AI, R2 and the Future of Everyday Driving | Rivian CEO RJ Scaringe

00:00:00By 2030, it'll be inconceivable to buy acar and not expect it to drive itself.Every single one of our cars, we want tohave the ability for it to operate atvery high levels of autonomy. Radars areextremely cheap. LARS are very cheap,but the really expensive part of thesystem is actually the onboardinference. In order to imagine moreexpensive than any of the perceptionstack, [music] my view is EV adoption inthe United States is a reflection of thelack of choice. As consumers, we needlots of choices. We need to havevariety. We selfidentify [music]with the thing we drive. The worlddoesn't need another Model Y. The worldneeds another choice.

00:00:36>> Hi listeners, welcome back to No Priors.Today I'm here with R.J. Scarring, thefounder and CEO of Rivian. We're here totalk about their autonomy strategy,proprietary chips, their coming R2model, whether Americans want EVs, andwhat our relationship to cars is goingto be in the age of AI. Let's get into[music] it. AJ, thanks so much for doingthis. Thank you for having me.>> So, Rivian's already uh an incrediblycool company. How did you decide it wasgoing to become an autonomy company?When did that happen?>> I mean, from the beginning, we thoughtof it as a transportation and mobilitycompany. And in fact, even before Rivianbecame Rivian, when I was thinking aboutwhat's the first products, it wasunclear what kind of car would be, butor even if it was a car, but it wasalways clear we wanted to be at thefront edge of helping to redefine whatdoes it mean to have access to personaltransportation. And so autonomy isalways been part of the strategy, butit's now fully coming to life with thetechnology that we're building.>> And when you think about the function ofRivian, there's transportation, there'salso the experience. Like when how longago did you guys start investing in theautonomy strategy here?>> Yes, we launched R1 in um very end of2021.>> Mh.>> And we used what I'll broadlycharacterize like a 1.0 approach toautonomy. So we had a perceptionplatform.

00:01:52We used a a third party, afront-facing camera that was essentiallya third-party solution that then pluggedinto an overall framework that we built,but it was all rules based. So, thecamera is fed a rulesbased planner. Theplanner would then make a bunch ofdecisions around the feeds from theperception. And it was, you know, themoment we launched, we knew it was thewrong approach, but it was the thing we'started working on uh well before thelaunch. And so, at the end of 2021,beginning of 2022, we made the decisionto completely reset the platform. And>> was that hard as a decision?>> No, cuz it was so clear when we made wemade the when you're building somethinglike this, you're you recognize you'regoing to spend many many billions ofdollars creating it. So we knew thislike at the core of transportation is isdriving and at the core of that is ashift to having the vehicle be capableof driving itself. And so we made thedecision to redo it like clean sheet,you know, no legacy of what we had builtin the Gen One. And that first launchedfrom a hardware point of view in themiddle of 2024. Uh so that was with ourgen two vehicles. You know, not a singleline of shared code, not a single pieceof common hardware on the perception onthe compute side. And uh and then we hadto build like the actual data flywheel.So we had to grow the car park to buildenough of a data flywheel to then startto train

00:03:07the model. And what we showedin our economy day late last year, latein 2025, was the beginnings of a seriesof really like super exciting steps ofhow this is going to grow and expand. Isay this all the time. I I think of notjust for Rivian, but I'd say for theauto industry in general, the last threeyears compared to the next three yearsare going to look very different. So therate of progress that we saw in autonomybetween let's say 2020 and 2025 or 2021and 2025 and what we're going to seebetween today and let's say 2029 2030are they're completely different slopesand that really comes back to you knowentirely new architectures now beingused to develop self-driving actuallytruly AI architectures whereas beforethese were not AI architectures in thein the true sense they were they wereusing machine vision but reallyrules-based environments that we definedas as humans, you know, we codifiedthem, which is very different than howApple today.>> You might actually have perfect timinghere in that uh I got to be part ofinvesting in sort of the first wave ofindependent autonomy bets that wereworking with the OEMs at my lastinvesting firm. Okay. But this is>> I would say 8 10 years ago.>> Yeah.>> And uh as you mentioned there's severalarchitectural revolutions since then.Yeah. And so for companies to

00:04:22make thatshift from you know we're going to havethese separate perception and planningsystems to more endtoend neural networks>> I I asked because I felt it was actuallyquite a hard decision for people inchoosing their partners and from a froma technical perspective.>> Well I think it I mean you can see it.So there's if you go back to the verybeginning of the idea of self-driving,a lot of effort, a lot of spend happenedfor companies to build these rules-basedenvironments and to build these moreclassic systems. And when transformbased encoding came along, you just acouple years ago and it shifted veryrapidly to it was clear that the futurestate was going to be neural net based.It was hard because if you're a companythat built all these systems, it's likedo I keep investing what I had? what doI what do I do with all this work thatwas was built before? And the reality isis a lot of it is the vast majority ofit is going to be pure throwaway. Umbecause it wasn't like a gradual shift.It was a complete rethink of how thingsare architected.>> How did you decide that this was goingto be a an in-house effort versus apartner effort that given most peoplewho made cars said we're going to gopartner or buy something here? I I guessthe emotional philosophical is on thingsthat are really important, we've takenthe approach of vertically integrating

00:05:38them. So electronics, our software, allthe high voltage systems in the vehicle.So things like motors, inverters,uh all the power electronics, these areall things we we develop and buildinhouse. And in a few cases, you know,we had to start with something that waseither off the shelf or partially offthe shelf. But today, all of that'scompletely in-house. And in the case ofself-driving, we knew that long-term itneeded to be something that wasdeveloped internally. We started as Isaid with a mobilecentric solution,which a lot of folks did, right?>> Particularly in like you that 2015 to2021 time frame. But when you reallylook at what's necessary to to besuccessful in a neural net basedapproach,>> there's a core set of ingredients thatvery few people have and I think weuniquely have them. So first andforemost, you need to have completecontrol of a perception platform. Youhave all the everything that the the thesystem is capable of observing, whetherthat's cameras, radars, or LAR, or somecombination of all three. You need tohave control of that. Meaning there's nointermediary company that's likeprocessing some of the information. Andso that's powerful because you can thenfeed raw signals into your system. Thesystem needs to be capable of triggeringunique or interesting or noteworthyevents that you can then use to trainthat triggered.

00:06:54you know those triggeredmoments need to then be captured savedon the vehicle and then when the whenthe time arises where you have Wi-Fiideally send it up and the reason I sayWi-Fi these are this is a large a lot ofdata so you could of course do it overLTE but it's expensive as you have tohave a really robust data architectureon the vehicle then you need to be ableto send it off offboard and use thatwith a lot of uh training so with a lotof GPUs to train a model companies thatare either developing independentsolutions that are not a car companythey typically don't have access to thetype of mileage that we do. So that thehuge amount of data that our vehiclesgenerate. Uh if you're developing thisfrom a sensor set point of view, youtypically don't have the vehiclearchitecture and the vehicle car park.So we just came to the view that we haveall these ingredients to do it reallywell. And [clears throat]>> it's like not an optional thing. It'sthe companies that do this well willexist. The companies that don't do thiswell,>> like I feel really strongly this. Theywill not exist. They will shrink toshrink to nothing. asmtoically approachyou know zero.>> You think it can only be delivered inreally a vertical vertically integrated?>> No, I think I think there's more thanone less than five companies outside ofChina that have the necessaryingredients to do this. The capital, theGPUs, the the car park with you know

00:08:09enough vehicles generate enough data. Isay more than one less than five. It'sprobably>> and the control of that whole trainingloop you're doing.>> It's probably like more than one lessthan three maybe four. Like there's verysmall number of companies that can dothis. I think the uni unique spot we arein time right now is the 1.0. Can>> I ask explicitly then? It's you, it'sTesla, it's Whimo. Is that the three?>> I would include all three of those.Yeah. And there's maybe one or twoothers in [clears throat] in the mix.But I think>> the challenge is you have to look at thenot just the moment in time forperformance where we are today.>> Do you have the ingredients to continuemaking progress at a very high like highrate over the next four or five years?>> And so a lot of the solutions that aremore 1.0 based and and are sort of stuckin that framework I think have a like atruly a 0% chance of progressing to becompetitive with a neural net basedapproach and the neural net basedapproach does take a lot of times youhave to build ton of inference on theyou have to have either buy it or buildit a lot of inference we decided tobuild it so we built an in-house chip todo this you need to have a car parkedthis large>> you just mean enough onboard compute toactually run the models the car yeah inthe vehicle and so you could you couldbuy that. Of course, Nvidia makes those.Um, but you need to be able to do thatat scale and have it in every car. Andso, we took

00:09:24the decision to make ourchip in house.>> Is that more a capability uh decision ora cost decision?>> It's a cost. And then like we want tohave it on everything. So, every singleone of our cars, we want to have theability for it to operate at very highlevels of autonomy. And so, we design,spec, and build the cameras.>> Radars are extremely cheap. LARS arenow, you know, very, very cheap. But thereally expensive part of the system isactually the onboard inference.>> And so that's like an order imagine moreexpensive than any of the perceptionstack. I think people focus on theperception because it's the things wecan like visualize,>> right?>> But the brain is actually the mostexpensive part. And so we brought thatin house as a way to remove cost fromthe system so that we can easily deploythis on on every car.>> You are taking like a sort of knowstep-by-step approach to levels ofautonomy. Yeah. and Rivian, how do youthink about um how quickly you approachlike level four or you know the safetycase around each of these things? Howfast your team goes against this?>> Yeah, I mean this is even this questionis unique because just a few years ago20 2019 2021 even there was like verylike very clearly delineatedways to approach autonomy. There was alevel two approach which was cameraheavy maybe with a

00:10:39few radars>> and then there was a level four approachwhich was of course had cameras but hada lot of lightars. It was sort ofinconceivable to think of the level twosystem becoming a level four andsimilarly the level four system was wayoverbuilt to even like conceivably thinkabout putting that on every consumervehicle.>> Well, you didn't want the the big wantall these parts. Yes. The tens ofthousands of dollars of perception. Sowhat's happened is those two worlds justI think have just started to veryclearly merge where the delineationbetween a level two, a level three and alevel four um in terms of perception andand in terms of compute has started tofade and it's now essentially justremove like how capable the system is ataddressing all these corner cases.And you know, this is what's hard for aconsumer to recognize. If you're drivinga level two system or a level threesystem or a level four system for99.9999,like three or four nights, feelsidentical,>> right?>> The difference is like the fifth orsixth or seventh nine on that is theselike extreme corner cases. And so Ithink it's actually led to a lot ofconfusion where you'll be in a level twosystem like the car could drive itselfand you're like yes it can under>> most of the

00:11:54roads conditions exceptthese very unique corner cases. And soto your point on safety cases, thequestion then becomes is like howconfident are we in the systemcapability in covering these reallyobscure unlikely rare events which ofcourse if they're not covered well itcan lead to really uh you know terribleoutcome you know the vehicle in a badcollision and so that's where the neuralnet based approach has just changedthings a lot. So the the thecapabilities are so much stronger andthe ability now I think for us to deployon a lot more vehicles have a car parkthat's very large. So we went from, youknow, few years ago state-of-the-art wasyou'd have a test development fleet ofmaybe maybe a few hundred vehicles,maybe maybe like high hundreds ofvehicles to now like thousands andthousands. Every single car on the roadis part of your data fleet that'sidentifying these unique corner casesand then running the model against themto test.>> And now of course we're simulating thoseunique cases and we can do a lot there.So the just the whole nature of it'schanged so dramatically that I mean Ithink by by 2030 it'll be inconceivableto buy a car and not expect it to driveitself. You know maybe that's sooner.Maybe like we hope it's sooner likewe're targeting a little sooner thanthat but

00:13:09certainly in like a very verynear future like that will become amustave in a car. Sort of like it's hardto imagine buying a car today withoutairbags or buying a car today withoutair conditioning. Um these things at amoment in time were optional. I think innot too not too much time, couple years,it'll be hard to concede buying a carthat can't drop you at the airport orpick up your kids from school.>> I would argue that right now um most ofthe biggest car makers do not have theingredients that you described to makethis a reality.>> So, do you think that um that's going toplay out in the market where likeautonomy will be so important as adriving feature, core feature of the carthat there's just going to be a bigmarket share shift to those those whocan figure it out. I I know you'rebiased here, but I'm like,>> "No, no, no. I think it's it's it's ahard question to answer." So, I thinkit's uh I I always characterize likethis.>> I think it's inconceivablefor a car company to continue to operateat scale like mass market. I think veryniche enthusiast realms sure, but likeat scale>> without a software defined architecture,which is even before you get toautonomy, just like can you do OTAAS? Doyou have control of a of a>> sorry can you define software definearchitecture?>> Yeah, that's like before we even get toautonomous

00:14:25like these are like basics.So the way car>> the core thesis of>> Yeah. Yeah. So the way car electronicsystems have been designed and built andhave evolved with the exception of Teslaand Rivian every car on the road haswhat is uh called a domain basedarchitecture. So you could also call ita functionbased architecture. So all thefunctions across the vehicle, let's saychassis control or door system controlor uh eight track, your air conditioningsystem, all have little computersassociated with them, right?>> What we call ECUs, electronic controlunits. And in a modern car, you mighthave 100 to 150 of these. And each ofthese run their own little island ofsoftware. And that little island ofsoftware is written by a supplier, morelikely a supplier to the supplier. Soyou go to a a tier one and they hire atier two who writes the code base to runyour H.>> That's why it's impossible to debug likea software system. And>> it's also why it's really hard to do anupdate. So imagine you have a 100different islands of software written by100 different teams uh that all have tocoordinate. And so if you want afeature, you know, something thatmanifests as a feature often involvescombining functions from differentdomains.>> So a simple one to visualize is when youwalk up to your car to get into it, youwant it to automatically unlock. Youwant the HVAC to go to your preset. Youwant your seats to adjust. You want it

00:15:40to make an audible noise in the outside.You want the lights to do something.>> You probably want the the audio systemto do something. Those are all differentlittle ECUs in a traditional car. Andthe coordination cost in it is reallyhigh. It's very unlikely that a carcompany will make a change to thatsequence because it involvescoordinating amongst maybe 10 differentplayers. In contrast on a on a approachwhere you build a zonal architecturewhere you have a very small number ofcomputers ideally you know one two maybethree depending on the size of the carthat are running one operating systemthat control everything. It's very easy.So that sequence you could make upupdates to you know in a matter ofminutes maybe an hour you could changethe whole sequence of what happens youwalk up to the car issue an overearupdate and it's very straightforward.How often does Rivian update?>> We do about one a month and uh it'stypically, you know, we add a couple ofnew features, we add refinements toexisting features. We're listening tolike what customers are seeing andasking for, but you know, every monththe car gets like notably better andit's created this really amazing dynamicwhere customers are like excited for thefor the update. They're like, when's thenext OTAA going to drop? The irony ofall this is these domain basedarchitectures goes back to like how dowe arrive at this it actually goes backto fuel injection systems.

00:16:55So up untilearly 1960s like every car on the roadwas completely analog. So there's nocomputers at all in the cars 100% analogand the first computers were there todrive the fuel injection systems and carcompanies said this isn't a corecompetency. Let's push that littlecomputer to run the fuel injectionsystem to a supplier and the supplierwill make that. You know, this is whereyou saw things like the Bosch fuelinjection systems and never planned.It's sort of like a field of weeds. Thenover the next like 7 60 70 years,everything that became, you know,computer controlled to any degreesuddenly started to have a little ECU, alittle computer associated with it. andit just like grew into this absolutedisastrous mess that is a you know todaythe the network architecture that's intruly every car on the road with theexception of of two companies that whatI just described is what underpins wedid a large uh software licensing deal a$5.8 8 billion deal with VolkswagenGroup, the second largest car company inthe world to uh essentially leverage ournetwork architecture and ECU topologyuh for their you know all their variousbrands and so it's an interesting finalpoint there on the on your firstquestion which is you know what happensto market share so I think it'sinconceivable that car if to

00:18:10be at scalethat you don't have a softwaredefinfined architecture that allows yourfeatures to become better and better andparticularly thinking about how AIstarts to integrate into the featuresthat's numberSecondly, it's inconceivable to thinkabout a car company existing at scalewithout the vehicles having very highlevels of autonomy. And so car companieshave a choice on both of those. They caneither accept that they're going toshrink. That's choice one. Choice two isgo build it themselves, which is reallyhard because they don't typically havethese skill sets. They're not softwareelectronics companies in terms of liketheir organizational DNA. Or they canfind a third party to source it from.And in both cases, there's not greatthird parties to go to. Uh, and in thecase of autonomy, most of the thirdparties that that did emerge over thelast 10 to 15 years tend to be very muchuh like classic rules-based what calllike AD or autonomous vehicle 1.0solutions. And those work pretty wellfor the business construct of sellinglike a sensor and a function. But thatstructure is really flawed when you wantto have like a large data flywheel andit's constantly learning and evolvingand you're issuing updates constantly.It's just um it's really hard to imaginethat with an arms length transaction.And so I think the vertically integratedstacks

00:19:25are going to naturally have somebig advantages.>> So this might be an irrelevant questionbut I'm curious. Um do you think thatthe autonomy like the models that maybethe three maybe the one maybe the fivecompanies that come up with this>> uh develop are fundamentally differentover time because I spent a lot of timein the AI ecosystem and the>> let's say the languageorientedfoundation models like feel like they'reconverging at this moment in time.>> I I look at a Rivian I'm like>> I don't know people adventure in thatthing. Do do you actually want it to dodifferent things, have different stylesor capabilities, or is it really justlikeas much autonomy as possible safetycase?>> Well, first I This is a great this is agreat question. Um>> I want my car to drive.>> So like in the LLM world, it a lot of ithas converged because it's the trainingdata sets nearly the same. Yeah. Sowe're taking the the breath of knowledgethat's contained on the internet andwe're training models off of that. Inthe case of driving a vehicle, there isno internet of driving data. And so youneed both a robust sensor set to be ableto capture the data and you need a carpark, you know, that has enough vehiclesin it. And so, of course, Tesla has thelargest car park of vehicles by far. Ourapproach to this is we have a a higher

00:20:40level of capability on our perceptionstacks. We have better cameras, we haveradar, and of course with R2, we'll havea LAR as well. A huge part of thatstrategy is not only those cover cornercases better. So the cameras haveincredible low light and you know brightlight performance. So the dynamic rangeof the cameras is stronger. We have morecameras, a lot more megapixels. Uh wehave radar which is great for objectdetection. And the LAR which is it's avery powerful tool for training the themodels. And so imagine800 ft in front of us there's a littlespeck into a camera. It's hard to figureout what that is. And historically, whatwe would do to train that is she wouldhave a LAR sitting on the vehicle on ona like a ground truth fleet to helptrain your cameras. Putting that onevery single one of our cars is turnsour entire fleet into this amazingtraining platform, this data acquisitionmachine. That was a core part of how wethought about our strategy is we'regoing to go, you know, not as heavy aslet's say a Whimo on perception,>> but heavier than let's say Tesla tobuild a really robust data platform on avehicle-by- vehicle basis and then witha car park that's going to grow growsignificantly with the expansion withR2. Yeah. So, I I think first andforemost is there is no common internetdata. So

00:21:56the data sets that we're goingto be picking up though are going to bevery similar>> but but you have to go acquire>> but there's still different decisionsabout what data you care aboutacquiring. Yeah.>> Well I think this is what to[clears throat] like how does a car feelultimately it needs to be safe and thedifferences in the way it drives orfeels are going to be more about likewhat's the UI the user interface of it.You know like even we just updated someof our features. We have three settingsfor how the vehicle drives. Mild,medium, and spicy.>> Spicy is the highest one. Yeah. And sothis is like a little bit moreaggressive over time and we've spenttime thinking about this. I think thiswill start to become part of a keydecision is how does the vehicle behaveand there's work we're doing to to thinkabout how the vehicle can behave in away that against a set of heruristics>> drives like you.>> So overall the overall model is trainedon how to performs in a safe way but itactually learns some of your you learnsome of your driving preferences andcreates a model around you. Of course,in a world where you never drive the carbecause you're just it's always drivingfor you. There's a way for you to set.I'd like it to aggressively changelanes. I'd like it to reside in theright hand lane. Like those kinds ofdecisions and those are those are lessaround the tech, more on what's the thethe product or the UI if you like.>> Right.

00:23:11The ability to collect thosepreferences.>> Yeah. Preference based. And I think wewill see that>> and that'll be a decision like a Teslamakes that may be different than howRivian makes it. you know, it's hard tosay today.>> Can we talk about what the R2 means forlike the company and some some of thekey design decisions here? I was justtalking to Jonathan, one of your leaddesigners, about the constraints and,you know, aiming for more mass marketand more volume here.>> Uh, I mean, yeah, you said it. It's uhso R1, it's a flagship product. It'saverage selling price is around $90,000.It's the best selling, the R1S is thebest selling premium electric SUV in thecountry. So it's electric SUVs over$70,000 and we're the bestsellingpremium SUV electric or non-electelectric in the state of California. Soit sells really well. You know, it outsells everything in its class like amodel Tesla Model X. It out sells like 2to1. But um because of the price, it'sjust limiting in terms of how muchvolume we can achieve with thatplatform. And so R2 is the our firsttruly mass market product with pricingthat's as we've said going to start at45 and allows people that are in thatyou know the average price of a new carin the United States is $50,000 in thatlike $45 to $55,000 price range.

00:24:26Uh Ithink to have a really great choice andto date there haven't been a lot ofgreat choices there. You know there'sI'd say there's like sort of singularset of great choices with a model 3model Y. Uh and of course that's that'sshown through the extreme market sharecapture of 50% roughly market share goesup or down but around that call it halfthe EV market is Model 3 or Model Y. Sothere's just such an untappedopportunity to pull customers out of ICEvehicles out of internal combustionvehicles with a choice that's you knowhas characteristics that are differentand unique relative to a Tesla. Theseare like too substantive to be rapidfire questions, but they're they'reimportant for me to ask you. DoAmericans want EVs? Like why haven'tthey adopted them faster?>> What?>> Yeah, I think to the last question, Ithink causality is always a hard thingto,you know, really understand, but let'szoom out here. The the overall adoptionrate in the United States of EVs isaround 8%. The vast majority of vehiclebuyers are buying vehicles that areunder $70,000 with the average saleprice of about 50. And so if you look atthe number of vehicle choices you haveat a price point that's under $70,000depending on the year. This of coursechanges year to year. There's well inexcess of 300 different vehicle

00:25:41modelline choices. Putting aside trims andperformance packages but just in termsof like overall vehicle types. And soyou can buy hatchbacks, minivans, SUVs,you know, two-seaters, convertibles. Imean there's a whole array of differentthings you can buy. And in the EV space,I think, and this is I think there'smore than one, less than three greatchoices. And I'd say Tesla with theModel 3, Model Y is absolutely one ofthose. But there's so few choices thatif you are looking for a form factorthat's not a Tesla.>> So, you think it's just missing productset that people want? Yeah.>> An extreme lack of choice is how you putit. Um, like a shocking lack of choice.And this is what gets into interestinglike corporate psychology, but becauseof the success of the Model Y inparticular, the EV choices that do existthat are outside of Tesla are often verysimilar to a Model Y. Sure.>> So if you were to like draw like anoutline, if you looked at the side viewprofile of a lot of its alternatives anddraw a profile and then put it next to aModel Y, it's almost identical. There'sa design sketch over here of basicallythe Model Y and all its competitors.>> They're all basically the same. It'slike if you want a Model Y,

00:26:56buy a ModelY versus getting>> you want something different.>> Yes. You have all these companies aretrying to create their own version ofModel Y. And it's like it's unfortunatebecause they didn't say, "Well, what canwe do that's unique and different?" Andso for us, we think the Model Y is agreat car. I've owned one. Many folks onour team have owned one. But the worlddoesn't need another Model Y. The worldneeds another choice. And so I think uhthis is a reframing of just how we lookat transportation is it's such a bigspace. It's such an area of personalexpression that we need as as consumerswe need lots of choices. We need to havevariety. We selfidentify with the thingwe drive. We just haven't had it. So Ithink my view is the EV adoption in theUnited States is a reflection of thelack of choice. Uh there's one set ofreally great choices with Model 3, ModelY. I think there needs to be many more.And so even looking at our partnershipwith Volkswagen Group, a big motivatorfor that which ties to our mission wascan we take our technology platform>> and allow that to be expressed through avariety of really interesting uh andvery story brands and different formfactors, different price points um ofcourse different segments. And I I thinkthe more choices we have, the more it'sgoing to lead to

00:28:11broader based adoptionof electric vehicles, whichcreates, I think, a a very positivelevel of momentum around the space. It'sit's worth noting on that point when welook at how we develop a car like takeR2, we don't think of it as this issomeone who's going to buy an EV, let'smake it good. We think of it as let'smake the best possible vehicle, youknow, we can imagine. So incredibleperformance and you know great range,great uh dynamics, tons of storage andthe person buying it will be drawn intoelectrification because the car is justthe best choice they have.>> And we took that same view with R1 andon R1 the vast majority of our customersare first time ever owning an EV is aRivian which is which is really good. Ifif all we were doing is moving customersbetween>> one or two brands it wouldn't beaccomplishing the goal. We have tocreate new EV customers with productsthat are so compelling that it justdraws people in.>> So that leads into my very last questionhere. I grew up thinking like a car is ahuge part of my identity.>> Love cards. Drew them.>> Still think they're pretty cool. Uh andyou know as they become more likeutilitarian services with uh the rise ofrobo taxis as a concept of like you knowserving

00:29:26some of the function which yourcar did before. How do you think ourrelationship with cars changes orvehicles over time?>> I do think it's we're going to see ashift. It's an interesting likephilosoph philosophical question. Whywhy are cars such a part of our societyand>> why do we have this affinity for them ina way that we don't have that feelingfor other things in our life that arereally important? Like I don't I don'tlook at my refrigerator and think Ireally love that. Um in the same waythat I do with a car>> and I think part of it is a car enablespersonal freedom. It allows you toexplore. Um, it's it's something thatyou not only ride in, but it be becomespart of an expression of self. And Ithink that's probably going to continueto some degree, but it is going toevolve. and and the way we look at it uhwith our products and even how we'velaid out and contemplated the thepurpose of the brand. We really look atit through the lens of the vehicles andthe products we make need to both enablepeople to go do the kinds of things youknow that they would hope to havememories of years to come. So we weoften say the kinds of things you'd wantto take photographs of but more thanjust enabling it which is a functionalrequirement like can it drive their youknow can it fit the stuff your your petsyour gear your friends your your all of

00:30:41your stuff more than just enabling itcan it inspire it and so can the brandand the way we present what we'rebuilding and the way we make designdecisions inspire you to go do thethings you want to remember for years tocome and so there's little like designdecisions we take that link to that. Soa flashlight in the door>> is a invitation to explore. It'sinvitation to go look at things thenight. Uh the>> or the treehouse.>> Yeah, there's exactly. So there's allthese little decisions you madethroughout the whole car that are justdesigned to like engage that element ofinspiring people to go like imagine thatlife they want to have.>> Awesome. Thank you so much, [music] R.J.Congrats on the R2 and uh on theautonomy program.>> Thank you.Find us on Twitter at no prior pod.Subscribe to our [music] YouTube channelif you want to see our faces. Follow theshow on Apple Podcasts, Spotify, orwherever [music] you listen. That wayyou get a new episode every week. Andsign up for emails or find transcriptsfor every episode at no-flyers.com.[music]