over 20 billion dollars worth of capital and bitcoin spiking to over 10k for the first time so again this is me at the end of 2017 peak hype of web3 and i'm waiting for my glorious decentralized future but at the same time you know i i think at the end of 2017 we're also seeing zero consumer applications beyond

accessing l1 blockchains and i still hadn't even ridden in a self-driving car that wasn't operated by a person in the driver's seat so that leads us to the the consolidation phase which happened after this sort of peak in 2017. where self-driving actually it turns out takes more capital and more time than expected and over the course of the years after 2017 we see a huge amount of consolidation this is amazon buying zuke's apple buying drive gem buying crews neuro acquiring ike aurora and uber self-driving division and so on and so forth and these span the spectrum from aqua hires to you know for pennies on the dollar to

strategic partnerships but regardless like the consolidation was very real and meanwhile the the biggest self-driving companies just kept getting bigger and now waymo and cruz for instance each employ over a few thousand people similarly in web 3 after the iso after the ico boom you get a lot of over-promising and under-delivering a lot of tokens not being tied to underlying contract utility and ico companies taking multiple years to deliver on their product promises or failing to launch a product at all so we're sort of in the trough of disillusionment but with the consolidation we also see massive successes emerge in the post

2017 time frame and this is some of our first consumer applications and um and and protocols that have long lasting value that are still around today and and that are core parts of the web3 ecosystem and interestingly to know much more than in there is a natural path towards consolidation in web 3 that's related to network effects especially for protocols network value increases so much more with each additional participant and the incentives around consolidation web 3 are really quite explicit and also to be fair the expansion and consolidation cycles in web3 um you know

aren't uh they're it's they're multiple cycles web 3 is broad enough to have multiple sort of hype cycles layered on top of each other through d5 crypto gaming stable coins dows nfts and so on and so it's not just store sort of this uh this one curve um so that leads us to the question like where in the world like where in this process are we in the in the gartner hype cycle i'm like where in this uh slope of enlightenment are we right now and there's very much this question around like are we there yet for both self-driving and for web3 um and uh and i think for self-driving there's a valid criticism and this is a

slide i would have never been able to show had i still been working in self-driving um but the vast majority of self-driving miles are driven still with one or two operators um there have been accidents in self-driving and i'm you know i i can't tell if we've saved people's lives yet and similarly there's some criticism in web 3 which is i'm still waiting for my glorious decentralized future but there's actually a lot of centralization whether this is asset ownership or along the web3 stack as moxie discusses recently but all in all i just take it to mean that across the board we're still in the early days and that there is still so

much more to come and so from this shared history of boom and bust and expansion and consolidation i think we have a lot to learn from both self-driving and from web3 and namely the first lesson is really one that is a takeaway for me from my experience in web 3 which is that in self-driving we should spend more time breaking down silos a lot of this reflection comes from why web3 boomed to begin with i think the ico boom the ico boom happened because of open access to capital and really allowing companies to have to be able to raise without being a part of the the sort of like boys club of

silicon valley and being able to raise from investors who weren't just private institutional investors but from the public and from people actively participating in those protocols and networks i think that web 3 has lived up to its promise of interoperability and that in web 3 by building in the open by layering up projects on top of each other and and working in in open source web3 has been able to move incredibly quickly and actually ship in a way that i think self-driving has a lot to learn from so an open secret in self-driving is that all the stacks are more or less the same which is

the the same algorithms run in the systems in systems that are composed of more or less the same building blocks and i've personally written the exact same geometry libraries multiple times over so i would love to see self-driving adopt the ethos of open of building in the open and of sharing that we have in web3 i think it would lead to like faster adoption and faster progress in the self-driving space in particular i think that for instance we could do a lot better in self-driving around data sharing which is sharing like a corpus of disengagement data where every time a

self-driving car makes a mistake on a public road an operator has to take over like i'd love to see those scenarios be shared in a corpus across companies and be used to build simulations that everybody can benefit but despite self-driving stacks all looking the same i don't think that we should make them interoperable so i'm not advocating for like a proliferation of self-driving standards that nobody uses and this is something that's been actually a bit painful for me in web3 where everywhere i look there's like 18 standards that people all want my t all want me to use um so i think there's some nuance as a part of this sort of exhortation to break down silos and on the topic of nuance like i get

why self-driving companies want to own everything themselves and it's because interfaces are sticky and that it's easy to over design interfaces prematurely and so i think that as far as breaking down silos and sharing goes there are things that make a lot of sense to share more of which is research algorithms infrastructure disengagement data and so on places where it's less useful to break down those silos so it's less useful to share like raw sensor data or internal module code and in the same way i think there's room for nuance around coordination and decentralization in web3 um the other open secret as far as web 3 goes is that web 3's a lot more

centralized than we want to be and then we say we are and i think the nuance here is that the centralization to begin with isn't always bad by its nature i think that coordination in decentralized structures is really important and can be a bootstrapping mechanism so um this is a sort of revision of my original lesson which is to break down silos but to do so in a way that's more thoughtful and nuanced and the second lesson is to invest in tooling which is that from self-driving i've learned the deep importance of tooling and infrastructure so in self-driving

we've found that in the history of consolidation of a history of expansion and consolidation the best companies the ones that have that are long-lasting and still around are the ones that instead of trying to launch a self-driving system in just a year or two are the ones that built for the decade they're the ones that invested deeply in infrastructure in tooling they're the ones that invested in building labeling tools or data infrastructure onboard execution infrastructure or or simulation pipelines have proven value as a as a foundation for sort of this decades-long self-driving project and i think the same lesson can be learned for web 3 which is that like remember we're

building for the long haul and i think there's a perspective around long-term value for web3 that is not to be underestimated i think there's an opportunity here to fundamentally shift how people think about data ownership i think there's an opportunity to invest in tooling and infrastructure for the long haul and i also think part of building for the long haul in web3 is focusing on usability and and building better tooling for usability which leads me to this second point in terms of lessons around tooling is that there's also you know in the self-driving space a lot of the companies that pushed the space forward the most um were the ones that

built great tools so whether that's scale ai which built labeling tooling or the various lidar and sensor companies i think that the companies in the self-driving space had really long lasting impact were also ones that were building tools themselves and so in web 3 i think there's similarly a really amazing opportunity to work on great tooling that powers the next generation of web 3 companies so i'm very excited personally to be building at the infrastructure layer in web3 so again revisiting these lessons i think that the takeaway from this shared history of bloom and bust of expansion consolidation across these two

two industries that i spent a lot of time in is to break down silos and do so with nuance and and do so with thought and to invest deeply in infrastructure and in tooling and so that's what we're doing at privy which is we build developer tooling that breaks down data silos we make it easy for web3 companies to take on sensitive user data and understand their users as more than just a non-chain address we make it easy for users to control and share their data cross-application and cross-chain so we're just getting started in self-driving i saw billions of dollars of investment lead to the

training of an entire generation of roboticists that wouldn't have existed otherwise and i'm seeing web3 draw incredible talent in the exact same way creating a generation of builders who think about decentralization distributed systems security privacy and more and so i am so thrilled to be a part of this space i think we're very much still in the early days and there's so much more to build thank you [Applause]