This Week in Startups - Pioneering physical AI with Archetype AI’s Ivan Poupyrev | E1951

Episode Date: May 18, 2024

This Week in Startups is brought to you by… .Tech Domains - Don’t miss our “Jam Session with JCal” contest, coming soon! To apply and get more details go to https://jamwithjcal.tech brought to you by .tech domains. OpenPhone. Create business phone numbers for you and your team that work through an app on your smartphone or desktop. TWiST listeners can get an extra 20% off any plan for your first 6 months at https://www.openphone.com/twist⁠ Vanta. Compliance and security shouldn't be a deal-breaker for startups to win new business. Vanta makes it easy for companies to get a SOC 2 report fast. TWiST listeners can get $1,000 off for a limited time at http://www.vanta.com/twist * Todays show: Ivan Poupyrev of Archetype AI joins Jason to demo Archetype AI’s product “Newton” and discuss the application of sensors in various industries (2:55). The two also dive into anticipatory interfaces (44:14), the role of sensor technology in AI (31:09), and the potential for robots in various industries (28:12). * Timestamps: (0:00) Ivan Poupyrev of Archetype AI joins Jason. (2:55) Demo of Archetype AI's product, Newton, which interprets and processes motion data (9:22) .Tech Domains - Apply for the Jam Session with JCal contest today at https://jamwithjcal.tech (10:48) Exploration of potential business model and use cases for Archetype AI (22:15) OpenPhone - Get 20% off your first six months at https://www.openphone.com/twist⁠ (23:30) Use cases in aviation, military, and automotive industry (28:12) The connection between robots and artificial intelligence (30:17) Vanta - Get $1000 off your SOC 2 at http://www.vanta.com/twist (31:09) The evolution of sensor technology and its role in AI (44:14) The pace of technological innovation and the future of anticipatory interfaces (49:00) Future plans for Archetype AI, hiring and partnership opportunities * Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp * Check out Archetype AI: https://www.archetypeai.io * Follow Ivan: X: https://x.com/ipoupyrev LinkedIn: https://www.linkedin.com/in/ivan-poupyrev * Follow Jason: X: https://twitter.com/Jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Thank you to our partners: (9:21) .Tech Domains - Apply for the Jam Session with JCal contest today at https://jamwithjcal.tech (22:17) OpenPhone - Get 20% off your first six months at https://www.openphone.com/twist⁠ (30:20) Vanta - Get $1000 off your SOC 2 at http://www.vanta.com/twist * Great 2023 interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland * Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow TWiST: Substack: https://twistartups.substack.com Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin Instagram: https://www.instagram.com/thisweekinstartups TikTok: https://www.tiktok.com/@thisweekinstartups * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

We also have conversations with people from printing press companies, and they're using, like, really old printing presses. Yes. And they're attaching sensors or printing presses because they want to kind of, like, you know, they're getting old and they need to, you know, adjust them to for the, for their reliability. And the same problem. Yeah.

I have all the sensors around this printing press. Can you tell me when things are about to go wrong, like predictive maintenance, yeah, recalibration? Because time is money for these guys, you know? Every minute machine doesn't work. That's money. This week in startups is brought to you by dot tech domains. Don't miss our jam session with JCal contest coming soon.

To apply and get more details, go to jam with jCal.tech. Brought to you by dot tech domains. Open phone. Create business phone numbers for you and your team that work through an app on your smartphone or desktop. Twist listeners can get an extra 20% off any point. plan for your first six months at openphone.com slash twist. And Vanta. Compliance and security shouldn't be a deal breaker for startups to win new business. Vanta makes it easy for companies to get a

sock to report fast. Twist listeners can get $1,000 off for a limited time at vanta.com slash twist. All right, everybody, welcome back to this week in startups. Obviously, we're moving into an era of startups, end employment and work and life. That is going to be driven by absolutely mind-blowing experiences powered by artificial intelligence. Videos by SORA, mid-journey, we've seen all that. We're starting to see robots like optimists and figure, music generation, all of this stuff is incredible. And it generally uses text-based prompts. But what if AI could understand the real world in real time. Well, archetype AI is here to bridge the physical world with AI.

And they call it just that, physical AI. They've created Newton. It's a first of its kind of AI model that understands the physical world. Okay. The innovation allows integration of sensors with machine learning. So you can have sensors, pull this stuff in here. And today, we're lucky enough to have the CEO and founder Ivan Puperev to explain what they're building and to show it to us.

If you're not watching us, you can go to YouTube.com. It's a new website that hosts videos. You're going to love it, by the way. There's a lot of videos up there, like thousands of them. And go to YouTube.com and search for This Week in Startups. She'll find the episode. Ivan, welcome to This Week in Startups. How are you doing?

Good. Great to see. Great to be on your show. Yeah. I've seen some demos of what you're building. Right. And it's really interesting. So why don't we get started?

And we'll just show the audience what you've felt. how it works. How do you like to start? Do you want me to show the demo? Do we just talk to a bit more about the company? Because I think it's like one of these things where once you see it, you start to understand it. You do such a good job of demoing it and explaining what's happening behind the scenes

and the other demos I got to see online. Right. Okay. All right. So, Ivan, show us how you take motion and you find some meaning in here. The idea of market API is to build a foundation model which can understand physical world. When you think about the physical world, you think about. sensing and sensor data, right?

Because human naturally observe physical world throughout and biological sensors. But when you go to, you know, machines and talking to the physical world environment and industries, they run on kind of all kinds of sensor data, motion data, radar, you know, spectrograms and so forth. So let's be sure the video, which is very much inspired by our conversations with logistics companies, like one of our investors is Amazon, and how they can track packages. through the long logistic supply chain and know what's happening

when the package moves because you kind of don't know like you send it somewhere you have no idea what's happening with the package. So how can we get sensors to tell you what's happening to the package?

So this is an example of the demo we built. So in this case, I'm going to pose here. In this case, you can see there is accelerometer in the box and you can see all the sensor data coming up

on the screen. So you can see. You have an accelerator box. which looks like a, you know, playing cards, a pack of playing cards. And you have an accelerometer in your phone, so you get the idea. And then you see like essentially a wave signal of some type, three waves, a purple, a green, and a yellow here. So you shake it, it's right. So you can see the box, and she's shaking and it moving.

And now she's putting the box inside of the package. And what you do is in our interface, you can ask Newton, pretty much tell me, tell me the truth. me the transit status of your package, how it moves through this thing. So, she put you inside of the package, and now the person that's up transit status and turns it on. And now what's going to happen with Newton is that as she moves a package around, the Newton translate this complex sensor data into the very understandable, you know, message.

The package is in motion. The package is still. So you don't need to go and understand what the sensor data means. but it's actually, you know, in simple language. And now you should change the prompt to reward package mishandling, so a dropping or shaking. And without changing the model, without reversing the model, without retraining the model,

the model can understand that sense of data needs to be tracked for the package dropped. And you can see now it's analyzing and see the package dropped. So this demo demonstrate that how you can, in real time, kind of steer the model to look after this particular, events or what they call behaviors in the physical world which demonstrate captured from sensor data, something you can naturally cannot understand. Especially one more demo because, and this is a difference between physical AI and, you know,

classic L. L.M. is not a chatbot. It's not something you're chatting. It's something you're asking a prompt, and then the model is looking for these behaviors, is trying to understand and report these behaviors to you based on what you ask it for. And the output doesn't necessarily have to be textual. Because if you see, imagine a worker at a factory or a doctor in a hospital or anybody who is working in physical environment, they have to be focused on the physical world.

They have to be focused on the task at hands. So the textual representation is not the most natural for that kind of environments, right? So the model has to also produce outputs in other formats, in visual. So let me show you. So this is a dash cam we're going to see now. So it's a dash cam recording what's happening in the world. It has some sensors, I don't know, LIDAR or just video, and then you're going to translate that into a language model, right?

Right. But in this particular case, what you see here. Or a visual model, I should say. It's a unique model that's not a language model. It's a visual model, yeah. Is it say model Newton? which can translate either in a text representation, but the same model can render that in a very different representation.

And later, during the show, I can show you a diagram which shows how that's happening and why that sort of translation is possible. But in this example, on the left side of the screen, you see the real video. On the right side, you can see the overlap, visual overlap, that Newton creates in response to the prompt, to a response to the question you ask. And let's be just show how it works, right? So this particular case, you ask a monitor for a car in front.

And you can see there's a car in front, and the Newton highlights where is the car in front. So obviously, when a cow crosses the road, you can see in this case, stop highlighting this. And when the car, the car passes by, it's going to continue highlighting the car in front. And the interesting thing here is that you can change the focus. You can say, stop doing the car and look for pedestrians. Show me how this pedestrian run. Now you can see on the right side, it's pedestrians who is being highlighted to give, you know, to direct your attention to them.

And for the same video. So you're steering your model to do things, which you need by text language. Here you're asking something to show me crowded areas, but the output right now is not the visual overlay, but on the right side you can see a hit map on the map based on GPS data which shows you crowded areas. So you can imagine a very simple use case where you have a fleet of vehicles, and that's actually a real use case we're discussing,

when you have a fleet of vehicles, fleet of cars, which drives around the town to deliver goods or products, you would like to report all other cars what's happening in the city so they don't get stuck somewhere because of the flooding or because of something, or some other events, right?

So dynamic update of the map based on the semantic understanding of the world delivered by Newton. That's one of the many use cases when it comes from understanding physical world. All right, you guys know I'm passionate about innovation and tech, and I love hearing from founders. I've got a crazy, exciting opportunity for you to consider. I'm hosting something called jam sessions with JCal. It's a contest.

It's powered by my friends over at DotTech Domains. And over the summer, I'm going to have five founders get the chance to do a jam session with me right here on this week in startups. It's really simple. You tell me in this one-on-one session what you're struggling with as a founder. What are your challenges? What's your vision for your startup? Tell me about your product.

Tell me out your customers. And we sit there and we jam out. I deeply listen to you. I ask you really deep, thoughtful questions. You give me deep, thoughtful answers. And we try to figure out how to grow your business. And then we publish it here on this week in startups. So everybody gets to learn.

It's really simple. There's only two rules here to get one of these five jam session slots. One, you got to have under $2 million in funding. So this is for new startups. And two, you just have to have a dot-tech domain name. So here's what you do. If you want to get more information, jam with j-cal.com. Jam with j-cal.com.

That's that dot-tech domain name you keep hearing about. Dot-tech domains and I are trying to find the most innovative founders. We have used the dot-tech domain for many things. And there's tons of people in the industry like rabbit. Dot tech, you know that a really slick AI hardware device, aurora.com, 1x.com. If you're using a dot-tech domain name for your startup's website, I want to hear about it. So apply for jam sessions with J-Cal. Jam with J-Cal. Tech.

Don't wait. There's five slots. You've got a good chance of getting one if you apply now. Okay. So this model that you're building, Newton, can take any sensor data. It could be LiDAR, it could be cameras,

it could be an accelerometer, put it into the model, and that let you ask questions to, I don't know, solve problems in the real world or understand the world better. That's exactly correct. Yes.

Okay. So is it an open source model or is it a closed model right now? Now, at this time, is a closed source model. We're not opening sourcing. Got it. And so you're building this model and then you're hoping to get a bunch of training data and then solve problems for businesses and then allow them to the business model here is obviously to make this a hosted services like an Amazon Web Services or something where people can give you sensor data and then query their sensor data and get some output. So how are you training this?

because you show different sensors, a camera sensor, and then you showed at accelerometer. I saw in another demo you gave, you showed somebody touching a doorknob, and then I think you have other models. So tell me what sensors, what inputs you currently have coming into it. So right now, at this point, we're focusing on four kinds of sensor types.

First of all, it's cameras, so with people using cameras, obviously, audio, time series data, and RF, which is pretty much radars, right? This is kind of sensors we were focusing right now. We're training our model for those sensors. The data is coming, the way of approach we're taking is that very early at the stage of the company, we did pretty broad review of the market. We went out to literally hundreds of companies.

We talked to them asking what sensors they're using and what kind of things they want to do with sensors. And that's how we selected this group of sensors initially. And then very early we started engaging, we built design partner program, and start engaging with us company to build specific, understand how our model can be solving their specific use cases. So the training data comes from, you know, either from partners that are giving us data to be able to train our model for specific use cases. And with every use cases, the model can learn more and more things.

Or for some of our partners, we're collecting data ourselves in the physical world, for specific where it's not. It would seem to me the number one use case here is self-driving and it was in your demo there. This is, I think, what Elon's gotten to when he shifted hard coding to a language model. So is what you're doing essentially a broader version of that that's available to anybody who wants to use it? That's exactly correct. We're building architecture of the Newton's architecture, the way we're designing Newton. It designed the way so we can take any kind of sensor within those categories with very small amount of modifications,

sometimes out of the box. that can use those sensors to solve their problems. So it's a very general purpose, universal model for everybody. Because whether you go to the physical world and the physical world businesses, you can't build bespoke solution for every single person or for every single business because it's so diverse and kind of messy the physical world in general. So it's universality of the model, which is extremely critical for being successful in this field. Do you believe what Elon's done with FSC?

and making this model and what you're pursuing will solve self-driving. And if so, when do you think self-driving will get solved? Because in your model here, you're asking it, hey, tell me where it's congested, tell me where there's a car, tell me where there's a car, tell me where there's a cow and obstruction, et cetera. So, you know, one of the core questions is, will we solve self-driving in all the edge cases by just watching humans drive and make mistakes and knowing it's a mistake or not? So knowing what you know, how close is Tesla to having perfect driving?

Yeah. I. Or better than human, because you must have used 12.4, 12.3, and you're building something similar. So just humanity in general, yourselves, Tesla, let's just brought it out because you obviously don't work there. Yeah, well, I don't work for Tesla. And we don't really focus on self-driving.

Self-driving is just one of the use cases. We're working with a few companies to help them with self-driving. But that's not one of one now. We actually build a building horizontal model across multiple, multiple modalities. We're working with a semiconductor company, working with a automotive companies, obviously, but also consume electronic companies and construction companies. Like this is sort of like we're trying to build a generic model. As always, it's very hard to predict with anything which happens in the future, like self-driving,

one's going to be solved. But I do believe that be able to understand contextual information beyond what's sensing, you know, from the direct sensitive, but understand the context information, behavior of the complex system, behavior with the people around it, and using large language models, style reasoning about the world around you would definitely bring full self-driving closer to solve all those complex use edge cases. So how much of what you're doing is predicated on having a large data set? You know, there are some people who have cameras in entire cities.

London, China are both known, you know, or different cities in China are known for having massive surveillance systems for safety, et cetera. And so they have a massive amount of data. If you had access to that, man, you would understand a large portion of the world. Then you have satellite data, maps, GPS data. And then I guess people walking around with sensors on them or bicycles riding around with sensors them, obviously a Tesla or the Waymo cars have massive sensor arrays. So what is your training data?

You know, people look at all the language models using open crawl or Reddit data or Twitter data. or Quora data or Stack Overflow, there's all these pools of data and oil. What are the ones you're tapping into to understand the world? Right. So, as I mentioned, we're working with design partners,

and depending on the use case, we're trying to solve for them. We're tapping in their data should they provide to us. We're also using, obviously, a lot of open source data out there, all these data sets which are available. We're using them to kind of cede our model

with the initial understanding of the world and kind of train the model on those. You know, like what we find out is, like, when you work, when you work with very specific customers, the specific customers has a very specific problem. You really have to work with those customers to get this data from them, and then you fine-tune the model on their specific use case.

But we also were quite surprised that in many cases, those customers are quite open to let the model to train on this data. They can keep the data later, But once the model is trained, everybody benefits. So the approach you need to do the piece by piece. You're solving for one customer and that empowers everybody else. Got it. And so factories are places where there's a lot at stake.

There's a lot moving around. They're complex environments. There's robots as humans. So getting into factories and just understanding what's happening in a factory, is that one of the early use cases here? And then do you need to make more sensors for those factories? Or they already have the sensors in there, don't they? That's exactly correct.

You know, like, one of our investors is Hitachi, right? And the Hitachi actually were interested in our company exactly for that reason, because they're already storing massive amount of data from the sensors, right? You know, when we talk to them, they say, like, look, we have all this data, but understanding this data and figuring out how this multiple data streams can be analyzed to understand not what particular sensor does, but how, all together they can draw holistic image of the factory, that's what we're looking for. And the use cases that's just like endless there, completely infinite amount of the edge

case. Well, let's double click on that. You know, you have a factory building, I don't know, robotic arms, right? There are factories that have robotic arms, building robotic arms, quite meta. Well, let's say you have a factory that builds robotics. And you, you know, get all the input for the last five years, if everything that's occurred in that factory, then what would they ask and what would the benefit be once they have all that data in the language model? Because we showed very basic proof of concept demos here, but in the real

world, what do you think they would then start asking it? What could they ask their factory that built, I don't know, cars or televisions? I can give it a real use case and the real example with the real customer we have. So we having conversations right now with a very kind of live. semiconductor company, right? So have these machines which are making chips. You know, literally they're saying like, look, we have something like a plasma reactor for etching the silicon wafers, right? And some of these machines have up to 400 sensors inside of the machine.

And 200 of them are critical, which means if they're off, you know, the value, which is supposed to be, it cannot work. And what they're saying is a problem there is that you take. Take this machine and literally you move it or you shift it or change it by a meter. And because the precision and tolerances are so high, everything goes out of whack right away. It means they all start getting false alarms. The machine stops, yields drops, and then somebody has to come and reset the entire machine, which can take days, and a lot of money lost, right?

Which is obviously being passed to end customer and eventually to us. So the question was like, can your model not just look for the threshold values of the data, but actually understand which data is correct, which data is not correct. So when you move it and it's actually moved around, the model self-adjust to itself. So this is one of the very specific use case where the factory is looking for solutions. Wow. Yeah. I mean, that's incredible when you think about it, these highly precise machines,

if monitored, they have monitors and sensors already. Of course. This is, now the language model is watching it, it could tell you what to fix. It could maybe even fix it in real time. I don't know if this, it can actually adjust those 400, you know, nuanced, or it can actually do the adjustments with the machine itself.

I don't know how the machine is configured, but at least being able to monitor, it's going to save a lot of time and money. I was thinking of like a printing press to go old school, not that we've print much, but you watch those newspaper presses or magazine presses, it was a very similar situation if they were off just a little bit because you see how fast they moved.

Exactly. The whole thing is just throwing away a lot of off-printed newspapers. Yeah, exactly. A lot of loss, right? It's surprising, by the way, and finally you mentioned printing press. We also have conversations with people

from printing press companies and they're using like really old printing presses. Yes. And they're attaching censors or printing presses because they want to kind of like, you know, they're getting old and they need to, you know, adjust them for their reliability

and the same problem. I have all the sensors around the printing press. Can you tell me when things are about to go wrong, like predictive maintenance, good collaboration? Because time is money for these guys. Every minute machine doesn't work. That's money.

Juggling multiple devices and apps to run your business is a mess. Open phone is here to make it simple. By simplifying your business communications with one easy-to-use app. Open phone has rethought every detail of what a modern business phone should be. And here's the magic. It works through a beautiful, elegant app on your phone, or you can just use it on your desktop,

making it super easy to get a business phone number for your entire team. And you know how brilliant open phone is? My teams use it every single day. My sales team loves it, my ops team. They use it all day long. And here's the features that we love. You can create a shared phone number like customer support with multiple employees fielding all the calls and all the text to that one number.

At my investment firm launch, we pride ourselves on replying to every single call or email instantly. And Open Phone is the number one rated business phone on G2 for customer satisfaction. So here's your call to action, super easy. Open Phone is already affordable. Starts at just 13 bucks a month. But Twist listeners get an extra 20% off any plan for the first six months at openphone.com slash twist. And if you have existing numbers with other services, no problem. Open phone's going to port them over easy, peasy, lemon squeezy, no extra cost. Head over to openphone.com slash twist to start your free trial and get 20% off. We talked about one type of vehicle cars.

I wonder battleships and airplanes also pretty complex and with unlimited sensors. My God, the sensors in an airplane or a battleship, I mean, incredible. Has the military and space, you know, started to come out and say, hey, let's just take a look at all the sensors we have. I could imagine a SpaceX rocket or a giant airplane, a complex Boeing airplane, having all these sensors reporting in and then being able to ask questions, I wonder if you could avoid accidents or maybe come up with insights on how to make those products have less drag or, you know, be more efficient in some way. We haven't yet yet from anybody from aerospace. We have this conversation from aerospace industry or anybody else from that. But, of course, we open these conversations and love to talk to them. I mean, if you think about automotive industries as a kind of a proxy for this complex machinery, so there's a lot of interest from automotive industry because, you know, the car generates

some just gigantic amount of data. I've just read like maybe yesterday an article that like once AI cars, AI said bring to the cars, even not full self-driving, just like AI, you know, cooperated cars. It's like 25 gigabyte data per hour going to be generated by the car. So how you even process all that amount of data and how you can make sense of that, how humans can understand that data. So you need a sort of something in between which can help you to analyze this data. And that's what Newton is.

Newton is looking at the physical world, all this data and get you, help you to make sense of that world of physical data. That's kind of forward. I was just wondering about environmental stuff. You've got the obvious factories that are pack with centers. But then we have the real world, and we're very concerned about the rainforest. We're concerned about oceans and temperatures and pressures and, you know, the amount of sunlight and et cetera, precipitation. And those sensors have also been deployed in many cases.

And those systems, you know, are, I think, incredibly complex. Weather systems come to mind, you know, global warming and CO2 and all, all of of those, have you started to think about how we might be able to use all the global sensors on the planet to maybe understand what's happening to the ecology of the planet? Yeah. I mean, obviously, like, if you, if you be, yeah, of course, certainly. Certainly the decarbonization and supporting kind of environmental, you know, environmental monitoring. It's one of the most interesting directions we can take to. I give you another example, which we discussed quite extensively with one of the

partners in the process of conversation. You know, the gigantic windmills, you know, those things, rotates offshore. So, they have a very specific problem,

is that vibration of the gearbox is a prediction of failure. And you have basically this windmill farm. of dozens of those windmills. And they all vibrate slightly differently.

To try to understand, is this normal vibration? Is it the ground vibration? Is it the wind vibration? It's making sense of this vibration would allow them to do either predictive maintenance or slightly adjust operation of those windmills to optimize their performance. So this is exactly this is one of those problems which relates to what you're mentioning,

how to control this gigantic infrastructures which we are building in the physical world and how to use the sensor data to actually predict the future of what's going to happen with those machines. And how do you think about the connection between robots and artificial intelligence? Obviously, we've got figure and optimists

and a bunch of people are starting to look at this and there's lots of sensors in these robots, Boston Dynamics, obviously has been doing this for a while. So are those going to eventually be out there in the world mapping the entire planet Earth to give us some more information than we currently have? Right?

Because what could be unlocked if you had perfect insights into everything occurring in a city? Everything. And we have seen this with perfect mapping, right? With GPS has had a profound impact. We don't get lost. As a species, pretty hard to get. lost these days. Pretty hard to be out of communication with satellites and, you know, SMS to

mobile phones now, et cetera. So, so if we could with these robots, you know, if there were a billion robots on the planet and you had all the sensor data, how does life change for humanity in your mind? And maybe you talk just generally about robotics and the impact here. Robotics is very interesting. Again, so just like with the satellites and space companies, is we haven't yet engaged in robotics companies. We're mostly focusing on our current focus is construction and automotive and factories and semiconductors and semiconductors particularly, like all that stuff, help them to solve the world. That's our initial set of customers.

But we're talking to a lot of people, obviously. I think the speculation is always dangerous, what's going to happen in the future. But what we see from the data coming in, one of the most important things people are asking for is some form of prediction and optimization. In the way, if you can, because the best way to predict the future is actually to understand the past, right? So if you have a certain amount of data captured about behavior of the factory,

a behavior of the building, behavior of the ecosystem of, such as a city, right? There is a particular to predict. If you can use all those data in a long term, just like with large language models, by using all this data, it is possible to predict potentially what's going to happen tomorrow, the day after tomorrow, a few days after the days after. And by doing this, you can optimize your energy consumption.

It can optimize your infrastructure control. You can optimize how people to live better lives, just be able to predict what's going to happen. Just like we predict the weather, you know, we should be able to give some sort of prediction, what's going to happen with your factory, what's happened to the city, like where, what happens to traffic, what happens to things. So start looking into the future with all this data and making better decisions now to either avoid unnecessary outcomes or prepare for them better. That I think would be one opportunity we can see we can see here. Listen, a strong sales team can make all the difference for a B2 startup.

But if you're going to hire sharks, you need to let them hunt and you can't slow them down with compliance hurdles like SOC2. What is SOC2? Well, any company that stores customer data in the cloud needs to be SOC2. compliant. If you don't have your sock too tight, your sales team can't close major deals. It's that simple. But thankfully, Vanta makes it really easy to get and renew your SOC2 compliance. On average, Vanta customers are compliant in just two to four weeks. Without Vanta, it takes three to five months. Vanta can save you hundreds of hours of work and up to 85% on compliance. And Vanta does more than just SOC2. They also automate up to 90% compliance for GDPR, HIPAA, and more. So here's your call to action. Stop slowing your sales team down and use Vanta. Get $1,000 off at a $1,000 off at Vanta.com slash twist. That's vanta.com slash twist for $1,000 off your sock, too. There are seismic networks and seismographs or whatever those are that are predicting earthquakes, right? And I don't think we're particularly good at it. It sounds like we could be a lot better

at it with more sensors and more language models, also tsunamis for NATO. So has that come up yet and have you studied those areas? We haven't yet, no. We try to keep, of course, obviously our aperture as broad as possible, look as many things. But the same thing. But the same thing, like we kind of don't want to boil the ocean. Right. So construction seems like a really good place to do it because there's a lot of construction. Yeah, construction was very interesting.

And actually we actually very actively engaged with one of the, I would say, largest construction company. They're based in Japan. And they building this massive projects, you know, like, you know, terraforming style, moving the mountains and changing direction of the rivers. The project which takes years, right? And the problem they have is that they would like to opt because the amount of kind of like resources span to build those projects is just humongous, right?

How can we optimize this process, you know, going forward with future projects, even to understand what's kind of like how the process was working right now, looking at the data like four years of the data and asking questions like, well, when this construction period started, what's happened then, how what was the throughput for that style, how many people were engaged in this part of work, just asking those questions is allow us to probably dramatically to reduce waste, increase speed of building construction projects, and reduce the cost of them. So that is one of the actually active engagements we are right now pursuing. Just take a look at the data they have, you know, and helping them to figure out what's actually happened during this construction period.

Yeah, there are some giant construction projects going on, obviously in the Middle East in Dubai, in Saudi, Neom. and then you have, let alone, some of these water projects, you know, whether it's moving water, China's got a giant project to move water from, I don't know if it's from the north of the south or the south to the north. I can't remember, but there are some major, major, you know, multi-decade projects and also you have things like Venice or Seawall projects in Amsterdam. And these are, you know, these are tens of billions of dollars, some of these projects. They're right. That's exactly correct. I think what's happening is that our ability to kind of build things to dramatically improve, right? So we can build those gigantic, you know, constructions.

We can build very precise chips and very precise technology at nanometer and like, you know, two, three nanometers, you know, parts. As the project is becoming more and more complex, they generate you more and more data. And it's not going to decrease. Like, we're going to have more and more data coming in. The avalanche of data is not stopping. And we can't control those projects. So this is going to control those technology without having very precise sensing and very precise understanding.

So this sort of like we kind of discussion with you too, right? So you need more data to control those things, but you can't analyze those data. So that's what we try to help in. We're trying to help all those industries to understand. You and I are of a certain age watching the last 30 years of development of sensor technology, which was absolutely catalyzed by creating billions of smartphones. The prices went down to nothing. Then you had storage and fiber, and the storage costs have gotten down to, you know, very much commoditized.

You have bandwidth very much commoditized. And it was just waiting for a technology to help us sort of analyze this at scale, yeah? And AI is that technology. I completely agree. I think when we started the company, we were discussing that, like, you know, there's several building blocks for the Newton to happen, for the archetype to happen. we need a few building block. We need, first, of course, is a cheap sensing technology and industry to be ready to use sensing technology.

It has to be a sort of amount of penetration of the sensing across the different industries. And that happened. There was a whole industry 4.0 movement, the IoT wave, which happened. It wasn't really successful, but it put sensors everywhere. Why didn't IoT, like, there was supposed to be this giant IoT, IoT of everything, and it kind of didn't happen, except in your smartphone, maybe in cameras, but why did that have a false start, do you think?

What was missing? Because of the problem for analyzing the data. You have a silo-silo data in a certain device. That makes total sense, yeah. The device by itself produces some so small amount of data, and the value from that just one device is not very high. Because, okay, on-off kind of signals, how much value they're going to give you. So it's like, okay, whatever. You know, like, I know my fridge is on, my fridge is off.

Why do I care, right? It's when you start connecting different types of data together. And then you try to place in the context of larger human life and kind of attach all this meaning of this data, which has became possible with this foundational large language model approach and transformers and do this deep prediction. that's when suddenly that's becoming possible

what something dream of IoT was in the future, right? So I think this is census one of a component, bandwidth another component, storage another component, and of course, AI, fundamentally this kind of this

transform space model which allows you kind of like use a huge amount of data to predict, you know, and understand this future. This is all components came together and this is sort of like vision of that.

is now it's becoming possible and super excited for that reason. It's almost as if AI was the keystone in this arch, you know? Like all the bricks got built up and was like, boom, we should put AI in here. And we're seeing it inside the human body. All these sensors, people have continuous glucose monitors,

heart rate monitors, pulse, oxygen levels, steps. And then people are getting pre-novo, full body scans, blood work. And nobody's put all of those together. That's the body. The body, have you considered, did that come up when you were doing your startup of like, hey, maybe we should just work on the human body and somebody should just take all that big data, all those sensors and put them into some language model of the human body? We will definitely, of course, we did. Actually, one of our advisors is a chief technology officer of the Arthropetics Department of the USCSF, right?

If you need to have a needy place, he's the guy to go to. So we actually discussed very deeply with him. He has this kind of this idea, the whole direction, pretty big direction, is motion as a new vital sign, which is very interesting. You're saying that if you understand how people move through space, you understand how healthy they are. Because the goal of the healthcare is to get you moving. Nobody is getting better because just to lay down on the bed and not do anything, right? The goal is to have active life. So by measuring motion, we can measure the success of the healthcare.

So he was one of the first kind of our advisors in the company, and we deeply looked at the health space. The health space is tricky, though, right? So there's a lot of regulatory, you know. Yeah, of course. Yeah. I mean, you can't. It's totally different than somebody's house. you know, we're starting to see houses and buildings also have this technology where,

right, as but one example, we, I now have in, you know, my house and my ski house, humidity sensors, water sensors, temperature gauges that are all remote. Obviously, we have cameras around the houses, inside the houses, etc. And, you know, when something happens, like there's a flood or water, we're now getting a handle on that quicker earlier and then, you know, avoiding damage, right? And that's just the tiniest of, and, you know, maybe one of the most common ones, but boy, it's going to get interesting over time, right? I think the nest is also doing some interesting things in terms of turning down the temperature or your air conditioning when the grid gets too high. So you have two different systems

they're interacting. It's really going to be a brave new world. Right. No, that's exactly. And again, it's like one, one of the interesting use cases we have, you know, like one, and like a lot of the stuff, of archetype was informed by our work at Google. Just to tell me a little bit about the team, it was, we all worked at Google kind of like of building models for sensor data, right? We kind of understanding how to use sensitive,

how extract meaning from sensor data and actually put to value, right? And one of the use cases, one of the things we built, we build this radar, solely radar, which is a project, which actually I talked about it at one of your events a while ago. Yeah. And at launch.

Yeah. And so we launched a very first sensor and built, it kind of invented the whole, the first sensor, which was consumer grade radar, tiny radar which you can put on the phone or can put in air conditioning. You can put into the, into the, I remember, yeah.

The pixel had this, right, to do the deck sensing. That's right. That's what we did at Google. And at that time, we were kind of like first time look at the radar sensor data. And we are realizing that human cannot extract information of the sensitive. It's impossible. It is too complex. Sensor signals is too complex. There was the first time we applied deep neural network to a very complex

sensor data which humans cannot understand. And it was very successfully to the point that, you know, our last product at Google was shipping a the sleep monitor, right, which can measure how well you sleep using radar. And that's in that Google home device that sits on your side table. That's right. And it watches you by radar and knows if you're moving around and gives you. It's so funny. You mentioned it.

I have one of those Google things a couple of feet away from me in my office, which I used to watch my nest cameras. And I was in the settings page and it had turned it on and off. Exactly. It's not, it's in my office, not next to my bed. But what an incredible concept is that the radar is watching that, right? And it's that sensitive enough to monitor humans in a bed. Yes. In the moment, also, you're breathing, your heartbeat.

It's extremely sensitive. It's extremely, and it's privacy secure, right? It doesn't have it. It's not a camera. It doesn't see you. It doesn't see your motion and understand how you, you know, act and then kind of like... I think also you were using, I think Nest cameras were also using this a little bit, or there are some... Not yet.

Not yet. Not yet. I know that there was talk of using this for sudden infant death syndrome, SIDS and watching babies because when you have a baby. if you're a dad, you know, like you put the camera in there and once you put a camera in your baby's room,

now you're being super vigilant and all of a sudden your anxiety goes, are they breathing or not? Did they stop breathing? I mean, there's more good to talk about babies dying,

but sadly, sometimes babies will stop breathing and they roll over a certain way and they could suffocate. It happens in all, every species. And these cameras could actually know

when that's happening and put an alarm out, I guess. That's true. Yeah, so there's a, There's a couple of companies where we use the radar for observing babies. There's people who put the radars into the regular cameras for the power consumption.

So if nothing happens, it's radar's looking around. And when something coming in, then the camera turns on. So you can extend the power life. And this improves also false recognition and false alarms. It's one of the particular use cases. But what I want to say is that, so this is the first time when Andrews, the data from the radar. That's how you can use deep learning to understand this really complex sense of data.

That was one of the inspiration for the company. Yeah. The company that was doing this is called Owlet Duo or Outlet Dream. And it is specifically using, I believe, radar and sleep to watch your baby and just maintain the environment. It's fascinating. It's great. Super interesting.

And I think it also has like a sleeve you can put on the the foot. Yeah, it does. So it uses, talking about combinations of sensors, you could put a sensor on the baby itself, and now AI is going to be able to tell you what's going on with your baby, if your baby's lethargic, or maybe it's got an upset stomach, maybe the formula using is disrupting its sleep or something. Like, this is incredible what we're on the precipice up. Exactly. What gets you excited? You know, you're deep in this, and you've been deep in it for a long time. I do remember you were at our lunch mobile event and this is way back in the day

when the pixel 3 or it was a very early pixel that you guys had this sensor stuff. Pixel 4. Yeah, it was very early. And so what gets you excited now when you're watching this progress and if you were to talk about the pace of change that's occurring? You've been a technologist for three decades, I believe. Watching this last three decades,

talk to the audience just generally about the pace of innovation and what, What makes you excited today? What excites me most is combination of the sensor data and artificial intelligence, right? I think that's fascinating. And that's, you know, like I used to work at Google. I used to work at Disney with building the census for the parks and resorts. Oh, yeah, right.

Before that, I was at Sony and we built the very first kind of mobile devices and again did you work on the Magic Link project or what was it called general magic and that stuff? There's a magic link project. At Disney, no, I wasn't involved in that, but I know really well of that project. No, I wasn't part of that. But we build the things like, you know, Avatar land, you know, we build a sensitive system there for the Avatar land and, you know, with all the rides and magic fountains. And you name it, right?

All kinds of sense and technology. So much fun. Yeah. It's a lot of fun. And it's, you know, like, when you work in Disney, you're realizing that the most important thing is a narrative, right? So it's all right the narrative. And narrative and distance

around the magic. This is a magic of technology, magic of things happening before you anticipate, like before they happen before anticipate you. These things which have, guess what you want.

They understand your sort of like, what you like to happen and they're happening for you. So that's what kind of like people's really surprised and excited and happy. That's what people makes happy, right?

Yeah. Our dreams come true, right? And I feel that this combination of sort of sensing and prediction can, you know, that can anticipate what people want, can solve our problems before we even see them, support us before we're asking that. This is kind of anticipatory interfaces and anticipatory use cases. That's on the personal level, I'm still kind of like, imagineer.

And that that makes me super excited, you know, like, that's kind of nice. Well, I mean, just looking at your face and understanding the mood you're in and, you know, people moving through a city like, wow, everybody is really depressed. Everybody's really anxious. Like, what do we do here? You're going to have, like, this incredible pulse on the world that we just didn't have insights into. And that's what Disney does. You go to Disneyland. That's why they say it's the most magical place in the world because they're anticipating, you know, your experience and then delighting you with laughter.

surprise, thrills, whatever it is. Yeah. And those are all going to be customized, right? A certain person might go on a ride and you could actually sense that they want more thrilling or they want more, I don't know, storytelling or more fun. You could actually adapt the ride to their particular age or desire. Right, right? Extreme personalization, right?

Extreme personalizations of everything. Like, that's, you know, because we're living, we live in the period of, mass production, right? Everything is mass producer. Things are cheap and we can buy them at the price. And they're really, really high quality. It's amazing. That product we can buy right now is amazing.

But they designed one product fits everybody else. Yes. So can we go back to the, you know, when you also have a personalization where every product is separate and works for you? And that's something, you know, what you want. I mean, science fiction is just, you got to work at Disney. So you got to see a little bit of this and a little bit of Pixar, a little bit of Star Wars. but, you know, the Minority Report film,

Minority Report was just so, so many little items in there because they did go to MIT and they, a bunch of futurists and technologists contributed to, I think that was Spielberg, who did it, yeah, and they contributed to, you know, the different interfaces. In fact, I think the gloves were an MIT

specific project that they just extrapolate on, but in that film, people are walking around, and when they look at a billboard, it tailors it to that person. So Ivan would get one ad, I would get another. You might like Chinese food. I might like Japanese.

It's going to direct us in the mall to our preference. It was, and here we are. You know, ads on the internet are as customized as they could possibly be. In a way that people think it's like listening to our voices, you know, and listening to our microphones, even though it's, in most cases not. This is amazing. Is it, is the API available for hackers to start hacking on yet? Have you, have you made a public or API or developer? it yet? Not yet, not yet, but we are planning to do this. We, you know, we're a process

of building the core technology first. Right now we're focusing on a few, as I mentioned before, we have a design partner programs and design partners programs open and we kind of inviting companies to join design a partner program. Come to us with your problems and see if they're there for us and we're building technology with them. I think once we have a few pilot cases built and demonstrated to the public and showing the value at the same time preserving

generality of the platform we would love, of course, to open to broader audience and let everybody try with their own sense of data whether it's from mobile phone or from kind of IoT devices they have some of you hacks

in your kitchen, it all connect to the model and try it out yourself. That's coming. Got it. And you can learn more at archetype. Are you IO?

Arctype. AI. It's archetype. Yeah, archetypeaI.io. ArchetypeaI.O. So you can understand the real world. If you're looking to do a partnership, it's interesting to you, go over there.

And I know you're hiring, so go to the website and go to the careers page as well. And continued success with this is kind of mind-blowing. It's very early, but I wanted to have you on early because I know next year, everybody's going to be talking about what you did. And I wanted to put this moment in time in 2024 here because in 2025, 2026,

this is going to get really interesting. So I hope you'll come back next year. and tell the audience about, you know, all these incredible use cases that you're kind of stealthily working on and share more updates. Great seeing you again. And we'll see you all next time

on this week in startups. Bye-bye. Thank you.