Home robots are coming faster than you think, with Sunday's Tony Zhao

So we have the hand and the gloves. You actually don't have to use a whole robot to be generating data to train the robot. Right now we are actually having like more than 500 people in the US as contractors. And that's how we train the robot without needing to deploy it in the first place. Helper or Terminator. Today's robotic startups are racing ahead and some are more styled in substance, being accused of fake demos and even suspicious data collection practices. So today on Access we've got Sunday robotics founder Tony Zhao. They're the creators of Memo, a lovely little household robot, but unlike everybody else who trains on videos and simulations, they've also got the memory glove, an exact replica of Memo's hand, which over 500 trainers are using to collect data on dozens of household chores. But first we hear from Alex live in Davos at the World Economic Forum, where we discuss nightcaps with big wigs, what was thinking machines thinking, and much more. Welcome to Access. Alex, what's going on over there? You buried in snow and whiskey and heads of state? Pretty much. I do not know what time it is. It's 12.30 a.m. here. What time is it in our life? It is 3.30. And it's a nice peaceful afternoon. I haven't been intimidated by any prime ministers today. Yeah, I guess for those who don't know what we're talking about, I'm in Davos, Switzerland, for the Illuminati IRL, the World Economic Forum, here doing some live interviews for sources, which is fun. And it is a trip being here. This is my first time to Switzerland, and obviously first time at the forum and the conference. And yeah, man, I just took Happy Van Winkle shots next to what I am and the White House's senior tech advisor. So that's been my night. I think they can arrest you for taking shots of Happy Van Winkle. That's expensive stuff. Don't let anybody see you doing that. But yeah, you look cozy. You got like a nice wood panel ceiling. I see the flag in the background. You don't look too much for wear, though. There's kind of like a sepia tone over your entire frame. I don't know if that's just like poor health in general, or just the greatness of the light. It's poor lighting and poor health. Well, is anybody actually talking about economics? Or is it just circular money exchange as we've come to expect from the Illuminati? No, yeah, I mean, it's kind of funny. I feel like, yeah, I feel like I'm just kind of on the sidelines here, because it's such a political thing. I mean, at least half of the G7 is here. Trump will be here tomorrow, and everyone's talking about. Apparently, they jam all the cell signals when he's traveling in an area. So from like one to four tomorrow, there's going to be no cell service. So everyone's like, does your starlink upload work? These are the kinds of things. You feel like you're in like a TV show. It's this just weird bubble that is really hard to explain unless you're here. I mean, I'm staying in a chalet that I'm sharing with the former head of the NSA. Why were you invited? I forgot to ask. I don't even-- Well, look like, yo, we got to have sources.News on the ground. [LAUGHS] No, my friend Ash at Brunswick invited me. And I'm doing interviews at the Brunswick home. Brunswick is a PR firm. But I don't know what this says about me that I can go to a Davos happy hour and like no people there. I think it means that I'm probably have over indexed on networking in my career. To a point where now I should probably just dial back and like hide in the hole somewhere. Or just go off on the mountain here in Ski. I mean, it's beautiful, man. It's, yeah, I'm like, I'm staying on a ski slope. And the juxtaposition of waking up every morning and seeing the most beautiful mountains I've ever seen and then trekking to the KPMG AI house to watch, you know, whatever's going on that day. It's quite something. What's the highest quality water that you've consumed? I mean, the Swiss are very good with the water. I will say the top water's great. They really love sparkling water, everything sparkling. They're an interesting people. They're very hospitable. But also, I would say independent, you know. Like, it's just the kind of a-- That tracks with what I've read on Wikipedia about that-- Well, sure. But like, to get back tonight, I was like, oh, yeah, I need to figure this out. I need to hike up the mountain to come back and pod the view. So-- What elevation are we at? Is that helping with the newsletter writing or hurting it? I don't-- it's like 6,000 feet, whatever that is in meters. It's pretty wild. I mean, you'll walk down the street here. And it's like Eric Schmidt, or like a head of state. And like, instead of an entourage, it's like one person with them. You know, there's just-- it's such a small space to have so many high net worth powerful people, whether it's from tech, business, politics, whatever, everyone's like regulated to the same space in a way that is just very unusual. I've never seen anything like it. So how do you, like, remember all the conversations? Do you have, like, a row of decks that you bring with you in, like, aluminum suitcase? So, like, how do you-- how do you do that these days? You've got to be, like, a context app master. My memory for now is OK. I mean, we'll see after a few nights of this how it is. But my memory is still serving me well. I mean, I talked about this on our CS episode. Like, I have a scratch pad where I just kind of know it app-chat things down, which is only desirable by me. And Granola is my best friend. I mean, I'm Granola meetings. And, you know, I'm kind of shameless about that now. I don't really know how I could do, like, do a live interview and turn around a newsletter three hours later without it. Because, like, the team has not even been able to process the footage from the interview by that point. So, yeah, man, I'm running hard and heavy. It's a lot. Do you know the article that I'm most famous for, Alex? I had probably something with airplanes. No. Back at Business Insider, I incidentally stumbled on the World's Best Slideshow series of Android or iPhone. Your favorite blank revealed. And I did it for, like, politicians, VCs, executives, sports stars. And it just went absolutely gangbusters. And so, I wish I was there to do a beautiful reprise of that series, find out who's actually using the Pixel phone. Or not. Or is it all iPhone, Pro Max, and Orange? It's mostly iPhone. I mean, you've got people from all over the world, right? So, there are actually a decent Android contingency. You could do so many good slideshows, like the dignitaries slipping on ice as they walked down the promenade here. The horrible tidelines on the houses. I mean, they take this little ski town. And like, the companies build these huge facades on the front of them for one week only that become-- it's kind of-- I mean, it's like CS a little bit. But the only store, original store that is still intact and has not been taken over is Rolex. And that's when you know the clientele is a bunch of billionaires. I put this in the newsletter, but it is the only place in the world where you're in the line for the bathroom. And it's like a tech billionaire founder and a head of state and Matt Damon. And that is kind of like this place in the nutshell. I mean, at a party tonight, I literally watched the co-founder of a very famous payments start. I have to go stand in an alt-weight line, because he didn't have the QR code to get in. The dude's worth like $9 billion. It's just like it's absurd, man. This place is absurd. Talking love to hear that. Everyone should wait in the line once in a while. But every-- I mean, I will say like topic-wise, everything here is AI. And-- but it's like AI. It's a bunch of people talking about AI who don't actually know how AI works. So the see people who do, like Dario from Anthropics here. I was watching him in a couple talks today, like Demis, from Google, who I interviewed, for sources. Some of the open AI people, like they're like gods, right? They're just all these hangarons and people just waiting, just like tell me about the impact on the economy. And it's like none of them actually know the impact on the economy, right? But they're here to talk about that. And I don't know. It's just a sign of, I think, how much money and power is in tech these days. Is there a drug of choice at Davos? Or just trading GPU credits? I'm sure there is. I haven't run into it yet. I'm probably not the right parties for me. It's Tylenol and Biden and Zink. But there are stories, of course. I'm sure that's happening. I had a good bourbon tasting tonight. There was lots of good wine. But I don't know. I mean, the Swiss are-- they don't seem like a drugie culture. I mean, maybe I'm wrong, but I don't-- I haven't really sussed it out enough to know. Hard to be precise when you're lagging behind. [MUSIC PLAYING] This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business. As a result, last time spent on opera More time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com/Spotify This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome? That's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online makes sense? There's no place like Chrome. Check responses set up require compatibility and availability. Very 18-plus. Well, back here in the states, we've had some interesting news. Some of which I've read on your newsletter. I feel like the one that was most interesting to me was the one that I knew the least about, which was thinking machines seemingly quasi-imploding before it even launches. Can you tell me some more of that? Yeah, this is Mirror Maraudi, so she was the CCO of OpenAI. She was a CO for like two days when Sam Alman was fired a couple years ago. She started this new NEO lab, which is doing, you know, foundational research in AI, and it's been kind of just kind of puttering around. I mean, they released something, but it was an academic thing, and they've hired really good engineers. They've raised a ton of money. I think the largest seed round in tech history, like $2 billion. Basically, there was this implosion where two of the co-founders left around six employees have left out of like a team of 40 to 50. When AI researchers are, you know, trading like MBA athletes, I guess that matters a lot more than it used to. So, people were super interested in this, and I think it's because a lot of people also look at companies like this that are pre-product, pre-revenue, that are like a talent play in AI, and like, what are you actually going to do? And if your company is your talent, and your talent leaves, what is left, right? And so, I don't know, man, I think it just says, like, how fragile all this is, and there's so few people at the frontier, and like, you know, they all boomeranged back to open AI. The people who left, they all came from open AI, they just went back, and now this really buzzy AI lab is like thrown into disarray. So, I don't really know what happens next, but I do. I don't think there's going to be as many foundational kind of frontier AI labs that are doing well in the future. I just think there's going to be consolidation. It's too expensive, and there's too much attention on it, and drama, you know. There was like drama about, like, there was a relationship that wasn't disclosed or something, and I'm like, "Hello." Oh, it is. But come on, like, every AI researcher in San Francisco was in like a polycule. Like, this, like, arguing that that's the reason you would, like, go, "Zoam on." I mean, I remember, like, one of the open AI co-founders, like, got married in the office. Like, this is just, I don't know. There's just a lot of drama and, like, dirt being slung around in the press on what happened, and I think it's just, there's just so much money at stake. Are you talking about yourself, dirt being slung around in the press? I don't think I contributed any of the salacious details, I will say. I said people were leaving, which is true. But then some others reported, like, "Oh, it was about a relationship," or there was misconduct, but wouldn't say what the misconduct was. Very clearly coming from, you know, the thinking machine's camp. I don't know. When you've been doing this as long as I have, you can very clearly see where narratives are coming from. And Maraudi's camp was very, like, I think, vehemently trying to push back on the fact that her company was imploding by, like, saying, like, "Oh, there was a relationship that happened months ago, and we're just now, like, doing something about." And it's like, "Okay, like, news that'll love them, like, people in AI labs are in relationships with each other." I don't know. Just didn't see that big of a deal to me. Yeah. It's interesting, given that they haven't really talked about much about what they're building. Yeah, right. I mean, I feel like this is one of the. What's your building? Right. And this is, like, one of the refreshing things I feel like about having Tony on the pod who we're going to cut to in a minute here is that, I mean, he's, like, humble about it. And as you'll hear, you know, I've, like, known him for a couple of years, but, like, when I first met him, yeah. And when I, when I first started working with him, you know, he was being super humble about having this, like, apparently, this goaded research paper that a lot of people were ripping off and that they really were on the frontier. And he isn't super loud and proud about it, but it's like, wow, someone actually with academic background and a technical innovation here, and now a couple of years later, they have an impressive product in Brandage just kind of feels like the full story. And that's at a moment where it feels like a lot of companies either have some very fuzzy technical idea or they don't have a technical innovation at all, and it's just a pure, you know, vertical play or an audience play or some type of rapper play. And yeah, it seems unusual these days. Yeah. For sure. His approach is super unusual. I would argue he doesn't really have a product yet. I mean, it sounds like his product is the gloves that we talk about that people get paid to use to train the data. I'm still very skeptical of, like, these household robots being in homes very soon. I don't know. We talk into a lot of robotics founders here in Davos recently, and I just, I think there's a lot of hard problems here, but I do think he's approaching it a super original way. And the branding is sick. The glove idea is very unique. He's obviously an OG in the field, and Sunday, I think, is considered in Silicon Valley to be, you know, one of the key companies and robotics that everyone's looking to. So definitely a great guest and cool that you worked on that and helped him kind of position all that back in the day. I don't know. It's cool to see it finally come out, right? I mean, honestly, I like at some point a year, a year and a half later. I don't even know what I contributed. Sometimes, sometimes these companies, I mean, I work with so many startups at a time in various capacities that sometimes they launch and I'm like, did I write that? And I actually actually have to search my notion for the tagline sometimes. And I'm like, sometimes it's there. Sometimes it's not. And as Tony said, even just kind of the helperverse terminator's economy, if it was even just that, I can be proud of that. I mean, that's what I try to do, obviously I can't sell it per se, but that's what I'm proud of is trying to guide people toward a more positive, less cynical vision. And to the extent that people feel that it helped them, awesome. But no, I did not come up with the name. Back then, we didn't actually talk about it on the podcast, but back then, they were called LemmyBot. Like, Lemmy do that for you, which is pretty cute and catchy in its own way. I think Sunday is just kind of bad-ass because, I mean, it's just one of those names that's like super simple and evocative and all the chores happen on Sunday. And so it's just a little too perfect, but LemmyBot was cute as well. LemmyBot's cute. Yeah. I mean, it's like access, right? Like, simple evocative and you listen to pods doing the chores on Sunday. I hope so. Shall we cut to the conversation with Tony? Yeah, let's do it. This episode is brought to you by Accenture. Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics and smarter workflows to simplify campaign delivery and access better data across the business. The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com/Spotify. This episode is brought to you by Google Chrome. That's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block or finally break down that long article you've had open for weeks. Ready to make anything online make sense? Check responses set up require compatibility and availability very 16 plus. How are things going in Robo World since I saw you last? I feel like you've gone through a couple company names at this point since I last saw you. When we work together, I think it was September 2024, I mean, I feel like so much has changed. Yeah, yeah. It was good. I think the naming thing is kind of funny, that we've went through like, I would say like three to four names. Like a week before we announced other stuff, we're a cost Saturday. And then that's when we already have all the marketing materials, all the website and all the logos, everything ready. And then we realized we made a very dumb mistake of like, hey, there is this other company also cost Saturday robotics and they have this trademark there. We realized that we should probably not be doing it. And like, I think everybody was like, it's all hands on deck, let's change our names. We have like seven days left, change all the swag, change all the websites and I think we did it. Yeah. And I actually think that Sunday might be a slightly better name than Saturday. So do you need a lawyer or do you need a new lawyer? Yeah. Like, it is a fun journey. Yeah. Well, I feel like you netted out with something kind of perfect. I mean, my favorite brands and tag lines are always subverting something unexpected. And Sunday is like the day of the week people hate the most. And I think that's it. I think there are a lot of people who would say, oh, we wanna avoid the negativity. And it's like, well, not if you're gonna solve it, you know? And I think it worked out well. I mean, was that the intention behind the name? - You think about like laundry and you think about like Monday, right? And it's actually bad if you are worried about Monday, well, at the same time, you need to do all your chores. So yeah, I think our goal is to give the Sunday respect to people as they can actually enjoy it, just like Saturday. - It fuck Monday.com. We're preempting with Sunday.com. - Tony was, was Elsa's idea Saturday or what did he contribute? Was did he have you on Friday at one point? (laughing) - Well, this has been really helpful. I think we've been talking to each other like way before we're even playing to be out of stealth, right? I remember we're having this early conversation about how should we position ourselves as a company, what kind of vibes like our brand should have and why we should not be building another terminator robots and try to put in people's homes. I think we're allowing very well on this route. - Yeah, we were talking about helpers, not terminators, but I mean, that was honestly just like full trust at that point because Tony had no robot and no brand at that point. And I was just like, if it's gonna look like a helper, then yeah, let's do that. And I feel like it turned out to be so cute and I've been playing so many Nintendo games with my four year old. I've come to realize that iconic oval shaped Kirby eye shape that you're kind of borrowing from a little bit or like the eyes on all of, I mean, Nintendo puts those eyes on everything. Whether in Mario Wonder, it's like a pipe or a star or anything else. How did you arrive at that? Who designed the face? - Yeah, so we have an industrial designer in-house that didn't like all the designs. But I think the funny thing here is like, it almost feels like slightly controversial then now the robot has a face because most of the human robots actually does not have a face. If I saw the G1 unitary robot, it's hollow. There's nothing. Or it's like a fake like a black mask, right? So I think we're almost doing the obvious thing that like if there's going to be robots everywhere, they need to talk to them, they should have a face so that you can actually make eye contact. But again, like, you know, all the great work coming from Shai, our industrial designer. And he kind of like, I saw the evolution of this whole design kind of starting from the hand, from the gloves that we're making that we probably should talk like a lot more about in this podcast as well. But kind of seeing it kind of design style propagated to the arms, which are also like kind of rectangular and smooth and to the body and eventually the face and the hats. - The other thing that jumped out at me from our collab, which feels like a million years ago at this point, was that your emphasis on time, I think was a little different than a lot of folks in tech. I mean, I don't want to like throw anybody under the bus, but it's like, I've just worked with so many AI productivity apps that talk about saving you time. And for the most part, when you're at work saving time, you're just gonna do more work, right? Whereas, with your saving time at home, it's almost like, oh, I could just go do something else now. It's like almost a completely different paradigm that tech isn't usually touching with that proposition. - Yeah, yeah, absolutely. It's kind of the thing that actually people don't want to do. Nobody want to do the chores. And yeah, I think they are just, it would be kind of ridiculous if the doctors, Autumn, I'm a journalist or like a scientist are replaced by AI's and we're left with autochorus. It just doesn't make sense. - Hey, hey, hey, hey, journalists are replaced by AI. Oh, God. - Alex, you're gonna be stuck doing laundry. That's gonna be your full-time profession. It's gonna be a subscription service. - My wife makes me do laundry, so it's, you know, there's nothing new. I will happily do it, it's fine. - Yeah, speaking of laundry, my very first job was at Urban Outfitters. I don't know if either of you know this, but I was scolded early on for not folding shirts well enough. And I will just like take that insecurity to my grave and like I go so fucking hard folding my shirts now. Like I started learning to fold with a board. I don't know if you guys even know that there is a folding board. - Wait, you do that? You actually fold with a board? - No, I'm good enough. Now I'm a 10x folder now. But like, I mean, I was wondering like Tony, is there a specific folding methodology that you are adopting? Are there any haters for this methodology? It's kind of weirdly controversial, depending on where you are in the world. How you fold your shirts? - Yeah. - I think folding is, I think relatively people agree on what is a good fold looking like. It doesn't like where you're at. There's more to urban affiders, but go ahead. But I think the place you have the most amount of tension is how like dishwasher should be loaded. Essentially, like, especially when they are like a ton of dishes, right? But I think when we actually need to design these tasks when we teach the robot, it's not just like, hey, we get a lot of people, collect data, whatever you want it to be, and then we just put whatever the end result is and into the robot, that's not what we do. And it's actually a big trade-off between whether the specific way we fold is easy to learn for the robot. And whether on the product side, it's actually something people want. And at the same time, it shouldn't be like, like ridiculously difficult that it might take like infinite data to learn. And I think we see that commonly that when we make all these decisions about the robot, it always tends to be like three or four categories of things that we need to jointly consider. And even how to fold the t-shirt was like a big part of it. - I know there's drama around the dishwasher methodology as you said it. Actually, I found a meme online that I see go around every so often. It was from Twitter. This person said in every partnership, there is a person who stacks the dishwasher like a Scandinavian architect, that's me. And the other person who stacks the dishwasher like a raccoon on meth. I would apologize to my wife, but she's not listening. - Oh, yeah, Chloe's the same way. So you are both like dishwasher orderly people. That's interesting to know about us. 'Cause we're different in a lot of ways, but we do that the exact same way. - The right way, you mean? - The right way. Glasses in the back, you have to put the stuff that's dirty first in the back. You don't want to put it in the front. You can tell someone's personality by how they stack dishes whether it's in the front of the back in my opinion. - Well, does memo have a point of view on whether you should rinse before loading or load with the food on? 'Cause I know there's controversy around that as well with the Cascade Platinum Reddit crew. - At context here, there's this whole, I think a whole debate, right? Of like the dishwasher will be able to clean your dishes better if you don't rinse as much. There's a whole like theory around it. But I think for us, like when we teach a memo, I think it will rinse. We do not have like big chunk of food and I'm inside a dishwasher, which I think is more, there's more downside than the outside I can give. - Well, you know that's big dishwasher that tells you that you need to rinse. That's the lot in operation. - That's a Bosch Enco. - For those who missed the last episode Bosch's tagline from CES, the more you Bosch, the more you feel like a Bosch. (laughing) That's not a joke. That was their actual tagline from the last time around. Well, I know like a lot of the models for robots and otherwise, we also talk to the CEO of Rivian recently. They talk about how they don't actually program rules of any kind. Bages gather more data. Is that similar with you guys as well? - With the glove. - Absolutely, yeah. It's just that I think if you want robots in the real world, in the open world, like you cannot, if else, all the things that can happen and trying to engineer that is kind of, you'll be kind of cursed if you try to enumerate all the things that can go wrong. But the nice thing about data is that all of these, if and else are somewhat included inside the data. So if we can get very diverse high quality data and enough quantity of them, we don't actually need to do much like real coding, but the newer network will learn from data and that will end up being better. - As someone who's very close to this stuff, are you more or less optimistic than the average person about riding in a waymo? Or Tesla self-driving? - I own Tesla. I self-drive to to office every day. - Used to work at Tesla, right? - Yes, I did. I spent a summer there back in 2022. That was actually before the whole end-to-end scheme happening there. Back when things are not working so well, like these days they work very well. I feel like I get more optimistic when seeing things that doesn't work and believing that it will go to 99%. But going from that to robot taxi that is fully, fully autonomous and can see. skillably be deployed in any place. I think there, people might have underestimated how much work there. So I think I'm more optimistic for always the first chunk of it, and the second lag is more challenging. - If we could get a little more technical for a second, and robotics are in the brain for me, Tony, 'cause today I interviewed Demis Ishabis. You used to work a deep mind, right? - Yep. - So Demis and then the skill AI, which just raised a big round and is doing robotics in the enterprise, and everyone is approaching it differently, but Demis and Patak from skilled were saying, like, they think enterprise is the right way to attack this problem from a data perspective. And it's interesting to see you do it from the home, you know, going towards the consumer route. And I'm curious like how you landed on that as the way that you all were gonna go versus just doing the more kind of boring, but stable, safe enterprise route. - Yeah. So I think this is maybe a hot tick, but I think there is a myth around like industrial automation or like, in factories, it's an easier thing for the robot to do. And that might be intuitive for people, right? There's more constraints, like there's clearly well-defined things that you should do. But I think the broader context here is more like, if you want to do robotics like pretty much everything is hard. Like a hundred percent of the things will be hard. And putting in that context, I think the relative, like the hardness difference between going for the industrial automation or going for like homes, I think it's small in that perspective. And then I think the question becomes like, hey, are we really asking the right question of like trying to find things that are easy because everything is hard. And what we end up with there is like, I think we just want to work on something that is really inspiring and exciting to people. And that will help us build the best team and that will help us gather people who are like truly excited about the mission and like putting the actual work, putting like all the ingenuities into making that a reality. So I think that's the, maybe the hot tick here is in a grain scheme of things. I think being like 20%, 30% harder or easier really doesn't matter that much. And I think there is also a different viewpoint here when it comes to home versus like say factories, right? One of the big thesis we have is, we believe robots will win on the availability before they win on a sheer like productivity or like the speed. So you're thinking about like factory automations, you're essentially competing with hourly wage workers and essentially humans. And human hands are so perfect after like millions years of revolution and we're so intelligent. And it's actually really hard to beat the throughput and the precision that human has. But these are absolutely critical in factories because if you're slow, like the whole line will be blocked by you. And if your successor is not high enough, maybe you'll be damaging some parts and that's like causing real economic damages, right? So if we think about it in that context, home actually becomes quite easy because the type of work that we need to do is not like rocket science. It's not like assembly and iPhone. It's actually things that a six year old knows how to do. And it's one of the few places that you can provide a lot of value by like maybe giving people a cup of water when they're tired and maybe refilling that water like time over time, making sure it's always like the perfect temperature and perfect volume there. The chores are they come like every day. Every day there is new like 30 minutes to an hour of chores getting invented in your homes. But you cannot just have a housekeeper to come in for like 30 minutes and you let them go, right? You either are like super rich and you have someone in your home the whole time there you're like a housekeeper, you pay like a full time salary for that or you actually need to do it yourself. And so the time we save becomes not like competing with an hourly wage worker, but becomes your opportunity cost of like hey, if every day after work you have like three hours you can spend. And if you can save one more hour there is actually quite significant. So it becomes like almost like a labor arbitrage that if you're making more, the robot will be more useful to you earlier. And that allows us to kind of get into the markets like gradually as opposed to like we're spot on competing with human dexterity like from day one. - I get so mad at people who have three hours after work every day 'cause clearly they don't have kids or they're super rich as you said and they have their own assistant or butler or something. I was like I have approximately like 43 minutes of free time per day. And if I have some music on you know doing the dishes or whatever is not the worst thing in the world but when that is very literally the time that I have to like read a book or play a game or spend the time with my wife it's like that is very valuable time. And so I do see some need there and I know the rumor is that Apple's working on something like this as well, right? - Yeah, I think pretty much everyone is. If you look at the latest demos coming from Tesla from figure is all about the homes, right? - Do you think we actually have robots in the homes of like everyday normal people though in the next even five years? I don't know like when you, what you have is so well thought out. It's like the videos, the branding, the positioning it really resonates, it's very intentional. Shout out to Ellis, I'm sure for a lot of that too. But when do you think this actually shows up for people in a way that's like cost conscious, that's accessible, that's scalable? - Yeah, so I think the way we think about it is that at the hardware cost get lower and as the capability of the robot in terms of like the AI, the newer network, expand over time. Let's say in the beginning and do one task but over time we can do like a thousand tasks, right? I think the product itself is inevitable that if it costs you a few thousand dollars and they can do it all the stores in your home, I think everybody will love it, right? But then the question really becomes like, how can we get there? I think we all agree that the end goal is very clear. Can we get the cost down fast enough? Can we get the AI capability to go up fast enough? So I think I would say like for us, we certainly believe that home robots will be in your neighbor's home, like maybe sooner than you think, definitely within the next five years. And we're actually running a beta program this year towards the end of it to be putting it in real people's homes. That's actually the thing that really drives us as the team is that it's not 100% sure yet, it's not like set in stone. So we're almost like driven by this agency to see how can we optimize things a little bit better so we can get this out of the door a little bit sooner. So I think it's probably the excitement that we have is that it's not clear yet. (upbeat music) - This episode is brought to you by Accenture. The result? - This episode is brought to you by Google Chrome. That's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50 page restoration block or finally break down that long article you've had open for weeks. Check responses set up require compatibility and availability, very 16 plus. (gentle music) - The Neo launch was super interesting to me 'cause again, it was really good branding. The keynote was awesome, and then you realize it's like, oh no, they wanna charge you like $20,000 to do data collection in your home with a tele-operated robot. I don't know if that's your strategy with Sunday. Sorry if it is. That's, I think that's not gonna work at scale. I think like you can sucker a few early adopters into doing that, but you have to have value and not like trick people into actually being the customer, right? In a way that they don't know that they're the customer. Like, you know, you can buy the robot, but you're actually feeding the data back to the company and that's the reason they're doing it. You're actually giving them more value than they're giving you, right? And you don't wanna do that as soon with Sunday. - Yeah, I think for us is our opinion is, it's not worth shipping a not autonomous robot. And the reality is like, when people think of a robot, robot should be autonomous. It should be by definition autonomous, right? It's a robot. It's not like a human wearing the shirt of the robot. - Okay, so I think what, but the reality is, right? Like, if we do not have data, like the AI isn't there and without the AI will not be autonomous. So we have almost this like chicken air problem there that without deployment you cannot get a robot to your autonomous enough to be useful. So we were actually like thinking about this problem since the early days of like how can we break this chicken air problems. And what we end up with is the whole idea of using gloves to collect data. So it's actually instead of like getting robot to your home and get someone else to like log into your robot and collect data. All we do is we build these gloves. Actually have one here right here. It just gave us the finger. Thanks so much Tony. So we have the hand actually here and the gloves. So the goal here is that you actually don't have to use a whole robot to be generating data to train the robot. All you need is to wear these gloves and do the task and we'll be able to like leverage that train the robot. So what that allows us to do is we don't have to get data solely from our customers. Right now we are actually having like more than 500 people in the US as contractors and it's part of their work to be like just wearing the glove and do the chores and we'll pay them for that. And that's how we train the robot without like needing to deploy it in the first place. How much are we paying for this? I mean can you hook a brother up? I mean I'm an entrepreneur now. I need some side income. I think they actually pay very well in terms of a wall clock time. Every hour they spend, they get paid like $40-ish, which is very good. And the additional nice thing, I think it's actually a very good job as in like you don't need to leave your home. You can just stay there and you can do it anytime. You do it for like 10 a.m. in the morning, 10 a.m. in the afternoon, and 10 a.m. in the night. And we don't control for that. All we need is to get like this amount of well-instructed like you know, clean data from your home, which brings us the diversity and the dexterity that's needed. Is that like an ongoing project in terms of the training or do you feel like at some point it's going to magically be done? That's a very good question. I think I think we'll be. I think the gloves will be here to say, we're going to be using it for a very long period of time. Just because, if you think about places where AI works the best, it is when it has almost access to unlimited amount of data. We have the whole internet to train the large language models, and these ASP are still paying like mercrose, like scale, like crazy amount of money to label their data. I don't think robotics will be a kind of exception to that. And is probably even in the other direction of like in robotics, we do not have a whole internet scale robot data to start with. So we have even more work that we need to do. So I think we need to like, yeah, like the data collectors and where the gloves, we're expanding them actually quite quite a bit this year so that we can. How many people are wearing gloves doing dishes right at this very moment, do you think? But how do their wives feel about it? Sometimes both them and their wife do it together, which is. That's really it. That's 80 bucks an hour. So at this point, we have around like slightly above 500 people all in the US collecting data for us. And we're going into the thousands this year. Yeah, and I know there is the design of the product, the brand, some stuff on the software side, I'm sure, and how you interact with it. But like, what percent of your moat would you say is this data that you're collecting with the glove? Because I'm not sure that anybody else is doing that, are they? Yeah. So I think our approach of going on glove, that is co-design with a hand, I actually. I don't think I'm aware of any company that is doing something. I think there are similar approaches, but not exactly the same. Everyone in robotics talks about the hand is a problem, right? Like it's one of the top problems, but you're right. Like to my knowledge, no one is doing this as a data source like you are. How can that be? Yeah, and this is actually kind of reflective of a broader thing of like how we think about systems in robotics. So people tend to think about robotics as like singular things like the hand, now the data, and maybe the legs, and like something like that. But it's actually a lot more interconnected than it looked like. So for example, if like it could be true that let's say the bottleneck is the hand, and you can build a very very dexterous hand. But the problem is like how can you get a lot of data for it? Like if without a data, the whole thing doesn't work, no matter how good the hand is, right? So what we do is you can make a really nice fake video if you want. I feel like many have been accused of that in the last year. Yeah, I will not comment on that. Our video. But I think it's actually more about like instead of building the most dexterous hand and call it a day, and that's the optimization that we want to do. We're like, hey, can we build a hand that is so capable, but we can also get a lot of data to unblock the AI part. So when we design the hand, it's actually not about like, hey, let's get the most amount of fingers out of it. Let's get five fingers. Let's get like 22, 27, whatever, degree of freedoms and pack all the motors in. But can we design it like together with the glove? And the glove is something that people can't wear and efficiently pluck data. So I think when it comes to like for us in terms of the mode as a company, I actually think it's less about whatever data that we have because we're going to have so much more than that. But always approaching these problems with a global knowledge of like AI of software of hardware and like the whole thing together like controls. And that kind of always tears us to making decisions that is slightly different from the rest of the field. And I think that will be the kind of the core competency that will continue to to own. And that will continue to help us make bets that are somewhat popular in the beginning, but our post office. Yeah, I mean, what can you tell us about the other stuff we've seen in the space? I know there's been different demos that kind of have or haven't been real visualizations of robots capabilities. I mean, does that bother you when you see that stuff? Like what type of stuff seems not real when you see it going viral on Twitter? The way I think about it right now is like videos are easy. But like uncut videos are slightly harder, but live demo is the boss. And real deployment is the final boss, right? So if you look at videos like let it be from researchers or companies, it could be the one time that works out of like a thousand Chinese, right? The success rate could literally be one percent. But as long as we interpret one video, it looks like things work out of time, which is like not really true. But if we move along the spectrum, and we see, for example, I see us, there are companies that are doing live demos in front of like hundreds and thousands of people. And I think that's something that I respect a lot, because that is when you have the true confidence to like show it in front of people, as opposed to like, like hedging it and be like, hey, as long as we get a video out, we're all good here. Yeah, did we miss you at CES? Yeah, were you there? We saw ping pong. We saw they were boxing. Yeah, it was cool. It was cool. Well, what are that stuff is real and what is it? I would say the ones that I saw that are as CES, namely like the, for example, in the boxing, and some of the like, I think grasping objects from the shop, I think they're all real. They're all two and only driven by real AI. They're not like, there's nobody. Well, the boxing one has someone like kind of remote controlling it, but the controls of the, like the fact that the robot is not falling down is autonomous. And to acknowledge the audio autonomous demos of the like, there's a folding a windmill and like taking photos with the texture's hand. Those are all like real and autonomous. Yeah, it can be tough to tell. Sometimes the outside looks shinier than then, you know, if you will, how well the inside works. If you look at one of the ones that we saw at CES, the unfortunately named LG Cloyd kind of has a cute face. I mean, is that the one that looks like most similar to memo? I'm not sure exactly what it can do or how it works. But it looks polished and I think as the average person, you're like, oh, it certainly looks like it works. But I mean, are you basically telling us that without the glove data, some of these household chores just aren't possible yet for anybody else? I think for now, that totally seems to be the case. And I think I will have, it would change my mind if they are willing to live demo it like many, many times in front of the, like, the audience. Yeah, speaking of memos, definitely a better name than Cloyd. Imagine welcoming guests to your house. Please come on in, allow Cloyd to serve us some amuse bush. Sounds a little like, I don't know, like, some evil things. I mean, the summer is summer, like, actively going for the evil vibes or the Transformers vibes, like Optimus or this or that. And like, Optimus is going to fall over and squash your cat. Like, Elon doesn't really seem to care. about, you know, that stuff in the short term, but if you look at it, I mean, logically, the way that you design it shows you what its goal is. And when I look at some of the others, the goal is to look impressive. And I mean, I assume you guys looked at the legs and you're like, oh, we don't need that unless we want to be vain about it. Yeah. And I think a lot is about really the safety. And if we want like flashy demos, the legs are great. They look like humans. But, and I think also people wouldn't disagree on the fact that like, if you look 10 years into the feature, like the robot may all have legs, we're all converging to that eventually. But I think the question is more about like, do we, can we earn the trust of real consumers or other customers with whatever the robot that we're building right now? And that will give us the ticket to build that ultimate robot that is like, can do everything, right? And I think just prioritizing safety in a, almost like a passive way that like, if we just unplug the robot, it's battery while he's doing something and we'll not fall over, is going to be like, helping us gain the trust so that people know that we consider those edge cases. Yeah. I saw on your site, you have something called compliant control, which is that you could interrupt it and it won't just like fall over this or that. How exactly does that work? Like, can you just kind of shoe its hand out of the way or how does that work? No more drinks for Alex. He's, he's being cut off. He did that tonight. He did that tonight, Chris. I think the idea there is, like when the robot is doing certain things, right, it's actually delivering the hand to where it wants to be. But the rest of the body, let's say, where elbow is doesn't have to be at the exact elbow position where it is right now, right? Let's say you're, you have the kids and they bump into the robot. The robot shouldn't be there, like, rigid, but it should kind of tucked in and maybe like, you know, when someone hit it. So this is when we mean by compliance that it is not like a industrial robot that all it does is to follow the rigid instructions going from point A to point B and no matter what you put in front of it, it will destroy it, right? That's not how we think a home robot should be. I think to achieve that is again, a mix of like very, very deep into the system side that we need to choose the right hardware and we need to do, build the right electoral system so that we can enable the hardware to be compliant. And then our model also need to consider that. So it's always like a complicated answer that involves all the parts of the system that we need to like put in like right at the beginning. I think I need to upgrade my hardware because I break glasses and plates constantly, expensive ones. And it just makes you wonder, like, are we going to see silverware, forks, knives, plates, bowls specifically designed for robot? Robot friendly. Yeah. Yeah, robot friendly. Or do we not have to worry about that? Because memo can handle your grandma's China just fine. I think that's a, there's a very good question of whether there, there exists like robot friendly, like utensils or, um, oxo good grips, the most goaded brand in all of kitchens worldwide. With those little rubber grips on them, I love that brand. Yeah. And I think maybe the broader question here is how can we, it's like when we deploy a robot into people's homes, I think the adaptation of course the robot will adapt to how you live your lives, right? But I think human will also adapt their behavior to the robot. For example, like, let's say you have a butter knife that is very thin. And for some reason, after you eat, you just really want to put it on a table and it's essentially flat with a table and the robot really struggles to pick it up, right? And what you could do is like, you know, keep doing this and the robot keeps struggling. Maybe it takes like five minutes to pick it up. But what you could also do is put it on top of a plate. And then it's infinitely easier for the robot to pick up without you actually intentionally trying to do things that are so different. Um, so I think there will be things like this happening in the future that we will also kind of evolve our lives out a little bit in exchange for more convenience that the robot can bring us. For example, when you're choosing a new coffee machine and you realize there's one one of them, the robot knows how to use but the other one doesn't and they maybe produce like a similar amount of like quality of coffee. And I think people will go for the ones that they can have the robot to do. Right? Um, so I think it's an exciting future for us as well as that we don't need to handle like a million types of coffee machines, but we all like only need to handle the ones that are more popular and we can kind of change user's behavior a little bit by steering them into getting like the ones that works well. How long before we get to what you're talking about, I mean, I've asked before like, when are these going to be in homes, but, uh, is this like a 10 year vision? Is this a five year vision? Like how far out are we from this actually in your mind? Yeah, it's definitely less than five years. Um, and really? Yep. Wow. Because like, if you think about where are the fundamental bottleneck is, is data? Like, I think the, um, I'll put another hot take here is I don't think robotics like training AI to, um, to like do households, right? It's fundamentally harder than like, math Olympiates or like all the crazy reasoning models that we see right there. In terms of the fundamental AI architecture, I think the rest of the fields like large image models were vision are actually quite a hat. And the only reason that the robots cannot do the chores right now is the lack of data that is almost like the singular biggest bottleneck that is hurting progress. Because without the data, you cannot put in like, do very meaningful research to iterate on figuring out what are the right methods, uh, and figuring out the science questions. We're just like blocked there. But once we're able to unblock the data part, I think, I think the capability will be more like an exponential growth than like, we're kind of stuck somewhere. Uh, so that's like kind of how we think about is all about the data. And we want to get as much data as diverse data as possible. And you think the best way to get the data is the hand versus just like training on videos of humans interacting with the real world or all that, you know, how Elon talks about FSD is the way that you're going to learn for optimists. Like, everyone has a theory of how to get the training data right now, right? And yours is a very unique one. I gotta have to give it to you. Like, I haven't seen anyone else approach it that way. And you obviously have conviction that that is the way. And I'm curious, like, you probably evaluated all the other ones. Like, are all the other ones just dead ends? I don't think that that's how I'll put it, uh, is I think it's true that right now, I think robotics is still at the, at the fun part that, uh, people are debating, right? And these days in, in large, one of people don't debate about like, whether you need really, right? They're like, this is, like, give me money. I'll give you the course one in results, right? We've got a fun part that people disagree that, um, there are many fun ideas. There are, uh, people are like kind of trying to, like, combat each other with narratives, right? Essentially. Uh, but given the amount of information and the concrete evidence that we saw, we think dealing with gloves, like, collected data, transferring to the robot is the best way. But this is not, it's a strong opinion that is, I would say, like, weekly held. As in, if there is another breakthrough that we see that changes some of the assumptions, we're more than willing to, like, update our prior and figure out how the best leverage that, right? And that can come in a lot of different flavors, uh, let's say world models that people talk a lot about these days. Uh, it could be simulation. It could be, like, anything. Um, but the piece that will never change in my opinion is, like, real high quality data. And that is going to be the concrete thing that we'll have either way, uh, that would differentiate us, no matter what the downstream algorithm is. Let it be, like, training the robot directly with, like, reinforcement learning, with imitation learning, work models. I think we're largely open-minded about how to use it. And in five years, do you have a sense of what market it's going to hit in first? I mean, I feel like intuitively, I would think, you know, elderly folks, like people who have the most trouble, uh, with the chores and may need it most. Is that your instinct as well? For the early, uh, customers, we do think there will be a period of time, especially in the beginning, that the robot will give them more trouble than it helps, right? And in the beginning, maybe it's like, you need to learn how to do things in a way that the, the user wants. And in the, it might make mistakes. So, I would say, when it comes to the early customers that will be addressing, I don't think we'll be aiming at people who, like, almost, like, desperately or, like, have to use it. But we'll start with people that the robot is very nice to have. But, you know, just elevates their, their standard of living, uh, by a lot. Um, so I don't think we'll be going for, like, elderly care, uh, from day one. But that's eventually the goal for sure. What's the robot narrative at Davos, Alex? Oh my god. I mean, it's all like jobs, you know, it's like, this is, yeah, this is such a different group than what we normally talk about on the podcast. But I do think everyone's like, oh, you know, if robotics really hits, and there's millions of these robots out in the world, you know, these are a bunch of policy folk, like how's that going to impact hiring and GDP growth and all these things. And I don't know, everyone in tech that's building this stuff seems to think that it's actually going to be an accelerant for everything that it's not going to hurt economies, but it will free up more people, you know, to do things that they want to do or just create net new jobs. So yeah, there's a lot of like, dueling narratives and dueling viewpoints. But I think everybody agrees that like, robotics is at an inflection point, right? And that's like, that seems to be the consensus here for sure. Tony, like, I assume you agree with that. You wouldn't be doing this company if you weren't, you know, believing that. But yeah, like, how do you, how do you feel about the jobs piece of it? Do you feel like we're going to find different kinds of jobs, more fulfilling kinds of jobs that get taken away, you know, from robots or how does that work? Yeah, for sure. I think the whole mission for us is to give people time back so that they can enjoy, they can spend time with their families, they can spend time with their interests, right? And I think there's almost like, like, not human part of doing chores that is so repetitive, that is so boring. It doesn't highlight what makes us unique as humans. But at the same time, I think it's like because of our positioning and the current type of work that we want to tackle is really not going to replace anyone's job at all. It's all about, like, saving people's time because they need to do it themselves, right? We're replacing, like, a software engineer's time to do, like, course as opposed to people who really need to feed their families, to do, like, housekeeping. I don't think we're touching their, kind of, their unique proposition at all because, again, it might be funny to say, but we as a robotic company, we do not bet against human dexterity because it's so good. And we always find place that we can compliment them as opposed to replacing them because, like, from hardware, software AI, we're not close in terms of getting, like, a human hand, let's say. That's why people are talking about it, like, the hand is the bottleneck, right? Yeah. Yeah, I wonder what simple chores you now have the most respect for after digging into the hand dexterity, or otherwise, it's like, oh man, cleaning a toilet, unbelievable skill. Yeah. I think, just even just using, like, sure, folding as an example, right? We can fold these shirts in the mid-air. We fold in a very dynamic way. We like fling it. We sort of hand to, like, kind of, tuck it in. And these are things that are very hard to learn for the robots. I think I definitely come to appreciate that more, like, as we work on these tasks. The other thing, like, as I said, for bathroom cleaning, there are, like, harsh chemicals that we really don't want to get to the robot, right? Am I, like, a beautiful paint on it? If you got, like, the bleaches on it, it will not look great. And so, when there are, like, I think when there are, you need to interface with very harsh chemicals and you need to apply a high amount of force, those are still, like, I think, remains very difficult for the mechanical side. Just had a vision of somebody walking in in my apartment, me saying, "Sorry, guys, my cloy just covered in shit. Tony, thanks for joining us, man. It's been great to reconnect after all this time and, yeah, super fun to see what y'all are building. I mean, it just, you know, there are so many problem statements in tech that either, I mean, there are so many reasons things can be weird, whether it is someone who just comes straight from business school versus a researcher, like yourself who's spent, you know, a decade on this stuff, or just attacking problems that just kind of don't really need to be solved quite yet. But, I think we could all agree, chores are annoying, so nice to have someone doing that on the show. For sure. Good to meet you, Tony. Yeah. Great to see you guys. Yeah. Enjoy it. That's it for this week's show. Thanks, Tony, for coming on the podcast and Alex for staying up late. I know you have very important people to be meeting with, so thanks for giving us your time. No one as important as you, Alice, always. Don't forget to like, subscribe, everywhere you get your podcasts where access.show, too. We have a website and you can find us in video at access pod on YouTube. You are so quickly fading, Alex, it's hilarious. It's obviously, it's obvious that you've been in like back-to-back meetings for like the last 16 hours. And if you like this episode, please leave a five-star review or thumbs up and share with a friend on the internet or in your local Swiss village. You can find me at hamburger on Twitter and at meaning dot company. You can find my newsletter at sources dot news. Access is part of the Vox Media Podcast Network. The show is produced by hooked creators. Buh-bye. Bye. To a state surrounded by nature. That's nice. Go on, book it. It's easy booking dot com booking dot yeah. Fall is the perfect time to refresh and reorganize your space. At the Home Depot, find power tools and tool sets starting at $50 to help tackle DIY projects, home updates, and more, whether you're drilling brackets to support new shelving or sharpening your hedge trim or blade with an angle grinder. The Home Depot has the tools you need to check projects off your list. Shop Labor Day savings at the Home Depot and gear up for fall projects with the right tools to keep your projects moving.