The Robotics Stack
00:03:15 · Varun Nair
[00:03:15] S2: I think there is actually a very common thread here, and you'll see this with a few different folks. A lot of people that were in crypto are finding their way into robotics, whether it's data or at a different level. And I think it's something in the psyche of people who are chasing the future, and SF is a great city for that. But in general, the goal is that, hey, crypto made a promise that the financial system is going to be changed, be decentralized. The promise was that there's a better tomorrow out there. And I think it's similar with robotics that this techno optimism sort of tech progress that can be created with newer technologies is inherent. So it's the same reason I enjoyed learning about and building in crypto. And it's the same reason I want to do robotics, and it's the same reason why I'm excited about space exploration in the next twenty years. So it's the same muscle of techno futurism, techno optimism around how these technologies are going to help us go forward as civilizations. That being said, more grounded answer around Coinbase in specific, I learned a lot there. I was actually doing within security org, I was doing a lot of data engineering, which gave me the building blocks of understanding how like data flows at like an enterprise company. How does that even look like? What the technologies are? And it was right out of college, so it was pretty early on and I learned a lot that ended up becoming very transferable and applicable in the future when I was working with robotics data.
00:04:39 · Oskar, Varun Nair
[00:04:39] S0: Yeah. Crypto is a very universal technology because it's finance, of course. Do you already see some potential use cases where robotics might be even connected to crypto? I think we have seen PrismaX. I'm not sure if you know them, but I think they dabbled in crypto for a bit, tried to create something. We've seen it in AI. You know, sometimes it's a bit of a let's just merge crypto. For the sake of it, do you actually decent potential there? [00:05:05] S2: I think it's a bit forced right now. I do think there is potential there, but to be very honest, I think current implementations are either forced or too nascent. There are examples of how you can use like decentralized systems for verification, like of work, which can be used for, let's say, distributing and getting data. There are a couple of companies I know that could be using DataQC through a certain kind of blockchain like system. That being said, it's not needed. It's novel, it's cool, it's exciting to my student like mind, but I wouldn't say it's like something that economically makes sense, and someone can make sense of it, but I haven't seen something like that actually pop off so far. [00:05:44] S0: Yeah. Neura Robotics, for example, from Germany, they raised with Tether. Somewhere I read about it that they're kind of planning to put a wallet essentially on each robot, but as you say, you know, it's very far into the future that I see a robot working or going somewhere and paying with their undevised wallet or something like that.
00:05:44 · Oskar, Varun Nair, Leo
[00:05:44] S0: Yeah. Neura Robotics, for example, from Germany, they raised with Tether. Somewhere I read about it that they're kind of planning to put a wallet essentially on each robot, but as you say, you know, it's very far into the future that I see a robot working or going somewhere and paying with their undevised wallet or something like that. [00:06:02] S2: I definitely see like there are potential use cases, and it can happen, it's just that you're in the robotics field, you're like, okay. Can we get, like, more deployments actually, like, out there? And that's a p zero question rather than, like, can we get wallets on the robots deployed? Because that's, a p two question. Yeah. [00:06:18] S1: Yeah. One question I have as a follow-up from your background and your experience is how do you define physical AI and where do you draw the lines between classic robotics, foundation models? Is physical AI the same thing as embodied AI? How do you define the space?
00:06:35 · Varun Nair
[00:06:35] S2: I think they are different. AI so far has been on our screens. It's been in the world of bits. And physical AI is all it's saying is that, hey, can I make intelligent decisions in the real world? I think Waymo, Zoox, Robotaxi are fantastic examples of physical AI, and they're even embodied in a sense, right, they're a core. But I wouldn't call them robotics, even though they're similar. I think it's important to make the distinction that there is robotics, which is a big part of the narrative of physical AI today, but physical AI in general is just any intelligent decisions being made in the real world. We have solved rigid body dynamics through simulations, so that's why things like Waymo's can exist, quadrupeds like Spot can do navigation and locomotion very well because again, bodies have been sort of figured out. But dextrous manipulation is still on the frontier of things because in robotics, that stuff is not fully solved yet. But if you've heard of companies like Spectre, they are not in the robotics field at all. They are an intelligent system for facilities. It plugs into your cameras, your other sensors, everything in like, let's say a factory or a warehouse or a building. It's like Palantir for the building. Like it understands everything. Can say like if there is a robber outside, there's a car that's parked that shouldn't be there, it can like intelligently pick up on those things. Assume a building has just come alive and it is sentient and intelligent because of all the sensor data it has. I think that's physical AI. Like in the Venn diagram, it's clearly physical AI, it's clearly outside of robotics. And that's also a big problem that's getting solved. So largely speaking, physical AI is the entire domain where the world of atoms makes smart decisions, but robotics in specific is the frontier because it's still unsolved for in a lot of ways. And that would be embodied AI, brilliant companies doing some cutting edge work around that. And I would break that down into data, simulation, deployments and foundation models, obviously, that I like to learn from these, broadly speaking. Yeah.
00:08:31 · Leo, Varun Nair
[00:08:31] S1: I see. That also bridges the gap with something interesting you said earlier. You've been into data, you've been into deployment, and lately you've been interested in SIM simulation. Usually, people are into simulation before doing deployment. So how did you, like, get into real world stuff before getting into into SIM? [00:08:50] S2: I mean, you could go either way. The thing is you do deployments or you just, like, evaluate in the real world and you see that policies don't work very well and you're like, hey, there might be a better way to scale this up. There might be better ways to run a thousand different trajectories rather than resetting one manually again and again. And that's how I found my way. It's not like I hadn't dabbled in SIEM before, I have, but it's interesting because you have data on one side, you have deployment on one side and you have SIEM sitting in this middle where learning in SIEM is smart, but there is a real gap that exists between transferring a policy that exists in SIEM versus in actual deployments. But it's the other way too. When something fails in deployments, it needs to be rerouted back to SIEM as well to make the simulation even better. And that's a piece that's missing. So it's not just that SIEM informs deployment, but it's also that deployments will inform how to make better SIEMs. I'm actually putting something out there about this later today or tomorrow. So hopefully it'll give you better context about it as well. Yeah.
00:08:50 · Varun Nair, Leo
[00:08:50] S2: I mean, you could go either way. The thing is you do deployments or you just, like, evaluate in the real world and you see that policies don't work very well and you're like, hey, there might be a better way to scale this up. There might be better ways to run a thousand different trajectories rather than resetting one manually again and again. And that's how I found my way. It's not like I hadn't dabbled in SIEM before, I have, but it's interesting because you have data on one side, you have deployment on one side and you have SIEM sitting in this middle where learning in SIEM is smart, but there is a real gap that exists between transferring a policy that exists in SIEM versus in actual deployments. But it's the other way too. When something fails in deployments, it needs to be rerouted back to SIEM as well to make the simulation even better. And that's a piece that's missing. So it's not just that SIEM informs deployment, but it's also that deployments will inform how to make better SIEMs. I'm actually putting something out there about this later today or tomorrow. So hopefully it'll give you better context about it as well. Yeah. [00:09:48] S1: Yeah. We recently discussed with an engineer saying that simulation was still struggling to accurately represent actuators, for example, since robots are pretty much a bunch of actuators. It tells you a lot about how accurate sim can be. Right?