⚡TERAFAB: ELON’S FACTORY FOR THE POST-SMARTPHONE ERA

Kind: captions Language: en The smartphone era is ending. The next dominant consumer device will be Tesla Optimus, which couples conversational intelligence with physical agency in the real world. Its intelligence lives in Star Mind. Tesla provides the body. SpaceX enables the mind through Starlink and Star Mind. I'm quoting from Phil Bisel's recent article on X that has become all the more important to understand now with the announcement of Terrafab in Grimes County, Texas. Phil is a visionary founder of technology teams at Apple and Rivian driving innovation in the automotive and technology sectors. Welcome, Phil. It is so great to have you back. >> Oh, thanks. Great to see you again. It's been a while. >> It has. It has [laughter] unfortunately it has but uh >> lovely to have you back all the same and fortuitous timing. >> Oh my gosh. Yes. >> Yes. Um what a day. >> Just what a day. Mind-blowing. So, I you know I I I have to begin by So, you you've you've been a long-term SpaceX uh investor and um I thought I'd ask you this whole kind of a a fear-mongering that was going on about, oh, you know, the unlock's going to happen, long-term uh investors are going to sell. None of that seems to have happened in the way it was feared. Not this one, >> you know. Yeah. No, I don't think so. I'm so glad we're over this hump, uh, the August 6th hump. It's the big first unlock. And most importantly is it kind of previews the next unlock. You know what I mean? It sort of tells us what the pattern's going to be. And what is the pattern? It's a nothing burger. I mean, >> I mean, since people, they're uh I think I characterize it as the largest shareholders in Tesla, the early investors that have the big swaths of stock, they're not moving. They're in it for the long haul. I mean, they might move a little bit. Um, some of the employees will sell to get out of their apartments and buy a home, and that's totally fair. And um but it won't make up a volume difference that I mean if it if it dragged the stock anywhere it dragged it down yesterday in anticipation and today what's a little up today u after the fact and I feel it really takes one of the big weights off of of SpaceX right now. Is does this is this indication that people still don't understand what SpaceX is all about >> and people who invest in SpaceX for the long term? >> Yeah. People don't understand what SpaceX I mean they think >> I mean they think it's a rocket company. I guess it is a rocket company, but I don't think they understand that. Every time I have a conversation with someone that's not, you know, they may have actually have a couple sheriffs, but when I tell them about compute and AI and that they don't really know that story. We do. >> We do. And and we we take it for granted that everyone does. That's not true. Um, >> I was recently on a little vacation back east and I talked to, you know, dozens of people, friends, not dozens, but about a dozen about SpaceX. Uh, luckily they're still talking to me. [laughter] No, I mean, I've, you know, I walk around on a beach in my SpaceX shirt, people ask me questions. Um, but, you know, it's just quite stunning. I mean that during that particular vacation, Starship 13 did its thing and it was a sight to >> and um I sat down and watched it with my parents and uh even my mom who doesn't really >> Well, she's coming back. Starship 40 is coming back. Um I think she'll reach uh Australian shows in uh maybe a few days and then across the Pacific thereafter. >> Yeah, maybe >> if she Yeah. >> I mean, what a time to be alive. What a flight that was. I >> was uh it was amazing. >> Fantastic. >> Yeah. But but you're right. People don't understand the company and um >> understand the enabling platform that Starship is and >> you know they don't they don't understand every flight is really a test. You know they think of it as some kind of >> I don't know you see a lot of uh people saying oh it's a marketing stunt or something. It's like what? That's a completely uh no you know you're trying to get to when you try to get to reusability uh you're looking at heavily high frequency iterative process of which are each of these flights you're testing some major components of of these subsystems and and some of them are you know literally existential tests you know like pushing things much hotter faster and bigger than before to look for failure points. >> Yeah. >> And >> and with more urgency, wouldn't you say? I was surprised that they put up those 20 rail Starlinks for a test >> with IoT 13 and now they're going to actually launch uh regular commercial V3s uh with 14 and uh that flight >> and catch the ship as well. >> Yeah, that flight should be um effectively orbital because in order to catch the ship, you got to kind of go around the Earth. So, you're going to get you're going to get into an orbital insertion trajectory for those for those sites. And I think that it would behoove them and they would probably on their plan to push, you know, 20 V3s into real orbit. Um I think people were surprised. You put these things out there and they last 20 to 40 minutes or whatever it was. And but it was a huge test for them and it was a very successful test. They were able to uh unfurl the solar arrays on them and uh power them up and turn them on and get them in the uh you know get get them in the grid of of uh you know turn them on and and get them as part of the entire um Starling system. and uh you know probably did some uh ground tests with uh the the antennas at least to the base stations and potentially uh you know that I'm sure they have a bunch of um private terminals that you know terminals that can access uh specific uh those specific sats you know they have a way to hone in on that and so [snorts] yeah I think it was a really successful test I mean obviously I think pretty clear that they can inject these things or emit these things from the pest dispenser. That seems to be real. Uh very real. Um and uh so there we go. >> And give us these shots, these, >> you know, looking shots. >> I think it's just mind-blowing what they're doing. I mean, like, you know what's really interesting, too, is you think about it is they're using their own tech. I mean, here they are launching um Starlink's satellites. the the irony of getting the photos should be there because uh what's ironic is that they're you know they're using all this video comes to us by virtue of these things connected. So >> yeah, >> what a time to be alive. >> It it's uh it's it's getting real. I mean I know >> you know there's a there's a steady apprehension in the investment community community and the doubting community which not necessarily are the same thing when it comes to uh Elon companies uh you know I mean there's overlap it's a ven diagram of things but um you know it's like a lot of people are like a lot of pieces of the puzzle but too many pieces of the puzzle when do these things kind of come together but man when they come together. They come together in a way that shapes a a puzzle like we've never seen before. And um this is where we're at. You know, I don't know if you're a near-term SpaceX or Tesla investor, okay? Um you're probably a little bit frustrated, but if you're you're going to hang in there for [snorts] a fiveyear plan, you're probably going to have your socks knocked off and in uh stock appreciation. >> Oh, for sure. And I mean, is that an alien dreadnot or what? Tell me that's an alien dreadnot without actually telling me you that's an alien dreadnot. >> Yeah. >> Crazy terap. >> What is Yeah, that's that's a design drop of the terra building uh this morning, which was quite a great little video. I mean, it's it will be the largest building on Earth. It has a highway that goes through the spine of it. >> Yeah. and and a very you not only is it a cool design but the highway goes through it for the purpose of basically feeding the pod you know in other words it's you have a huge you have you have a factory this big you have to feed I don't know if people have ever witnessed this and sort of directly but if you have like large automotive factories or large um any any factory that's of of large scale you'll see um you know dozens of semis pulling in delivering ing the supply of the supply chain that makes up the product. Um, in my neck of the woods, I can drive north of here about 20 minutes and be um, right next to the it's a I guess it's a Tesla factory in Fremont is about 5.5 million square feet um, give or take. Uh, and you see they have a a specific kind of roadway where the right-hand lane is just a lineup of vehicle of semis delivering parts into the factory. >> Yeah. >> And that's happening, you know, literally 24 hours a day. Um, >> and the scale of this is just phenomenal. >> Yeah. This is a roughly 100 million square foot building by to give an idea how big that is. It's 10 times the size of Gigafactory in Texas which a lot of people think is a pretty big building which I think I think Gigafactory on a square footage basis might be the largest factory in the United States. The Boeing factory I think is bigger on a volume basis only because of necessary um height required but um so 10 times that the Pentagon in the United States it's 15 times the size of the Pentagon it square footage of the of the factory alone would would be equal to about 2.5 central parks in New York City. >> Yeah. >> Uh and that doesn't include the property. I would the building I would estimate is about four miles long. So it would take you roughly 75 minutes to walk from one end of the building to the other. >> Yeah. >> Um it's a city. >> Yeah, it's a city and and you know u I see people have already been and I was looking through it and I I spot the Cyber Truck. Uh I cyber cab um the semi. I spot Optimus. >> It's it's just symbolic of everything in the musconomy >> coming together. Mhm. >> Seamlessly all the little bits of the puzzle, the many bits of the puzzle coming together. By >> by the way, you know, if you just >> take a little run into Texas, I mean, we've got gigafactory there. We've got the whole house area, which is going to be 11 plus million square feet for the factory for building the um star mine satellites particularly. Mhm. >> Uh we've got the Houston um Mega Pack Factory came online, I think. Um this I'm sure I'm missing something. Um you know, I think Tesla SpaceX must be Oh, of course we have Starbase. >> Starbase. Yes, of course. >> Gee, I forgot about Starbase. Um a rocket factory in in another part of Texas. Um, >> Texas is becoming quite the epicenter, isn't it? >> I know. I know. >> We've got three trial cities there for robo taxi. Houston, Dallas, Austin. Um, I'm sure we're not done. San Anton's probably coming. >> And, uh, what do they call it? The Texas triangle there. >> Yeah. >> It's amazing. And and by the way, good good on good on Texas and its sensible government to uh >> to attract uh the best factories in the world in on on his properties. >> I mean which sensible government would not let me just flip that around. >> I know a few. I live in a state where that government is the least sensible you could imagine. I live in California. I mean, they've basically >> done everything they could to push Elon out of California. >> Yes. >> You know, um, >> he wouldn't choose to. I mean, you know, Fremont, the Fremont factory is is massive. Of course, there's a reason that it is here and it was the first factory and um, used to be a Toyota GM plant, of course. uh Athorne down in southern Cal sort of remains a uh SpaceX kind of epicenter of of engineering, but a lot of that's moving uh to Texas and of course I I'd say the brain trust of a lot of R&D of course is up here in Palto and >> in my neck of the woods and will remain for obvious economic reasons. I mean just the kind of economy that's built around that type of >> also the the the density of talent that's available there. >> Yeah. It's just it is what it is. Um for the time being but uh but those are expensive employees. I mean a lot of them deserve what they're paid but this [clears throat] is a very expensive place to put the put that and uh think Elon will do his level best to um start moving that out as well slowly over time. I mean, Twitter used to be in San Francisco. That's completely gone. >> Yeah. >> Completely gone. >> As Elon said a while back on X, all my exes live in Texas and they're all coming together. [laughter] >> That's funny. That's funny. So, >> in Grimes County, no less. >> In Grimes. >> Indeed. Indeed. >> Life is a box of ironic uh pretzels. I don't know. As forest comes, life is a box of chocolates. >> You never know what you're going to get. But let's get back to your article and I think uh as I said it's even more pertinent today to discuss these layers the future device. >> Um curious before I ask you to give us a highlevel kind of understanding of what you're what you said here and for everybody watching I'm going to link this article in the description. Please do go and spend time understanding this. Why the future device? Why not the future ecosystem? It's a great question because um I think it's more catchy to use the word device because everybody is their mind trust their mind you know everyone's thinking what is the future device what is the future device you are absolutely correct and observant the reality is that the approach here when you read the article is that unlike building the device and forcing the EOS uh not forcing but over time building the ecosystem This is more of an ecosystem that yields devices. It's kind of a reverse plan. At least I think it will be. So, um >> I mean it's the anti-device. >> It is. It's very weird because uh Optimus, you know, my my my argument here is while Optimus is an extremely physical thing and manifested in its humanoid form, the reality is that Optimus really is the brain of Optimus and that lives in Star Mind. And now that is not to say that there isn't a hell of a lot of AI compute being done on the um you know the body of Optimus and in the in its physical form in the sense that you know an Optimus will be powered by an AI5 chip then an AI6 and so forth. Uh it will use that chip to do a ton of inference. I mean mind popping amount of um inference compute to uh to take in all the sensor data. >> Yeah. >> And and you you need to control the actuators. You need to be able to move the hand that meets the world to the place that you want to deploy the hand for it to be a tool and to beat the world and and >> you know >> Yeah. Yeah. I mean the basic thing is if we if we think of it as the device and let's just focus on on that makes it simple and that's why I titled the article is I mean I basically what I what I'm saying is we have come through in the last you know in the kind of consumer electronics space we've come through in the last I don't know if we call it 50 years for for for sake of argument um the personal computer error the mobile phone or smartphone error and you know now Everybody is saying, "What is that next device?" And obviously, I think there's probably no one left that wouldn't think it's heavily powered by AI. And most people would probably say, "You're crazy. The next device is clearly a wearable or a set of glasses." And I'd say those things will absolutely exist and they'll be made by the players that are already making them and will continue. I think Apple will probably in the next year release a product that's based on its uh vision pro that will be a much more sensible uh lower cost set of glasses and include a lot of the tech from that product. And I think um then you've got Open AI and a few others thinking about these small wearable AI devices that >> watch the world the way you watch the world and record your interactions with the world and hopefully do some good for you. Um, and I think those things will largely exist. And I think they will um never quite get as big as uh the the sort of confab of of uh iPhone and Android, but I think those things will grow to substantial uh market in substantial markets. But my real lean in is I think that Optimus is the big device. And the reason I say so is um you know if we go through the the by the way it seems as I point out a bit of lunacy to make this prediction. I and and I understand why people might think so because really where most of the history of where we've been with this consumer electronics interfacing is is smaller and cheaper over time >> and here it's something that's larger and more expensive. Um but yet there are some things about this and the way I think Tesla will approach it that totally changes that game. And the first one is that it has agency. This is the first consumer device that uh we at at you know at at sort of general scale general purpose that in fact actually can do you know the word do. Uh it's important to recognize that your your smartphone cannot make you an omelet. um it cannot clean your house, it cannot go to the store on your behalf, it cannot serve drinks in a restaurant or cook food. Um so this is the first device that actually has this sort of physical agency and I think that is a huge component of uh the attraction to it because the amount of work needed is really off the charts when you really start to think about it and uh even even in the home setting you know people I think are very well I don't get it you What would you use it in the home for? Well, um, look at just economic the economic offset of having something like this. Um, you pay delivery fees for food and other types of things. >> Um, you go to restaurants and um, out of sometimes convenience, not just entertainment reasons. Um, but what if you had something in your home that could bring you the food and cook the food and clean up after it? You sometimes pay for a gardener or a cleaning lady, uh, cleaning person. Uh, you may pay for a nanny or, uh, someone for elder care or to, uh, be a dog sitter, dog walker. U, all these things add up to a huge number of dollars. >> Sure. And um and now you have something that probably is rentable probably by the hour and possibly very fungeable in that hour. Meaning like you may not need it for 24. You may be able to get it for five hours a day if you only need it for five hours a day and then it >> walks off down the street and goes to somebody else's house. And I mean that maybe quite literally. >> Um >> yeah. So it's the first thing that we can have at our disposal that in fact >> Phil it it it does to labor what the cyber cab does to transport >> right and I think the other thing about it is I think positionally it's very different and what I mean by that is sort of its emergence from a technical platform is different than people think at first you would say okay I think what this is is it's a robot And maybe they sell it for 20,000. I don't think they sell it at all. I think they lease it. But um and you know it it looks like it's this big and I don't know how that could replace my iPhone because after all you know when I get on an airplane I'm not taking Optimus and buying a seat but that's misunderstanding I think what they're doing. I think the the reality is that the mind of optimist lives in in fact lives in star mind and in fact you could initiate uh a product purchase or product relationship with optimist before you have a physical personification of the humanoid >> and in effect I think that's what digital optimist >> I found that fascinating because that is so true because I mean each of uses AI different AI models for different purposes but if you could bring everything together it it's >> it's kind of way are being prepped and we are prepping our future selves for Optimus, >> right, >> for Star Mind, right? And you know the the I get your your article got me thinking and it it kind of this this thing that Elon says time and again the best the best part is no part. It got me thinking what you're really talking about in this article is the best device is no device. >> Right. >> Right. I mean yeah I mean in a sense what I think um you know if you look at we talked about like why is this you know lunacy to suggest that something at a higher price of greater physical size would be the next device when after all most of the trend would be towards miniatururization and and lower cost but the reality is um suddenly you don't need this fairly expensive and um compute ccentric sort of computentric device known as an iPhone or an Android phone. You can have a a lighter weight device that is effectively just a conduit through communications to Star Mind and you know with the >> in other words what what what this is a very interesting product from a different perspective and we should talk about the combination of Tesla and SpaceX in this regard. But SpaceX contributes >> and one more and I will add that once once you finish. >> Okay. Uh SpaceX contributes Starlink and Star Mine and what two pieces of that puzzle and some of the AI. But leaving that aside, uh what two pieces of the puzzle are those? Well, first off, Starlink becomes suddenly um ubiquitous ubiquitous internet, ubiquitous connectivity. And I say ubiquitous because it's a very different world than what we live in today. We think we live in that world. We do not live in that world. We do not live in a world where we can be connected to the cloud or to the universe, the internet on a consistent basis. Um, we do in urban areas, but we certainly don't in the middle of the ocean. We don't at the top of mountains. We don't in live in that kind of universe in rural areas. Starlink makes that >> network be a ubiquitous thing on on terapirma anywhere you are on planet Earth um land or sea or plane. So, um, that's that's something that's a big change. Now, once you have something like that, you might ask, well, what does that connect to? Well, you might say it connects you to a data center. Well, what if that data center is just a few miles above its head and it's called Star Mind and it's an infinite compute plane. So, you've got an infinite infinite communications plane uh, and an infinite compute plane. What do you put in the compute plane? You run instances of Optimus. The Optimus brain runs there and then it's transportable fungeible across devices and one of the devices is the humanoid version of Optimus. That's the thing that we think of when we think of Optimus. But Optimus can live as an app on your iPhone or your Android phone just as easily. And the personality or the interactions that you have with it can start before you get the physical product. You know, you could, it may be that Tesla, the way they start is they say, "Okay, download an app and start your relationship with Optimus. So, when's my bot coming? It'll be there in three weeks. It'll be there in 10 weeks. Oh, you didn't buy it yet? You don't need to buy it yet." Yeah. >> Um, >> then when it shows up, it literally takes on the entire context and everything you've ever had, conversations, and it's all part of that physical being. And then when you need to leave it behind because of um you know physical realities of maybe getting on an airplane, you simply take it with you again in your so-called device uh your you know your your iPhone or Android device. You show up at your destination. You instantiate it in your physical optimist that might be available for renting at the airport or your hotel or wherever. Um and um and that's quite a different thing. And and you notice that uh I wrote a a little piece in there as is Elon building a phone and and I wanted that to be as provocative as it sounds because I I answered it as yes, no, maybe because in effect >> I say yes and I say it's going to be called the star phone. >> Yeah. And and I think uh the reality is the word phone can disappear because >> uh really what the device the device is merely an inference endpoint for Optimus. That's all it is. You know >> one of those communicators from from Strat. >> It is it's effectively that it is effectively that because what it's not is it here's what it's not. It's not a heavyweight compute device. >> Oh no. >> Here's what it is. It's very much a communications device. Yeah. So, it leans in heavy on comm. Why? Because all it is is an endpoint to extend from Star Mind through Starlink to the physical reality of of the thing that you interact with. >> Yeah. >> And that allows Optimus to be very transportable. You know, it can appear in your car. It can appear in the physical persona and it can appear on on >> isn't that what digital optimist is all about? >> Well, digital optimists I think >> oh will evolve to be. >> Yes. Yes. The answer is yes. The the first um the first kind of thought of digital optimist is it's sort of this agentic technology that lives potentially on your desktop and um can drive your screen and do work. Notice the word work here. uh it can do work on your behalf, you know, and and and very horizontally applicable. It's not a programming agent. It can drive Excel as well as it can drive Photoshop. In fact, at some point, it doesn't need Excel or Photoshop because it can be just purely generative in that that space. In other words, >> and that's exactly what Elon said or described as this device, this edge device, right, on the Joe Rogan podcast where he said that you won't have app stores anymore, right? It'll just be an AI device that that generates for you whatever you want on demand >> on on demand. I mean, it's if you think about it, we it's it's a little bit strange that we have apps. I mean, we think of the universe that way. We think we pick up whether you have an Android phone or an iPhone, they're indifferent in the sense that they both are an app ecosystem largely and and yes, they're a communications conduit because these apps all talk to something. But if you want to change your flight on United, you go to the United app. If you want to um book a reservation at a restaurant, you might go to Open Table. If you want to see the weather, you go to the weather app. You want to see the stocks, you go to some stock app. Why are you doing any of those things? Why wouldn't you just be asking the thing or um you know, change my flight? Um why do you want to interact with an interface? Now, it doesn't mean that you don't have a screen. Doesn't mean you wouldn't want to see the information updated. doesn't mean when you open up your so-called iPhone, you wouldn't want to see the weather in your current location and know the ocean temperature if you're going to the beach or whatever the thing might be. Uh but the reality is right now we dig for information and we dig for that information through applications that are very stovepiped within their function set. uh that's going to be blown up and most of the interaction you know part of my article starts off with the idea of what is HMI human machine interface and how has that evolved over time and when we started this journey with consumer devices you know we started with the personal computer I mean that's the first time most of us could get our hands on something um in the compute realm at least and that was enabled by a keyboard and then Apple came along ong through Macintosh and enabled it with another device which was the mouse and and then of course that's evolved and and uh over time and then the evolution of HMI continued or maybe revolution through direct touch uh on our on our on our mobile devices and in particular the invention of multi-touch which really is a stunningly important uh component of the HMI revolution right because >> what would we do without the guey Yeah. And we if we couldn't manipulate the screen directly, we would be stuck in the mouse keyboard land and um and and we are certain extent with our our laptops. So the evolution now to human machine interface at least this next generation is just natural voice uh both directions. You know that optimist can speak to you, you can speak to optimist you can interrupt optimist it can interrupt you. It's very natural. It's very human. And >> so here's that's a beautiful segue to what the layer I wanted to add to your article and to your thought and kind of spark this part of the conversation is that I think this what you've just described is an interim phase because I think 5 years down the line perhaps maybe seven we're going to move from HMIs to BCIS >> right >> and then if you have neurolink blind sight Right. Um, and a few others that URL is is working on then. >> Wow. It just >> that's a crazy future, right? >> I contemplate and have talked and thought about a lot about Neuralink and a brain machine interface and uh and and purposely kept it out of this article. I mean, literally purposely I didn't want to lean in and make it sound too futuristic. I wanted to look at I wanted the reader to come away and say yeah this is possible because I think there's a bunch of readers that will read it and say you're crazy and and you know I'm willing to live with that because um because again I think that there's a a few things that don't fit you know it's a bit of square peg round hole the important reality is and and to your earlier point this is the first I don't know if it's the first the way to think of it this way but perhaps it's the first time when ecosystem emits device as opposed to device fits and develops ecosystem >> and that's a very different thing and it's in a sense because both Tesla and SpaceX as an entity um can approach this it's an extreme mode because no one else has all those piece parts you can't develop this type of ecosystem unless you have Starlink and Star Mind now you might say well you may not need Star you could have a terrestrial data center fine Um, you need Starlink. Uh, oh no, maybe you could just do it with mobile communications technologies. Sure, that's fine. You you don't have ubiquitous internet. You don't have ubiquitous compute. Um, and you don't have largecale compute, you know. So, this is something that's quite different. It's sort of like ecosystem precedes device or emits device. Um now in terms of neural link sure I believe in a universe where the human is more directly connected and doesn't use doesn't need to use what is effectively a very low bandwidth. I mean we are when we communicate >> about 32 kbps is that >> by voice >> uh by thought. Oh, I don't know. I don't I don't >> I think I I remember that's what uh >> Yeah, that's that's a number that Elon had thrown out a couple of years. But I mean this speaking is a lot faster than typing, right? I mean for most people >> speaking is as as a sort of interface modality. Speaking is absolutely faster than typing in terms of communicating information. You can see that I can talk faster and faster if I want to or I can slow it down. Um but the reality is I can only um move data so fast through this mechanism and you've got you know you've got to receive and I've got to send and um and the good news is we can kind of a bit overlap each other not completely you know if you start to talk I can I can be interrupted um so it has some virtues in that world um >> and and certainly optimists would have the same pattern although I think optimists because of the way the tech works, we'll probably be able to listen and process multiple conversations concurrently without the same kind of um you know the same kind of problem we have in terms of that. But but ultimately you are right. I think the the direct connect there's a the brain direct to machine is the direct connect and I think >> um we can start to see the the the shaping of that with Neuralink >> uh and there's some other companies that are trying to build non-invasive versions and maybe Neuralink ultimately becomes non-invasive >> you know it's a very funny thing it has it's it's highly invasive in the sense that it's these little tentacles electrodes that go inside your brain. But it's also in another way pretty um pretty clean, you know. Um and uh certainly for a lot of people with various physical and other types of handicaps, it um it it'll fix their lives, you know, and but I guess the reality is can we command a machine? Can we talk to optimists directly from our brain? Yeah, maybe someday. Um, and maybe that someday is actually sooner than we think, but I want to lay I wanted to put that aside. >> Mhm. >> Yeah. Yeah. I I mean I I feel you completely. You know, I I I face the same thing. I I speak to my friends about this and they think they look at me as if I'm crazy, >> right? I mean, I I think for for many, you know, you need to lay out a plan that makes sense and that they can and it's all, you know what, I think um part of the receiving of new information has to do with comfort. You know, if you if you >> upend somebody's world with fantasiful thoughts of the future and things that start to concern or get people nervous um you know they stop receiving information. >> Yeah. They shut down. Yeah. >> And um you know it's I give you an example where you can go with optimists in that direction. Once you start thinking of this star mind, you immediately start thinking of thoughts of Terminator and you know uh what's the thing called something that um uh you know you start to think of uh an AI that's controlling you and uh you know so I could have said things like um you know here's an example what I think will evolve. Uh so right now I'm thinking of of optimist and its physical or digital manifestation as as a thing like a like a thing with a effectively a personality uh you know like like like a human like an object like you can say I have this iPhone versus that iPhone. I have this iPhone versus that Android phone. I have a relationship with this person versus that person. But there is a point where if you host all these so-called instances in one large uh compute fabric like Star Mind that you can start to think of it as a hive collective, right? >> And and and and what that can do is both good and bad. I mean the good is you can share intelligence in ways that um you know optimists can share intelligence with another optimist in in ways that we started to talk about BMI. We started to talk about that but yet this is just a natural act for the thing right in other words it could communicate to another optimist over a network. You know what's interesting is like you could have two optimists in your house. They could be talking without talking um talking to each other. >> Um we we're already seeing we've seen that. I don't know if you saw the um the demo that figure uh robotic state with the two bots of theirs working in a I mean in a bedroom setting, you know, >> picking up the pillows from the ground and setting the bed and all of that. And they were just looking at each other talking. you know the coordinating >> and yeah working working it coordinating working and um I remember asking Brett Atcock who was on the on on on over the horizon for a live stream when they did that first full autonomous uh shift of sorting and he said well you know it's just it they already can do that it's just that they will have to kind of uh indicate physically in some way just to make the humans around them comfortable or to to let the humans know. >> Um so I think you know that's a signal that it's already happening. >> The other part I just want to circle back to the point about the voice interaction that you making and uh I don't know if you've used the latest uh chat gy voice mode and it is phenomenal >> because you can interrupt it interact with it in real time. >> Correct. >> And it is it just blew my mind the first time I used it. I I I watched the demo um and I was a bit bit of a skeptic because I'd seen these demos before uh released by OpenAI and I mean they they just never matched you know my expectations but this time it was like whoa you know it's just so smooth it's so it and then um if I understand it correctly it's kind of they've got two kind of neural layers where um one is listening while the other is thinking. And it can do that's how it does both together. So I it's it's almost as if we already living in this future, aren't we? >> Yeah. Which you you know effectively what you have is you have these independent asynchronous processes. You know one can receive voice, one can emit voice and one can do let's just call it the LMM LLM processing of the data, right? And all those things obviously need to work together, but they're not serialized necessarily. They're indep. >> Yeah. So, and that's why I said asynchronous with respect to each other because they're not synchronized um in steps. There's not a mode where it goes, you know, years ago, um we used to think of of of computers with modes. And this was back in the early days when we were commanding a computer. It would be in a mode to accept that command and then in that command and we could then directed and so forth. And then one of the things that um came out of Apple's research really research that started at Xerox Park and um a friend of mine Larry Tesler who worked on this and then came to Apple and in fact his license plate on his car for years was no modes >> and [snorts] Yeah. uh and and what he meant was that you could interact without the thing being in a mode of operation that you know like effectively Macintosh was the first no modes computer because you didn't need to be in a particular function to direct another function. You could just go to the menu and select something else. It was kind of a you know we look at it today but if you think about something like interacting with optimist it's exactly the thing it's a no modes model and then and then you can sort of and we think of that as natural interaction but when you go back into the early days of computer you were sort of stuck in these in these modes um so we've broken we break out of that with something like optimist now um certainly optimist can communicate with another optimist that may be present in its physical space or nonpresent in its physical space, meaning it can interact with another optimist literally across the world uh without emitting any voice or listening to anybody else's voice. But it may be that if you had two in your home that you may want to say to them, I only want you to talk. I only want you to communicate through through voice because it may not it may feel weird, you know, the two things coordinating behind your back. um even at your direction still um there there's a lot of um you know one of the things I wanted to point out about this is like any technology that enters our world especially our world at scale um any one of these things in small instance you know I mean like we've had to adapt to these Roomba style uh robots running around our house by, you know, sometimes picking up before it or moving the dog into another room or making sure it doesn't go here, you know, these types of things. Um we always have this uh we any product uh that enters the consumer space u particularly the consumer space needs to really really be I don't know curated to a great degree of uh of acceptance and I think one of the points I make in this article having worked at Apple for 10 years and deeply respected Apple's way of bringing product to people at scale, to consumers at scale, is that, you know, they make uh very purposeful direct action in terms of how that product interacts with people and the ecosystem that develops and so forth. They don't always get it right, but uh but they've led the industry if we're being honest. if we're being honest. And I think that my as as Tesla begins to have Tesla, SpaceX, you know, I'm having a hard time actually really separating the companies in conversation, which literally means that in my mind they're effectively the same. I mean, if >> Well, they are the same. >> If you read this product, they're the same company. They can't be separate companies. Yeah. um there's too many moving parts that are between them. Anyway, leaving that aside, um I said that it it is going to be difficult for Tesla to really have that good approach, the proper brand, the proper I use a weird word here, mouth feel for the product >> to have it accepted at a way that is, you know, it needs all these very subtle qualities. It's futuristic but not dystopian. It's, you know what I mean? There's so many little moving parts here. The name of it to me right now is perfect. Optimus to me means two things. It means optimism and it means agency. >> And I think it captures that perfectly. An example of a product that doesn't capture that perfectly in Tesla's universe is cyber cab or cybert truck. The word cyber, while it's a cool word and Elon thinks it's a cool word. I'm sure he does, but it evokes sort of this um this weird future state that people sometimes are a little uh reticent to want to you know >> to want to participate in because it feels like it's um you know very Terminator or something. Uh and by the way, I think that the names of the products are fine. I'm just saying that there's a very subtle thing that has to happen. You have to make your optimist just right. It's a Goldilocks principle. It can't be too techy and it can't be too cutesy. You know, these robots that have cute little funny little human faces, that doesn't work either if you're being honest. You know, uh if you're going to go to I forget there's a word for this, but if you're going to go totally into that representing the human, then you really need to get there. You need to get >> anthropomorphism. >> Yeah, that's it. You need to get fully there. you know, you can't really get halfway there. So, you shouldn't try to pretend you're there >> or else you're caught in the uncanny valley and that's >> Yeah. Yeah. >> Yeah. Yeah. The uptake is not is problematic in terms of users adopting it. But I mean that's a very very >> I think that's a very pertinent very deeply insightful point that you raise. Um I'm just I stepped back Phil and the other thing I felt when I finished reading your article was like I uh personally from everything that I from my conversations with roboticists and uh founders of of bot companies and people who look at the humanoid robot industry as it develops is that the home is still a bit far away for an ecosystem to deploy these humanoid robots in including Optimus. >> Y >> and that the factories are going to be the first and even that is going to be at least a year year and a half away. >> Yep. And I think >> that's an incredibly interesting point um and something I want to um you know double click on a bit. First off, I think our friend Elon isn't thinking of it the way that way at all. Mhm. >> Um, every time E Milan talks about Optimus, he talks about who wouldn't want their personal CP3 R2-D2. >> Mhm. >> I think what he's thinking is what I'm sort of thinking about how they they're approaching this problem. It is effectively it is a highly human interactive device. It is a it is the ultimate HMI to a human because it presents itself as a human. it is that makes it sort of a consumer device even if the if the purpose is business and I think that's the weird crossover here too you know just like I said it's sort of like an ecosystem that emits product I think it's a consumer product that emits use cases and we start to lose this line between what is business use and what is personal use between certain things >> so labor the line between labor and agency Well, the the line between business labor and look, it's a it's a labor device in one way. And that labor that labor can happen at scale on a let's call it one to many scale, one human directing many bots in a factory setting. >> And that's a very acceptable and often thought of as first entry point for these things to attack the world. you know this is the place this is the the attack surface the beginning place but if you enter the factory with the humanoid bot um you have not developed this bot in a way that solves the human interaction brand consumer kind of problems the these things will be uh for companies that enter through that door they will find themselves trying to build towards the other place and I I think Elon is thinking of it differently. I think he's thinking of it top down. It's acceptable in the home. It's acceptable in a small business. It's acceptable in a nursing home. It's acceptable in a hospital. Interacting with people, interacting with people, interacting with people in service industry, right? It's acceptable in a factory where it's sure it's interacting with people, but it's a different kind of idea. I think when Elon thinks about it and he looks in a factory, he sees robots. He thinks very special purpose robots at doing high velocity things. He also thinks of his workforce. You know, he emits uh, you know, at shift changer emitting a couple thousand people in and out of of of Gigafactory or Tesla uh Fremont factory and and you know, he's obviously saying, okay, that robot can be purposed, my optimist can be do a lot of those skills, a lot of those things, but I just think Elon thinks of it top down, not bottom up. And I think the the bottom up approach while and and and by the way this is not to say in any way, shape or form that Tesla intends to sell it into the home and not the factory That's not what I'm saying. I think >> sure the first couple million of these things end up really in factories and stocking rooms and behind we'll call it you know back of house activities where they're not interacting with humans. But I think he's thinking of it. This is very general purpose. He's made the hands be general purpose. You know, I think he I think he's as happy to deploy one of these inside an airport sitting there playing one of those pianos in the middle of um you know, terminal A uh as he is uh stuffing it in uh the dark uh recesses of a factory uh stocking, you know, a line or something like that. >> Yeah. Yeah. But I think I mean to to me it seems that um the so each of these modes or or features shall we say is or functionalities is um is a component that will be sequential and I think the the components that allow Optimus to be deployed in warehouses and on the factory floor will be the first to be achieved out of Optimus Academy. Um, and I think a lot of the experimentation uh and the the testing phase will continue from the academy onto the factory floor initially, but I think it's it's so much more complicated as you said to deploy it in people's houses because you've got, you know, this acceptance you need to you needed to mesh with the the home ecosystem. You don't want to change the home around the bot. you want the bot to adapt to the home um to the environment to humans in that environment safety issues. I so I feel that um the the the the challenges in developing the components needed to deploy Optimus in the home are underestimated from everything that I've seen uh speaking to you multiple companies that are uh involved in robotics right now. >> Yeah. And um I think I I I I always urge people to kind of stay grounded and I I think um useful work and the lowhanging fruit exists on the factory floor but it's a but even simple lowhanging fruits equivalent to the factory floor in the house have are going to be a lot more complex because of of a lot of things like safety around humans. >> Yeah, that might be the case. I think though um what so we that is what I meant by there is this set of unknowns about how it you know it makes its way into the general consumer space because after all I'm calling it the next device and that means that something that's you know I'm not calling it the next factory thingy um I'm saying it effectively kind of replaces iPhone Android as the next platform device that we use as humans. >> Um I hope that people don't take that as so literal that they think it's sort of this changeover that has to happen next year. >> What I think is true that that Tesla has in the way I've written this article is that again I said this earlier in our conversation that Optimus can introduce itself to the human uh in a virtual way. It can it can it in a digital way >> and only later. So imagine that you spend a good year before. Okay, so imagine this. It's uh 2028. Let's make up a day here, >> January 1st, 2028. And there are a million of these working in Tesla factories and SpaceX factories and somebody else's factory. And then um and then there's uh you know 50,000 deployed into uh Home Depot restocking uh trial you know take it. Well in the meantime there's a digital optimus that's been um emitted by Tesla to the uh the consumer and the consumer starts to build a relationship with that thing. you know, a context over time. It talks