WEBVTT 1 00:00:03.986 --> 00:00:04.647 Hey, Bankless Nation. 2 00:00:04.687 --> 00:00:06.729 In this episode, I talked to Shaheen Farsi. 3 00:00:06.809 --> 00:00:11.754 He is a general partner at Lux Capital, which is a hard tech venture firm. 4 00:00:12.415 --> 00:00:17.300 They invest in frontier technologies, AI, automation, biotech, robotics, just to name a few. 5 00:00:17.900 --> 00:00:18.481 Not in crypto. 6 00:00:18.501 --> 00:00:20.103 This is not a crypto episode. 7 00:00:20.523 --> 00:00:23.546 I wanted to go learn a little bit more about robotics. 8 00:00:23.566 --> 00:00:24.087 Robotics. 9 00:00:24.407 --> 00:00:29.230 And Shaheen is a veteran of automation and robotics. 10 00:00:29.290 --> 00:00:30.770 Automation is a very important word. 11 00:00:31.050 --> 00:00:36.213 There's a lot of hype and excitement about humanoid robotics. 12 00:00:36.233 --> 00:00:39.495 You know, the Tesla Optimus Prime, the figure robot. 13 00:00:39.975 --> 00:00:47.479 And Andrew Kang really came into the crypto industry really talking about like robotics and humanoid robotics. 14 00:00:48.119 --> 00:00:49.680 And I want to learn about that. 15 00:00:49.700 --> 00:00:52.942 I want to learn about robotics and the investing space around there. 16 00:00:53.582 --> 00:00:56.503 Jaheen has been investing in robotics for a decade plus. 17 00:00:56.903 --> 00:01:02.244 And so he's seen a thing or two and he's been around a few hype cycles in robotics. 18 00:01:02.284 --> 00:01:06.605 And so he's just probably the foremost expert on this industry. 19 00:01:06.625 --> 00:01:19.527 And so I'm pretty honored to be able to host him on this conversation and just learn about a sector outside of crypto that's in frontier technology that I think is going to be very, very impactful from somebody who just knows all the ins or outs. 20 00:01:19.687 --> 00:01:22.528 And so with that preamble out of the way, Shaheen, welcome to Bankless. 21 00:01:22.709 --> 00:01:23.609 Thank you for having me, David. 22 00:01:23.969 --> 00:01:27.051 Shaheen, there has been a ton of hype around robotics. 23 00:01:27.731 --> 00:01:33.754 And I think a lot of people are learning about the robotics sector for the very first time. 24 00:01:33.774 --> 00:01:43.319 And so this is why I reached out to you and Lux, because I want to get a veteran on the show to ask a veteran what they think about the robotics industry. 25 00:01:43.339 --> 00:01:46.081 There's just been a crescendoing of hype lately. 26 00:01:46.901 --> 00:01:53.545 And while hype is exciting and it's fun and it's an opportunity to learn, it can also be dangerous. 27 00:01:53.605 --> 00:01:59.228 And so I think the first question I want to throw at you is, is the hype around robotics justified? 28 00:01:59.448 --> 00:01:59.989 What do you think? 29 00:02:00.269 --> 00:02:02.090 Yes, it's absolutely justified. 30 00:02:02.610 --> 00:02:02.850 Okay. 31 00:02:03.511 --> 00:02:03.991 Okay. 32 00:02:04.031 --> 00:02:05.032 That's a pretty simple answer. 33 00:02:05.893 --> 00:02:12.599 So the hype around robotics, I think, comes downstream of like a lot of products or companies coming live with humanoid robots. 34 00:02:13.219 --> 00:02:20.406 Is that what gets you excited about robotics, the humanoid element, the figure and Tesla Optimus Prime robots? 35 00:02:20.466 --> 00:02:21.026 Is that the same? 36 00:02:21.046 --> 00:02:22.548 Because that's my sector. 37 00:02:22.668 --> 00:02:25.470 That's what are getting people excited in my neck of the woods. 38 00:02:25.510 --> 00:02:26.171 Is that the same for you? 39 00:02:26.531 --> 00:02:26.771 No. 40 00:02:26.851 --> 00:02:29.592 So I'm not particularly excited about the humanoids. 41 00:02:30.473 --> 00:02:45.340 I'm excited about the most recent wave of innovations in AI and cheap manufacturing and the ability to make things that are extremely complicated, systems that are very complicated at a very low price. 42 00:02:46.040 --> 00:02:54.724 So you combine the intelligence with the cheap manufacturing together and you have products otherwise wouldn't have been possible for certain use cases. 43 00:02:55.364 --> 00:02:56.585 Robotics is not new. 44 00:02:58.147 --> 00:02:59.167 Automation isn't new. 45 00:02:59.488 --> 00:03:07.294 We've had automation for decades, if not a century, with the advent of the assembly line, you know, with the creation of Ford. 46 00:03:08.015 --> 00:03:19.205 And so my view is that over the past, I would say, 15 years, with AI in the form of convolutional neural nets in the early 2010s, 47 00:03:21.534 --> 00:03:33.318 and the continued slope of cost reduction with robotic arms, we've seen a whole new wave of new robotics applications. 48 00:03:33.538 --> 00:03:36.739 And I think the humanoid robotics are a manifestation of that. 49 00:03:37.339 --> 00:03:43.101 But my personal excitement in robotics and automation isn't rooted in the humanoids. 50 00:03:43.721 --> 00:03:50.824 It's rooted in this more macro tailwind around the software that drives the robots, 51 00:03:51.744 --> 00:03:55.746 And the commoditization of the hardware that the software runs on. 52 00:03:55.866 --> 00:04:12.093 If you rewind back to, say, the 1970s, there was a similar opportunity, a similar moment around the integration of microprocessors, which were the equivalent of AI of our time. 53 00:04:12.868 --> 00:04:13.968 into robotics. 54 00:04:14.609 --> 00:04:21.952 And that's how you saw robotics enter into, for example, automotive manufacturing. 55 00:04:21.972 --> 00:04:35.557 You had these robotic arms that had a level of intelligence in them that allowed them to be programmed to very fine specifications, to do very specific tasks like welding, riveting, gluing, painting, 56 00:04:36.237 --> 00:04:37.798 various types of inspections. 57 00:04:38.459 --> 00:04:45.986 And that's what began this rapid proliferation of robotics into the broader manufacturing settings. 58 00:04:46.592 --> 00:04:57.837 Whereas today, when you walk into most automotive manufacturing facilities, a lot of the basic steps in automotive manufacturing are completely automated. 59 00:04:57.857 --> 00:05:12.044 You don't see any people involved at all until the later steps where various parts of the interior, the wiring harnesses, the glass are starting to be installed in the vehicle. 60 00:05:12.504 --> 00:05:25.454 And what we're seeing today, to the point that you brought up, is a lot of the companies that sell into these larger companies also having the benefit of automation. 61 00:05:25.714 --> 00:05:30.518 So when you walk into a Ford factory, you see a ton of automation. 62 00:05:31.189 --> 00:05:42.366 But if you walk into the factory that sells the components that are sold into the Ford factory, you may not see as much automation because they don't have the budgets. 63 00:05:42.586 --> 00:05:46.032 They don't have the investment to be able to invest. 64 00:05:46.838 --> 00:05:49.980 the way Ford can invest in its automation. 65 00:05:50.420 --> 00:06:02.147 And now what we're seeing with AI is just like how it is now easy for anybody to generate code, it's becoming as easy for anyone to program a robot and these robots are getting very cheap. 66 00:06:02.847 --> 00:06:06.990 And so now you're seeing a lot of these companies who otherwise didn't have access 67 00:06:07.550 --> 00:06:08.231 to robots. 68 00:06:08.311 --> 00:06:15.898 These industries otherwise didn't have access to robots, now getting access to them in a way we hadn't seen before. 69 00:06:16.378 --> 00:06:27.208 So that's what catalyzes my excitement around robots, which may be slightly different than what's catalyzing yours or perhaps others' excitement around robotics. 70 00:06:27.489 --> 00:06:29.370 And that would just perhaps be in the minority of 71 00:06:29.991 --> 00:06:39.923 Folks that probably, you know, I'm optimistic, but I'm not as excited about what we're seeing right now in humanoid robots yet as perhaps others are. 72 00:06:40.251 --> 00:06:48.973 I do want to talk to you about humanoid robotics, but I think maybe it's worth putting a pin in that and just saving that for later and really diving into what you just discussed right now. 73 00:06:49.033 --> 00:06:57.795 Maybe if I could just try and summarize your excitement, it's really about the integration of intelligence into automation. 74 00:06:58.215 --> 00:07:09.298 And inside of automation, in addition to that, intelligence is allowing for cost reduction, deflation for what it means to automate things. 75 00:07:10.418 --> 00:07:20.649 And so it's about the collision of automation and intelligence and also the accessibility of automation to be applied to more and more industries. 76 00:07:20.769 --> 00:07:24.152 And that's what you really get excited about as an investor. 77 00:07:24.172 --> 00:07:24.793 That's right. 78 00:07:24.813 --> 00:07:28.517 And as somebody who is excited about like growing the GDP of the United States. 79 00:07:28.557 --> 00:07:30.739 This is a fair summary of your excitement. 80 00:07:30.959 --> 00:07:32.140 As an investor, David. 81 00:07:33.130 --> 00:07:43.513 Many VCs are excited about, you know, giant markets and unfair advantages and monopolistic businesses in those markets. 82 00:07:44.193 --> 00:07:49.115 I'm particularly excited about opportunities to create markets where none exist. 83 00:07:49.675 --> 00:08:00.078 And what we're seeing with robotics is that markets for robotics are being created where markets for robots previously didn't exist for the reasons that you mentioned. 84 00:08:00.778 --> 00:08:09.063 the AI, the software, the commoditization of the hardware, which is now creating a market where one didn't exist. 85 00:08:10.164 --> 00:08:13.966 And there's companies out there that are in a position to now dominate those nascent markets. 86 00:08:14.421 --> 00:08:16.362 Can we try and define some terms a little bit? 87 00:08:16.402 --> 00:08:19.662 We're using robotics and we're also using automation. 88 00:08:19.782 --> 00:08:21.023 Are those the same terms? 89 00:08:21.203 --> 00:08:21.963 Are they different? 90 00:08:22.203 --> 00:08:25.784 And how should we think about these things when we're talking about this industry? 91 00:08:26.184 --> 00:08:26.864 Good question. 92 00:08:26.924 --> 00:08:31.906 So I look at automation as a solution to a problem. 93 00:08:32.126 --> 00:08:35.487 So there are many aspects associated with automation. 94 00:08:35.507 --> 00:08:36.887 There's a financing aspect. 95 00:08:37.207 --> 00:08:38.508 aspect associated to it. 96 00:08:38.609 --> 00:08:42.853 There's an engineering aspect associated with it, which has nothing to do with the robot. 97 00:08:42.893 --> 00:08:44.955 How do you engineer your factory floor 98 00:08:47.024 --> 00:08:54.429 and your workflows to be able to benefit from robotic participation. 99 00:08:55.529 --> 00:08:57.250 It is the physical robot itself. 100 00:08:57.611 --> 00:08:59.532 It's the software that runs on the robot. 101 00:08:59.912 --> 00:09:07.176 It's the systems that are put in place to maintain the robots and make sure that they're up 99.999% of the time. 102 00:09:08.057 --> 00:09:16.342 And so I view all of those resources culminating in the end goal of, 103 00:09:17.320 --> 00:09:30.945 the automation of a task to fall under that envelope of automation with the physical robots and the software that is running on that robot being a component of that? 104 00:09:31.365 --> 00:09:32.326 But that's a great question. 105 00:09:32.546 --> 00:09:42.810 Automation is the broad category, and we need a bunch of robots to create the process of automation, but automation and robots are not the same thing. 106 00:09:43.150 --> 00:09:43.470 Correct. 107 00:09:44.451 --> 00:09:51.093 Robots obviously automate, but automation is not just limited to robots. 108 00:09:51.753 --> 00:09:54.574 Self-custody won, but it still has a usability problem. 109 00:09:54.834 --> 00:09:57.695 A seed phrase on paper is still a single point of failure. 110 00:09:57.855 --> 00:10:00.656 Phones get lost, devices break, backups disappear. 111 00:10:00.856 --> 00:10:04.277 BitKey is a self-custodial hardware wallet built for that reality. 112 00:10:04.437 --> 00:10:09.598 It uses a two-of-three multi-sig with keys split across your phone, the BitKey hardware device, and Block. 113 00:10:09.778 --> 00:10:13.179 No single key can move your Bitcoin, and Block can never move it alone. 114 00:10:13.319 --> 00:10:15.100 There's no seed phrase to lose or expose. 115 00:10:15.280 --> 00:10:19.801 And if you lose your phone, your BitKey or both, built-in recovery gives you a path back. 116 00:10:19.981 --> 00:10:24.002 The hardware screen also lets you verify the destination before approving a transaction. 117 00:10:24.162 --> 00:10:24.862 And that's the point. 118 00:10:24.942 --> 00:10:27.743 More control without one mistake putting everything at risk. 119 00:10:27.883 --> 00:10:32.524 So download BitKey today and use promo code BANKLESS to get 10% off of your BitKey. 120 00:10:32.784 --> 00:10:34.485 This episode has been sponsored by BitKey. 121 00:10:34.665 --> 00:10:36.747 Markets don't move one asset at a time. 122 00:10:36.967 --> 00:10:40.391 One day it's Bitcoin, the next NVIDIA, then gold, and then the S&P. 123 00:10:40.471 --> 00:10:45.956 But most traders are still managing their portfolio across different platforms, different accounts, and different pools of capital. 124 00:10:46.096 --> 00:10:47.477 BitGet just changed that. 125 00:10:47.658 --> 00:10:54.104 Their new Stocks 2.0 product lets you trade tokenized equities directly with USDT, all inside the same app you already use for crypto. 126 00:10:54.124 --> 00:10:54.224 Yeah. 127 00:10:54.224 --> 00:10:56.325 This is not just another tokenized stock product. 128 00:10:56.485 --> 00:11:05.287 Stocks 2.0 is designed around deeper liquidity, faster execution, and the lowest fees on the market at just 0.04% and one-to-one economic exposure to the underlying stock. 129 00:11:05.388 --> 00:11:12.590 Dividend, stock splits, and other corporate actions are reflected automatically, helping your position stay aligned with the asset you actually want exposure to. 130 00:11:12.770 --> 00:11:18.292 One platform, one account, multiple markets, crypto equities, commodities, and more, all accessible with USDT. 131 00:11:18.392 --> 00:11:19.372 BitGet, trade smarter. 132 00:11:19.452 --> 00:11:21.553 Start trading today through the link in the show notes. 133 00:11:21.613 --> 00:11:22.733 This is not investment advice. 134 00:11:22.813 --> 00:11:28.855 I've been trading crypto for almost a decade, and I've used so many different wallets, exchanges, aggregators, different front ends over the years. 135 00:11:28.975 --> 00:11:30.815 And I'm always kind of looking for the same thing. 136 00:11:30.975 --> 00:11:40.978 Just one interface with deep liquidity across a bunch of chains and assets where I can access all the markets like perps, earn yield, trade confidentially and still control all my own funds. 137 00:11:41.318 --> 00:11:43.279 And I've never really found this experience. 138 00:11:43.299 --> 00:11:47.402 And I'm always switching wallets, juggling gas fees, and just getting eaten by slippage. 139 00:11:47.682 --> 00:11:49.363 Near.com is not that. 140 00:11:49.663 --> 00:11:51.244 It feels fundamentally different to me. 141 00:11:51.324 --> 00:11:55.746 I can do everything I want from any chain, and I keep all my activity confidential. 142 00:11:55.946 --> 00:11:58.728 I can even earn yield confidentially. 143 00:11:58.968 --> 00:12:00.709 It's the way crypto should work. 144 00:12:00.909 --> 00:12:10.615 The Near.com wallet is powered by Near, and it's moved over $25 billion cross-chain using post-quantum signing and has run over five years on mainnet with zero downtime. 145 00:12:10.755 --> 00:12:14.017 Near.com is simply the best way to be on chain and be in control. 146 00:12:14.137 --> 00:12:19.099 Get 20% of your trading fees back using the bankless link in the show notes, not investment advice. 147 00:12:19.419 --> 00:12:23.941 And then there's a spectrum of generalizability, I feel. 148 00:12:23.961 --> 00:12:30.865 I think people probably have an image in their head of the forward floor and there's like a fixed arm that can actuate and move. 149 00:12:31.145 --> 00:12:33.466 And it does a lot of the actual work. 150 00:12:34.086 --> 00:12:35.708 But it feels very precise. 151 00:12:35.788 --> 00:12:36.768 It feels very locked in. 152 00:12:36.788 --> 00:12:38.290 It's meant to do that one job. 153 00:12:38.690 --> 00:12:46.757 Maybe people can also think of an Amazon floor of little robots running around moving packages and they're not fixed. 154 00:12:46.817 --> 00:12:51.441 And maybe they're slightly more generalized than the car building arm 155 00:12:52.503 --> 00:12:56.908 Is there a spectrum to discuss about the generalizability of robots? 156 00:12:57.109 --> 00:13:01.034 And are you particularly bullish on a region of generalizability? 157 00:13:01.094 --> 00:13:06.822 Or is this not really a subject that is discussed much when it comes to robotics and automation? 158 00:13:07.583 --> 00:13:15.128 I feel like the automation solutions are as broad as the problem set. 159 00:13:15.489 --> 00:13:29.839 So as you stated, the problem associated with order fulfillment in an Amazon warehouse is very different from the problem associated with assembling the components of a vehicle. 160 00:13:29.859 --> 00:13:31.400 They're two very different problems. 161 00:13:32.020 --> 00:13:35.602 And they demand very different automation solution. 162 00:13:36.103 --> 00:13:52.493 Now, I can see from the perspective of a founder or from the perspective of an investor in a company to be motivated to champion their investments or their company as the singular solution to all problems. 163 00:13:53.434 --> 00:13:54.595 There is an economic motivation there. 164 00:13:57.085 --> 00:14:04.667 in presenting that picture that, hey, I'm building a widget that's going to solve every problem associated with automation. 165 00:14:05.367 --> 00:14:14.470 And it has a human form factor because humans do many tasks, if not most tasks. 166 00:14:16.290 --> 00:14:22.972 I would argue that that is more of an economically driven argument, more so than a practical one. 167 00:14:24.560 --> 00:14:38.550 The way I view the world and the way I view the opportunity is that specific applications are best suited to certain types of robotics and automation solutions. 168 00:14:39.811 --> 00:14:46.736 And so my expectation is that there will be an opportunity for a 169 00:14:48.098 --> 00:15:03.932 humanoid form factor machine that does some subset of tasks that are performed by humans today, but they will be one of many automation solutions that will take on many shapes and forms. 170 00:15:04.032 --> 00:15:05.313 So if you look at the robot 171 00:15:06.360 --> 00:15:34.604 that is used in the Ford factory today, that is an arm that has, for example, a welder and defector, it probably has zero resemblance to the robotic cart in an Amazon warehouse that shuttles boxes from one location of the warehouse to another for either stocking or order fulfillment. 172 00:15:35.253 --> 00:15:40.779 These robots have practically nothing to do with each other, except perhaps both being made from metal and plastic. 173 00:15:40.799 --> 00:15:47.025 And I think that's what having some computer chips in them and cameras on them, but that perhaps that's where the resemblance ends. 174 00:15:47.585 --> 00:15:52.010 Yet both bring significant value to their end use cases. 175 00:15:52.670 --> 00:15:54.752 And so it's my expectation that, 176 00:15:55.802 --> 00:16:05.166 those specific form factors will continue to proliferate alongside the humanoids, which will have their own place in the market. 177 00:16:05.606 --> 00:16:22.272 But I personally, as an end of one, don't see a single solution dominating all problems because I just don't think a single solution will be able to be best served to most problems. 178 00:16:22.753 --> 00:16:25.998 So would you say that there is the humanoid end of the spectrum? 179 00:16:26.178 --> 00:16:33.748 And maybe one of the reasons why that's fun and sexy to talk about is that the form factor is going to be consistent across companies. 180 00:16:34.389 --> 00:16:37.914 And I mean, they look like us and so we can get excited about them. 181 00:16:38.635 --> 00:16:44.177 But I think what I just heard you say is that that's kind of going to be the only consistent form factor. 182 00:16:44.717 --> 00:16:59.663 And when we talk about the rest of robotics that it takes to create automation, the form factors are going to be highly heterogeneous and non-overlapping and specific and tailored to their needs. 183 00:16:59.803 --> 00:17:07.666 And one of the inputs that we need to make that world work is the deflation and cost of creating this whole industry in the first place, which we talked about earlier. 184 00:17:07.846 --> 00:17:09.067 Would you agree with that assessment? 185 00:17:09.488 --> 00:17:11.370 I would say yes, broadly. 186 00:17:11.790 --> 00:17:18.498 There are components that go into these robots that will be consistent across use cases. 187 00:17:18.838 --> 00:17:28.869 The cameras, the force sensors, the reduction gear sets, the encoders, the power management systems, the... 188 00:17:31.571 --> 00:17:33.193 the actual arms. 189 00:17:33.413 --> 00:17:34.334 The building blocks. 190 00:17:34.354 --> 00:17:36.235 The building blocks will be consistent. 191 00:17:36.255 --> 00:17:49.266 For example, if you look at a smartphone, it may have little resemblance with, for example, a laptop, a small, inexpensive laptop. 192 00:17:49.626 --> 00:17:54.910 The use case for the cell phone may be very different from the use case, let's say, for example, for an iPad. 193 00:17:55.991 --> 00:18:10.178 Yet, a lot of the technology that goes into the phone, the ARM-based processor, the memory, the Wi-Fi interface, the video driver, even the LCD screen itself, there are different form factors. 194 00:18:10.198 --> 00:18:17.101 They have different specifications, but they share the same basic technological architecture. 195 00:18:17.161 --> 00:18:21.683 Even their operating systems may share many, many thousands of lines of code. 196 00:18:22.275 --> 00:18:28.577 but they're used for very different use cases with a similar technology basis. 197 00:18:28.597 --> 00:18:40.180 So when you see a robotic arm that is doing welding, it may have the same motors, encoders, force sensors, cameras, as the robot that's doing the riveting. 198 00:18:40.700 --> 00:18:46.182 But the robot that's doing the riveting may have a larger, for example, range of motion. 199 00:18:46.762 --> 00:18:50.103 It has a different end effector attached to it. 200 00:18:51.219 --> 00:19:00.764 And so it is engineered for that use case, yet that use case is very broad, but not as broad as doing everything for everybody. 201 00:19:01.184 --> 00:19:17.212 So I generally take issue with the notion of we're building a robot that's going to perform all tasks because I think it'll be very difficult for a robot that performs all tasks to be better than a robot that's specialized. 202 00:19:17.797 --> 00:19:19.898 to a certain task? 203 00:19:20.698 --> 00:19:24.619 And why would someone not use a robot that's specialized to that certain task? 204 00:19:24.639 --> 00:19:42.445 And so on the specialization end of the spectrum, the specialization to generalization spectrum, I guess the bottleneck is if we indeed do have all of the raw ingredients that it takes to make a robot, we have the power, we have the actuators, the ball bearings, 205 00:19:43.285 --> 00:19:51.688 really the constraint is the ability to put them in the right package according to the actual task at hand. 206 00:19:51.748 --> 00:19:52.949 Is that the current constraint? 207 00:19:53.069 --> 00:19:56.890 Is great, we have the pieces, but we need to put them in the right shapes. 208 00:19:57.591 --> 00:20:08.575 And that part can be more difficult because there's more possible shapes to order all the pieces in than could, than, than, you know, there's so many possible shapes that, 209 00:20:08.735 --> 00:20:11.036 to order all the pieces that it's a little bit hard to bear. 210 00:20:11.216 --> 00:20:12.037 Is that a fair assessment? 211 00:20:12.157 --> 00:20:17.520 And that's why you've only seen robotics proliferate in these larger industrial settings. 212 00:20:18.280 --> 00:20:24.124 Ford has tens of millions of dollars that it can deploy to companies who specialize in this. 213 00:20:24.204 --> 00:20:27.125 Companies like Honeywell, Dermatix, Symbotic, Winright. 214 00:20:27.606 --> 00:20:33.409 These companies who specialize in engineering robotic solutions for a particular product 215 00:20:33.882 --> 00:20:34.342 use case. 216 00:20:34.502 --> 00:20:36.403 In Ford's case, it would be an assembly line. 217 00:20:36.843 --> 00:20:46.407 So the robots that, for example, put a sheet of aluminum into a stamp, the robots that take those sheets of aluminum, the stamped aluminum out of the stamp, put into another stamp. 218 00:20:46.747 --> 00:20:52.769 And then the robot that has a camera at the end that inspects the stamped part to make sure there's no defects. 219 00:20:53.149 --> 00:21:05.933 and then the robot that puts two pieces of stamped metal together, and the other robot, a third robot, doing the weld between those pieces of sheet metal, those stamped pieces of sheet metal. 220 00:21:06.353 --> 00:21:12.175 And then robots doing the riveting and the gluing, and then lowering it into a bath of acid, and the robots that are doing the painting. 221 00:21:12.902 --> 00:21:22.289 So there are many tasks that need to be automated, that need to be engineered by the integrator, which is extremely expensive. 222 00:21:22.930 --> 00:21:30.996 And there are few companies today that can afford to do that with the legacy technology that was available to us until today. 223 00:21:31.256 --> 00:21:40.203 And my enthusiasm is rooted to the earlier part of the conversation and these robots becoming less expensive and more. 224 00:21:40.943 --> 00:22:04.163 the engineering that goes into making them useful becoming far more accessible, just like how anybody can now speak to a computer and generate code, I see a future where someone can speak to an interface powered by AI that will then program the robot to do a specific task rather than you having to hire an engineer. 225 00:22:05.123 --> 00:22:06.044 to do that for you. 226 00:22:06.164 --> 00:22:13.888 And we'll see the more rapid adoption of automation and more and more settings that until now didn't have that access. 227 00:22:14.368 --> 00:22:15.709 So you're right. 228 00:22:15.969 --> 00:22:25.895 It is taking the individual components and building all the software and the tools that you need to actually make that system that you build for your use case useful. 229 00:22:26.075 --> 00:22:42.146 It feels like kind of building a developer platform of sorts in that we have all the basic ingredients and we need to be able to allow people to tinker with, play with, combine all of the pieces in order to suit their needs. 230 00:22:42.526 --> 00:22:54.635 And making this more and more accessible is going to allow the market to place automation in deeper, more niche, more specific parts of the market so long as we can just figure out how to open up 231 00:22:55.315 --> 00:22:58.518 robotics to just and be more accessible. 232 00:22:58.759 --> 00:22:59.840 That's kind of what it feels like. 233 00:22:59.980 --> 00:23:02.723 And that's what companies are doing today, which fuels my enthusiasm. 234 00:23:02.743 --> 00:23:03.724 I'll give you an example. 235 00:23:04.244 --> 00:23:05.285 So tell me about them. 236 00:23:05.586 --> 00:23:11.231 So if you wanted to develop a robot that does some kind of unstructured task. 237 00:23:11.251 --> 00:23:17.738 So when you're talking about welding two pieces of sheet metal together, it is a very structured task. 238 00:23:18.387 --> 00:23:22.250 You are within the, you know, thousands of a millimeter. 239 00:23:22.770 --> 00:23:27.694 You know where this robotic end effector is going and you know exactly what it's doing. 240 00:23:27.734 --> 00:23:34.479 When you're talking about the less structured task, it becomes extremely challenging to develop automation for. 241 00:23:35.759 --> 00:23:46.427 If you fast forward to today, you have companies like Physical Intelligence that we invested in that are building models that specialize in various unstructured tasks. 242 00:23:46.947 --> 00:23:56.233 And they're putting many of those models on open source repositories for roboticists to use and fine tune for their individual use cases. 243 00:23:56.253 --> 00:24:01.637 This is extremely powerful and a huge enabler for the community. 244 00:24:02.124 --> 00:24:06.127 that didn't exist until very recently. 245 00:24:06.487 --> 00:24:26.859 So what would have required many PhDs and millions of dollars of funding and many years now can be achieved by an undergraduate student, maybe even a high school student, downloading and installing one of these models and fine tuning the robot to perform a specific task. 246 00:24:28.158 --> 00:24:32.961 Your question, follow-on question could be, well, why don't we have robots everywhere today? 247 00:24:33.702 --> 00:24:44.709 It's that we still have some work to do to make these robots more reliable and to make them faster and to make them competitive with labor. 248 00:24:45.029 --> 00:24:52.254 And we can get into that, if you'd like, on the role that labor has in the proliferation of robotics. 249 00:24:52.614 --> 00:24:53.575 I do want to get into that. 250 00:24:53.595 --> 00:24:55.216 Let me ask you this one last question 251 00:24:56.517 --> 00:25:02.681 It seems, I was going to ask you, are we on the cusp of something big in robotics? 252 00:25:03.201 --> 00:25:13.848 But it sounds like, and what I mean by that is that, you know, we have AI, we have, you know, decreasing costs of automation, feels like we have the raw ingredients. 253 00:25:14.208 --> 00:25:16.250 And so we're close, you know, we're close to a cusp. 254 00:25:16.910 --> 00:25:19.410 it kind of sounds like we're actually, the cusp is behind us. 255 00:25:19.571 --> 00:25:23.971 And actually, the ball is materially rolling towards this direction. 256 00:25:24.511 --> 00:25:31.233 To your point, we don't, as me as a consumer, you know, somebody who walks around the streets of New York is not impacting my life yet. 257 00:25:31.713 --> 00:25:38.014 But in the factories, the frontier of this intelligence is progressing, and progress is being made. 258 00:25:38.594 --> 00:25:42.555 We don't quite yet have the very high reliability that we need. 259 00:25:42.615 --> 00:25:43.375 But in terms of 260 00:25:44.055 --> 00:25:55.825 you know, a step function change in progress in the world of automation and robotics, it sounds like that ball is already rolling and it's no longer an if, it's already like happening now. 261 00:25:56.005 --> 00:25:56.526 Is that correct? 262 00:25:56.706 --> 00:25:57.166 That's right. 263 00:25:57.386 --> 00:26:04.733 And the analogy that I like to use and the evidence that I have that points to the theme that you're sharing is compute, compute. 264 00:26:05.750 --> 00:26:12.897 If you look at the 70s and 80s, people regarded computers as pieces of hardware. 265 00:26:13.437 --> 00:26:22.886 When you thought about a computer, you thought about what the processor speed was, how much RAM did it have, what was the graphics interface. 266 00:26:23.347 --> 00:26:28.132 That was what defined personal computing in the 80s and into the early 90s. 267 00:26:30.606 --> 00:26:38.417 the rapid proliferation occurred, that, I guess, that tipping point occurred with the internet. 268 00:26:39.078 --> 00:26:48.130 And compute was no longer regarded as a piece of, let's say, capital equipment, something physical or a machine 269 00:26:48.791 --> 00:26:50.473 It was regarded as a capability. 270 00:26:51.014 --> 00:26:59.526 You went on the internet, you consumed content, you stored your files, and the physical hardware was completely abstracted away. 271 00:27:00.267 --> 00:27:03.972 I feel like right now we're in the equivalent of the early 90s, 272 00:27:05.053 --> 00:27:11.655 of compute with robotics, where when you and I have a conversation, we're still talking about the physical robots. 273 00:27:11.675 --> 00:27:12.876 We're talking about the machine. 274 00:27:12.896 --> 00:27:18.037 We're talking about the plastic and the bearings and the software. 275 00:27:18.437 --> 00:27:22.719 We're still not talking about the output of the machine. 276 00:27:22.739 --> 00:27:31.762 The emphasis is not on the output of the machine the way the emphasis today is on the output of compute, which is evidence of this mass 277 00:27:32.943 --> 00:27:33.724 proliferation. 278 00:27:34.124 --> 00:27:40.392 So I feel like we are on that trend towards abstracting away the plastic and the metal 279 00:27:41.488 --> 00:27:47.050 and the cool demos and focusing on what these robots are going to do for us. 280 00:27:47.891 --> 00:27:50.912 And I feel like that is a direction that we're headed towards. 281 00:27:50.932 --> 00:27:52.613 And that's what makes me very excited. 282 00:27:52.813 --> 00:27:58.555 Let's go into what you were talking about a second ago with a labor input into robotics. 283 00:27:58.635 --> 00:28:00.756 I'm actually not sure where this conversation leads. 284 00:28:00.776 --> 00:28:03.217 So I think I need you to actually take the reins here. 285 00:28:03.577 --> 00:28:05.978 What is significant about this part of this conversation? 286 00:28:06.298 --> 00:28:17.506 Yeah, so there's always this ongoing debate about the question of how robots affect labor. 287 00:28:17.686 --> 00:28:21.969 Human labor, human jobs, are we going to be automated out of a job like this question? 288 00:28:21.989 --> 00:28:29.433 There's this general concern, which is a warranted concern, it's justified, that if you increase automation, 289 00:28:30.334 --> 00:28:36.678 then you are reducing opportunities for labor. 290 00:28:37.318 --> 00:28:43.342 And the enemy of labor, the enemy of work is automation. 291 00:28:44.043 --> 00:28:47.165 And if you observe historical trends... 292 00:28:47.950 --> 00:29:01.096 you'll see that economies who benefit from more automation, who adopt more automation, tend to also benefit from less unemployment. 293 00:29:02.097 --> 00:29:08.039 And economies who do not automate tend to suffer from more unemployment. 294 00:29:08.059 --> 00:29:11.301 So that's been the general trend. 295 00:29:11.921 --> 00:29:15.943 And so I would say that the enemy here of both 296 00:29:17.277 --> 00:29:24.062 labor and jobs, as well as automation, is cheap labor. 297 00:29:24.242 --> 00:29:31.467 So cheap labor here is the common threat or the common, let's say, enemy here. 298 00:29:31.967 --> 00:29:41.254 When you have cheap labor somewhere else, then you are now threatening the jobs or the employment 299 00:29:42.274 --> 00:29:43.735 in your market. 300 00:29:44.055 --> 00:29:55.541 And by the same token, when you have cheap labor available, then you're increasing the hurdle rates that automation needs to cross in order to be able to proliferate. 301 00:29:55.961 --> 00:30:03.165 And the observation that I've had is that when you automate, yes, you may be displacing 302 00:30:04.105 --> 00:30:14.811 jobs, but you are disproportionately creating higher quality jobs where you have less churn, where you have greater worker satisfaction. 303 00:30:15.251 --> 00:30:17.092 We're investors in a company called Formic. 304 00:30:17.473 --> 00:30:21.535 And what they're doing is offering robots as a service. 305 00:30:21.575 --> 00:30:23.676 What they sell is not a physical machine. 306 00:30:24.116 --> 00:30:26.818 What they sell is the output of a machine. 307 00:30:26.858 --> 00:30:28.619 They abstract away the machine 308 00:30:29.576 --> 00:30:39.200 And they make it as easy for you as the owner or operator of a factory or a warehouse to automate as it is for you to hire labor. 309 00:30:39.981 --> 00:30:49.365 And these customers are not deploying automation because they want to reduce their headcount or save money. 310 00:30:49.965 --> 00:30:53.747 They're trying to do this because they can't find workers. 311 00:30:54.607 --> 00:30:57.649 And these are jobs that have extremely high... 312 00:30:58.389 --> 00:30:59.010 churn rates. 313 00:30:59.730 --> 00:31:15.524 And what they've actually experienced is creating more and more jobs that are higher quality that do not suffer from the worker dissatisfaction and the churn that they experienced prior to adopting automation. 314 00:31:16.287 --> 00:31:24.077 And again, the enemy of all of this is not, the enemy of jobs and automation, again, is not the robots. 315 00:31:25.780 --> 00:31:27.862 It is the cheap labor, if that makes sense. 316 00:31:28.163 --> 00:31:29.424 Yeah, yeah, I think it does. 317 00:31:29.445 --> 00:31:33.870 And I think that takes us to where I want to go next, which is just the impact of, 318 00:31:34.531 --> 00:31:38.252 on the economy of a successful automation industry. 319 00:31:38.793 --> 00:31:40.533 I think you talked about it just a little bit now. 320 00:31:40.713 --> 00:31:42.914 We get more higher quality jobs. 321 00:31:43.414 --> 00:31:48.576 Yes, there may be some short-term dislocation of jobs, jobs from A to B. 322 00:31:49.096 --> 00:31:54.158 Perhaps there is short-term economic pain in some industries, but in the long term, 323 00:31:54.778 --> 00:31:59.081 in the long term and the trends point towards more higher paying jobs. 324 00:31:59.561 --> 00:32:12.369 What about just like the rest of the economy, the cost of goods, the cost of food, the cost of like building a physical like non-software, non-SaaS based startup 325 00:32:14.270 --> 00:32:18.873 say the automation industry does everything that it hopes to do in the next decade. 326 00:32:19.353 --> 00:32:26.418 What would that mean for just the average consumer's life in terms of prices and just what the impact on the economy would be? 327 00:32:26.638 --> 00:32:36.385 I mean, I'll give you the broad answer, and that would be more selection, more competition, and ultimately a benefit to consumers. 328 00:32:36.625 --> 00:32:42.189 So if a factory, so factories, for example, that our company Formake is selling automation into, 329 00:32:42.867 --> 00:32:45.008 They have more productivity. 330 00:32:45.268 --> 00:32:46.649 Their workers get paid more. 331 00:32:47.469 --> 00:32:51.511 They generate from better unit economics. 332 00:32:52.412 --> 00:32:56.734 And they ultimately are able to sell better products at better prices. 333 00:32:57.194 --> 00:33:00.015 And the benefit trickles down to the end customer. 334 00:33:00.075 --> 00:33:09.260 So it's my expectation, my belief, that if you're able to make automation simple enough to adopt and make it reliable enough, 335 00:33:09.980 --> 00:33:22.766 then everybody from the people that sell products to these automation customers to the people that purchase the products from them will ultimately benefit. 336 00:33:22.806 --> 00:33:28.248 And I think the key point here is educating our workforce. 337 00:33:28.948 --> 00:33:34.211 If we make sure that our workforce is educated and has the opportunity to grow their skill set, 338 00:33:39.877 --> 00:33:45.903 more higher paying and higher quality jobs. 339 00:33:46.044 --> 00:33:55.634 And so I think that's something that we as a society have to take upon ourselves to keep our workforce, to make sure our workforce is prepared for this new generation of work that's going to come about 340 00:33:56.342 --> 00:34:03.848 as a result of automation, as opposed to looking at it as, oh no, jobs are being eliminated, so what do we do with our workforce? 341 00:34:04.508 --> 00:34:05.929 We should be more proactive about it. 342 00:34:06.209 --> 00:34:14.376 Is there a particular industry that you get really excited about when it comes to the potential of said industry with automation? 343 00:34:14.616 --> 00:34:21.461 Like agriculture, I could imagine, gets scaled, but I'm sure there's a handful of industries that it's possible to talk about. 344 00:34:21.501 --> 00:34:23.783 Does one stick out to you as particularly exciting? 345 00:34:24.503 --> 00:34:28.905 Yeah, I mean, if you look at agriculture historically, agriculture is one of the first industries to be automated. 346 00:34:28.945 --> 00:34:38.188 And as a result of that, we went from, you know, food shortages to food being completely abundant as a result. 347 00:34:39.248 --> 00:34:44.230 You know, you look at industries like automotive, consumer electronics. 348 00:34:44.370 --> 00:34:46.450 These are industries that are highly automated. 349 00:34:47.151 --> 00:34:48.071 And as a result... 350 00:34:48.501 --> 00:34:51.483 You know, we have cars that are extremely safe, extremely efficient. 351 00:34:52.223 --> 00:35:01.109 And, you know, many of us have the benefit of either keeping our cars for 10 years, like myself, you know, that are reliable and run for a long time. 352 00:35:01.349 --> 00:35:04.331 Or we can go out and buy the latest and greatest car. 353 00:35:04.952 --> 00:35:08.174 And the newer the cars get, the cheaper they get and the more capable they get. 354 00:35:09.103 --> 00:35:10.824 And the more safe and efficient that they get. 355 00:35:10.904 --> 00:35:14.425 So these are all benefits that come to us. 356 00:35:15.225 --> 00:35:25.549 I feel like the zeitgeist today, David, going back to your earlier question about excitement, is around robots in domestic settings. 357 00:35:26.289 --> 00:35:29.330 So we're talking about companion robots. 358 00:35:29.590 --> 00:35:36.353 We're talking about butler robots, you know, robots and household settings. 359 00:35:37.513 --> 00:35:45.521 I am a little cautious about those kinds of use cases for two reasons. 360 00:35:46.381 --> 00:35:55.670 One, because I feel like there's still a huge opportunity for automation in factory and warehousing settings. 361 00:35:56.302 --> 00:35:58.082 There is plenty of work to do there. 362 00:35:58.122 --> 00:36:01.803 There's plenty of opportunities to automate in those settings. 363 00:36:02.743 --> 00:36:15.446 And I feel like it may not be as sexy as a companion robot or a maid robot, but I feel like those applications are still very, very real. 364 00:36:16.726 --> 00:36:19.646 We're seeing the proliferation of autonomous cars. 365 00:36:19.946 --> 00:36:22.747 I mean, they are a fantastic example of a robot. 366 00:36:23.548 --> 00:36:25.750 Um, there's a lot more room there. 367 00:36:25.830 --> 00:36:33.397 We still haven't, we were promised, um, uh, you know, the, the big rigs and the trucks, uh, to become automated. 368 00:36:33.437 --> 00:36:35.099 They still haven't become fully automated. 369 00:36:35.139 --> 00:36:40.464 So we'll see hopefully in the near future, more autonomous vehicles, autonomous trucks. 370 00:36:41.833 --> 00:36:59.543 And I feel like the future of a humanoid robot tackling the challenges in a domestic setting, cleaning up after a child, helping an elderly person, doing basic tasks like emptying the dishwasher, I think we'll see that come. 371 00:37:00.104 --> 00:37:02.025 But I feel like that is further out. 372 00:37:02.705 --> 00:37:05.827 And I'm more excited about these near-term opportunities now. 373 00:37:06.187 --> 00:37:09.470 and more, I would say, industrial and commercial settings. 374 00:37:09.870 --> 00:37:16.335 Because those are the opportunities that scale to all of society in a very efficient way, right? 375 00:37:16.435 --> 00:37:19.217 It's getting a humanoid robot into everyone's home. 376 00:37:19.618 --> 00:37:21.079 That's like a huge last mile problem. 377 00:37:21.099 --> 00:37:23.641 There's huge constraints with the actual design of the robot. 378 00:37:24.121 --> 00:37:37.057 But I think one of the reasons why I think you're excited is that there are a few places where you can put robots and automation into that have just massive benefits for society as a whole. 379 00:37:37.077 --> 00:37:37.898 That's right. 380 00:37:37.978 --> 00:37:43.605 And so it's a little bit of the scalability of the effect, I think, is something that gets you excited about the automation side. 381 00:37:44.025 --> 00:37:53.712 I don't want to say easier because all these problems are challenging, but I feel like the problems are more constrained in these industrial and commercial settings. 382 00:37:54.673 --> 00:38:04.700 And the value proposition or the economics that are more well-defined than robots going into a domestic setting. 383 00:38:04.800 --> 00:38:07.862 When you buy, for example, a piece of furniture, 384 00:38:08.871 --> 00:38:16.014 Um, the, you know, my, you know, I may sleep on my couch, you know, uh, every afternoon and take a nap. 385 00:38:16.434 --> 00:38:22.577 You may not even sit on that couch, you know, uh, for months, depending on like, you know, how we look at couches. 386 00:38:22.657 --> 00:38:22.977 And so. 387 00:38:23.650 --> 00:38:38.146 given the widely varying utility that consumers get from these kinds of discretionary purchases, I think it would be challenging to introduce a piece of technology like that into the domestic setting. 388 00:38:38.166 --> 00:38:39.207 I think I wouldn't say that it's 389 00:38:40.420 --> 00:38:42.421 that it's, that it's impossible. 390 00:38:42.762 --> 00:38:48.205 I'm just saying that it, it'll take longer than a lot of people think. 391 00:38:49.066 --> 00:38:51.448 And listen, like, I'm a huge robotics nerd. 392 00:38:51.488 --> 00:38:52.809 I was a big Star Trek fan. 393 00:38:53.549 --> 00:39:01.194 I, you know, I wish to be able to interact with a robot, like, you know, Commander Data from, from, from series from next generation. 394 00:39:01.214 --> 00:39:01.595 But like, 395 00:39:02.655 --> 00:39:07.818 I'm excited for that day, but I just feel like it's a little further out than what a lot of people hope. 396 00:39:08.118 --> 00:39:12.280 What are the constraints that are trying to be solved right now? 397 00:39:12.701 --> 00:39:21.045 Of the robotics automation companies that you guys have invested in at Lux, is there a common denominator of problems trying to be tackled? 398 00:39:21.065 --> 00:39:26.248 Or what are the modern problems in the proliferation of automation and robotics? 399 00:39:26.788 --> 00:39:35.958 It really comes down to the breadth of applications, the reliability, and being able to demonstrate the unit economics to customers. 400 00:39:35.978 --> 00:39:38.680 Because you're talking about a relatively nascent product. 401 00:39:39.021 --> 00:39:42.805 Like when somebody, when a customer buys, for example, a conveyor belt, 402 00:39:43.418 --> 00:39:51.900 or a package sorting machine, or a oven for their restaurants, there's very clear economics attached to that. 403 00:39:51.960 --> 00:39:59.642 If you have a coffee shop and you buy a coffee grinding machine for $5,000 or whatever it is, it's very clear as to what the return on that investment is. 404 00:40:00.062 --> 00:40:13.285 When you're talking about a new product, a new technology, with a sample set of 10 customers who've used this before, it's much more challenging to be able to justify that return on investment. 405 00:40:13.805 --> 00:40:33.640 So I think being able to demonstrate value, being able to create an opportunity for a customer to be able to properly underwrite is the challenge that a lot of these new wave of automation companies are facing right now, perhaps more so than just technology. 406 00:40:33.660 --> 00:40:34.361 It's being able to. 407 00:40:34.985 --> 00:40:38.086 quantify the value add for their customers. 408 00:40:59.633 --> 00:41:05.858 More than 2,000 podcast transcripts, 10,000 articles, and countless conversations with the people actually building this industry. 409 00:41:06.058 --> 00:41:09.421 And now we've structured all of that data into the Bankless MCP. 410 00:41:09.541 --> 00:41:17.047 So you can go and connect it to your Claude or ChatGPT, and suddenly your AI can answer your crypto queries with the entire Bankless archive behind it. 411 00:41:17.147 --> 00:41:23.252 And every new Bankless article or episode gets added automatically, so the context keeps staying up to date. 412 00:41:23.272 --> 00:41:27.175 The Bankless MCP is exclusively available to Bankless premium subscribers. 413 00:41:27.395 --> 00:41:31.496 So you can go to bankless.com, upgrade to premium, and connect the MCP in just a few minutes. 414 00:41:31.596 --> 00:41:36.518 And all of a sudden, your crypto queries to your AI, LLM, whatever you use, will get a thousand times better. 415 00:41:36.658 --> 00:41:37.298 So go check it out. 416 00:41:37.418 --> 00:41:38.599 There is a link in the show notes. 417 00:41:38.659 --> 00:41:42.840 And once you become a Bankless Premium member, you can hop into the Bankless Discord and let me know how you like it. 418 00:41:42.960 --> 00:41:43.641 Some exciting news. 419 00:41:43.721 --> 00:41:48.466 We are launching a new podcast to help people figure out the crypto cycle, how to navigate it. 420 00:41:48.546 --> 00:41:51.228 The best crypto cycle investor I know, his name is Michael Nato. 421 00:41:51.328 --> 00:41:52.369 He runs the DeFi Report. 422 00:41:52.430 --> 00:41:57.194 This is the guy that sent me a sell alert before the 1010 price drop happened. 423 00:41:57.374 --> 00:41:59.837 His cycle analysis has been absolutely on point. 424 00:41:59.917 --> 00:42:01.258 I've been following him for years. 425 00:42:01.419 --> 00:42:04.842 And this year we started recording weekly podcast episodes. 426 00:42:05.002 --> 00:42:12.768 Each one, we get into his portfolio, what he's holding, the market structure, entry targets, fair market value of Bitcoin and Ether, and where we are in the cycle. 427 00:42:12.868 --> 00:42:15.270 There's new episodes that are released every Wednesday. 428 00:42:15.330 --> 00:42:16.071 They're 30 minutes. 429 00:42:16.131 --> 00:42:16.591 They're short. 430 00:42:16.631 --> 00:42:17.272 They're punchy. 431 00:42:17.392 --> 00:42:19.774 I think this crypto cycle is harder to navigate than most. 432 00:42:20.014 --> 00:42:21.255 So let's do it together. 433 00:42:21.395 --> 00:42:22.596 Go subscribe to this podcast. 434 00:42:22.656 --> 00:42:25.258 Search The DeFi Report wherever you get your podcasts. 435 00:42:25.318 --> 00:42:28.381 YouTube, Apple, Spotify, or find the link in the show notes. 436 00:42:28.521 --> 00:42:30.082 There's a new episode waiting for you now. 437 00:42:30.342 --> 00:42:35.224 I want to learn a little bit more about the intersection of AI and robotics. 438 00:42:36.324 --> 00:42:37.745 We have AI now. 439 00:42:38.805 --> 00:42:42.246 We are now having people talk about robots coming into the home. 440 00:42:42.326 --> 00:42:47.948 And I think the naive, simple thing to do in your imagination is like, oh, we've got LLMs. 441 00:42:48.309 --> 00:42:49.529 Let's put them into the robot. 442 00:42:49.629 --> 00:42:51.810 And now we have smart, generalized robots now. 443 00:42:52.390 --> 00:42:55.273 While preparing for this interview, I've learned that that's not quite how it works. 444 00:42:56.114 --> 00:42:57.135 I wish it was that simple. 445 00:42:57.796 --> 00:43:09.249 Shaheen, can you explain that VLMs and VLAs and all the other details that we need to know about how AI actually becomes imbued in robots? 446 00:43:09.690 --> 00:43:11.211 So AI comes in many flavors. 447 00:43:12.112 --> 00:43:32.208 And if you look at the technology that's more commonplace today, when AI is being implemented in a robot on the field today, most of those robots use some combination of sensing with their cameras and radar and LIDAR or whatever it is. 448 00:43:32.849 --> 00:43:37.653 And then they have a processing chain that tries to perceive information. 449 00:43:38.233 --> 00:43:48.698 Their environments, which is segmenting the sky versus the ground, what the objects are, what object is movable, what object is not movable, what is a target object. 450 00:43:48.758 --> 00:43:53.540 They call that in general perception from what is being sensed from their sensors. 451 00:43:54.321 --> 00:43:55.561 And then there is planning. 452 00:43:55.681 --> 00:43:59.043 Okay, so how do I reach for that object? 453 00:44:00.103 --> 00:44:09.911 If you're a car, what path should I take to get to the other side of the intersection, for example, if there is construction going on on the other side of the street? 454 00:44:10.531 --> 00:44:25.843 And then there's the actual actuation, which is, okay, now I'm going to activate this motor, activate that motor to execute on that plan that I've generated from my perception of the environment that I have collected from my sensors. 455 00:44:25.863 --> 00:44:25.963 Okay. 456 00:44:26.563 --> 00:44:37.032 So that is the, I would say, common AI-based workflow that exists in robots in the field today. 457 00:44:37.492 --> 00:44:46.680 I'm not intimate with the technology behind the Waymo vehicles, but my guess is that the way they operate today is somewhere along these lines. 458 00:44:47.301 --> 00:44:53.867 Now, what we're seeing with VLAs and VLMs is some flavor of, 459 00:44:55.208 --> 00:45:11.439 of using language to interpret a scene and then generating language from that language should then take some kind of action. 460 00:45:12.160 --> 00:45:21.166 I am vastly, grossly oversimplifying this, and there's many people out there that can explain this better than I can. 461 00:45:22.621 --> 00:45:34.092 but the core of it is using language to interpret a scene and then using language to come up with some plan of action and then executing on that. 462 00:45:34.412 --> 00:45:45.182 And as you know, with us as individuals and animals, we don't necessarily talk through what we're seeing and we don't talk through what we're going to do. 463 00:45:45.943 --> 00:45:59.081 There's many other steps that come into play and many other sensations that come to play and many other pre-planned reflexes and heuristics that come into play. 464 00:45:59.922 --> 00:46:02.686 And many companies are trying to bake that. 465 00:46:03.247 --> 00:46:05.530 into their models. 466 00:46:06.071 --> 00:46:17.368 The jury is still out as to whether you can simply solve this problem with scale to just make these absolutely gigantic models that rely on language alone 467 00:46:18.029 --> 00:46:35.645 to perform these tasks versus the more quote-unquote traditional approach, which is this sensing, perception, planning, action process, which was popularized pre-LLMs. 468 00:46:36.766 --> 00:46:42.832 I think the jury is still out as to what's going to come together, but it's my expectation that it's going to be some kind of hybrid process 469 00:46:44.167 --> 00:46:45.888 of the two, if that makes sense. 470 00:46:46.248 --> 00:46:56.193 But I can suggest many people that you can bring onto your show that can give a pretty thorough lesson on how these VLAs and VLMs actually work. 471 00:46:56.513 --> 00:47:01.716 That would definitely challenge myself as an interviewer to go that far down the robotics and automation rabbit hole. 472 00:47:01.736 --> 00:47:03.657 But I do find it very interesting. 473 00:47:03.897 --> 00:47:06.018 I probably should have defined VLM and VLA. 474 00:47:06.298 --> 00:47:09.900 That's vision language model and vision language action model. 475 00:47:10.400 --> 00:47:21.704 Would you say, is it fair to say like we got LLMs on the cloud anthropic open AI side and we have VLMs and VLAs on the automation robotic side or is it just not that clean? 476 00:47:21.724 --> 00:47:25.146 LLMs are basically, you know, chatbots. 477 00:47:25.386 --> 00:47:33.369 And then VLMs and VLAs are basically interpreting a scene with language and then taking action. 478 00:47:34.549 --> 00:47:49.797 based on putting a set of observations and intentions through a model and generating a plan from that and then converting that plan into action. 479 00:47:49.897 --> 00:48:00.603 So for example, taking a picture of a scene from a sensor, okay, here's all the objects in the scene by putting it through a vision language model 480 00:48:03.744 --> 00:48:09.148 basically a chat bot to come up with some kind of plan of action based on the robot's goals. 481 00:48:09.708 --> 00:48:16.333 And then turning that language into like, you know, motor actuations to actually perform some kind of task. 482 00:48:16.833 --> 00:48:17.253 Okay. 483 00:48:17.293 --> 00:48:18.634 You know, you open the refrigerator. 484 00:48:18.714 --> 00:48:19.055 Okay. 485 00:48:19.115 --> 00:48:19.855 I want beer. 486 00:48:20.315 --> 00:48:20.896 Where is the beer? 487 00:48:20.956 --> 00:48:21.816 Oh, there is a beer. 488 00:48:22.397 --> 00:48:22.637 Okay. 489 00:48:22.677 --> 00:48:23.878 Now you have to pick up the beer. 490 00:48:24.438 --> 00:48:25.378 You go grab the beer. 491 00:48:25.919 --> 00:48:34.881 So it's been shown that these work, but they're still relatively nascent relative to the more traditional approach. 492 00:48:35.702 --> 00:48:38.643 And they may not be as quick and they may not be as reliable. 493 00:48:39.263 --> 00:48:43.264 And so, again, the jury is still out as to how you can make them more reliable. 494 00:48:43.304 --> 00:48:45.865 Do you just continue to refine them and make them larger? 495 00:48:46.705 --> 00:48:50.546 Um, or do you kind of hybridize them with these more traditional approaches? 496 00:48:50.646 --> 00:48:56.908 And I'm not a roboticist myself, but I think it would be a great idea for your next guest on the show to talk about that kind of stuff. 497 00:48:57.468 --> 00:49:11.992 It seems like it's an important ingredient nonetheless to add to the generalizability, uh, and, um, just the practicality of automation and robotics to fit into more spots in the world. 498 00:49:12.112 --> 00:49:12.212 Yeah. 499 00:49:12.232 --> 00:49:12.332 Yeah. 500 00:49:12.792 --> 00:49:18.453 Because it seems like you can take this VLM or VLA and apply it to a robot in different settings. 501 00:49:18.953 --> 00:49:20.334 And it just kind of works. 502 00:49:20.594 --> 00:49:21.014 Is that right? 503 00:49:21.294 --> 00:49:30.296 So the thought process behind these types of language models is that, yes, they're more generalizable. 504 00:49:30.916 --> 00:49:34.517 You can teach them to do things by simply just showing them. 505 00:49:36.105 --> 00:49:39.587 the task the same way you would show a child a task. 506 00:49:39.787 --> 00:49:42.048 That's, that's the thesis behind them. 507 00:49:42.728 --> 00:49:43.969 And I'm really excited about them. 508 00:49:44.869 --> 00:49:50.752 I'm optimistic that over time, we'll figure out how to make them faster, more reliable, easier to train. 509 00:49:51.673 --> 00:50:02.178 Cause right now, you know, again, you need a team of hundreds of engineers to teach a Waymo, for example, how to go from A to B safely and, 510 00:50:03.071 --> 00:50:05.797 Is there a future where you can do the same with a VLA? 511 00:50:05.977 --> 00:50:08.081 You know, perhaps. 512 00:50:08.422 --> 00:50:10.506 But the question becomes, how much does it need to be trained 513 00:50:11.759 --> 00:50:13.720 to get to what level of reliability. 514 00:50:13.780 --> 00:50:17.824 Like, for example, I'm trying to teach my four-year-old how to write numbers. 515 00:50:18.864 --> 00:50:23.188 And, you know, some numbers she can write, like, after I show her twice. 516 00:50:23.929 --> 00:50:30.794 Other numbers, for whatever reason, she has a hard time, like, stopping when she's doing a curve. 517 00:50:30.814 --> 00:50:32.015 Like, for example, write the number two. 518 00:50:32.075 --> 00:50:34.898 You have to curve and then stop and do a straight line. 519 00:50:34.918 --> 00:50:35.078 Right. 520 00:50:36.190 --> 00:50:37.191 That's the challenge for her. 521 00:50:37.291 --> 00:50:45.741 So maybe that challenge is limited to human children, or maybe it's a limitation associated with neural networks. 522 00:50:45.761 --> 00:50:47.022 Who knows? 523 00:50:47.082 --> 00:50:49.024 So that will be figured out in the near future. 524 00:50:49.125 --> 00:50:57.214 I would also imagine the chatbots had this very incredible advantage in that they just had to train on all the data of the internet, which was accessible to them. 525 00:50:57.774 --> 00:51:09.482 I would imagine that robots and automation don't nearly have the same qualitative and quantitative amount of data to train how to move your arm to grab the thing with the right amount of force. 526 00:51:09.582 --> 00:51:16.166 So we are investors in XDOF, which is specializing in generating this training data for robotics. 527 00:51:16.286 --> 00:51:18.328 Physical intelligence obviously also has 528 00:51:18.988 --> 00:51:25.235 a huge capability around amassing this data internally for training as robots. 529 00:51:25.255 --> 00:51:35.244 So yes, you're hitting on a very good point, which is companies that are able to access and build these training sets for these particular applications will certainly be advantaged. 530 00:51:35.645 --> 00:51:38.508 Shaheen, this has been very exciting and very educational. 531 00:51:40.108 --> 00:51:47.270 how are you hoping that automation and robotics positively impacts your actual personal life? 532 00:51:47.290 --> 00:51:52.631 So your day-to-day changes and your house gets an upgrade, your car gets an upgrade. 533 00:51:52.851 --> 00:51:59.612 Is there anything that you're trying to get your hands on as soon as possible to have a material improvement in just your day-to-day life? 534 00:51:59.992 --> 00:52:04.673 I'd like to have an autonomous car that is always available to me 535 00:52:05.502 --> 00:52:07.223 rather than having to call a rideshare. 536 00:52:07.283 --> 00:52:15.665 But the reality is that you kind of have that today with Waymo, and you sort of have that today with Tesla's FSD. 537 00:52:16.485 --> 00:52:23.107 But I wouldn't mind taking it a step further where the vehicle could figure out, you know, where to park and go off on its own. 538 00:52:23.828 --> 00:52:25.028 And for me not to have to deal with it. 539 00:52:25.048 --> 00:52:27.549 I think that for me would, and I'm just a big car enthusiast. 540 00:52:27.589 --> 00:52:28.209 I'm a car nerd. 541 00:52:28.949 --> 00:52:36.153 So having a car like that would be to own is something that I personally find fascinating and exciting. 542 00:52:36.514 --> 00:52:39.475 What about how you think automation actually enters the home? 543 00:52:40.236 --> 00:52:42.137 Not the humanoid robots. 544 00:52:42.477 --> 00:52:43.598 Yeah, yeah, yeah, yeah. 545 00:52:43.758 --> 00:52:46.440 So if it was just me, David... 546 00:52:48.061 --> 00:53:08.875 I would totally nerd out on having even a modestly capable robot, you know, in my home that could do simple things like, you know, turn the stove off, you know, or pour me a glass of water and bring it over, you know, just for the purpose of, you know, nerding out on something like this. 547 00:53:09.075 --> 00:53:11.337 I would find that just personally exciting, but. 548 00:53:11.721 --> 00:53:26.411 you know, being, you know, married and having two small kids, I just think it'll be, it's extremely unlikely that my wife would allow something like that in the house, um, until it's proven to be, you know, safe and, and never trip over a child or anything like that. 549 00:53:26.491 --> 00:53:34.757 So, um, you know, just for myself, just having a robot around that I can physically interact with, it would be a huge novelty and an interest. 550 00:53:35.082 --> 00:53:48.587 Of the companies that you guys have invested in at Lux, if listeners wanted to just go a little bit deeper about what we've been talking about today, are there any good companies that are doing something exciting that also provide an educational opportunity to just learn more? 551 00:53:48.607 --> 00:53:51.888 Any companies out there that are- They should absolutely learn more about physical intelligence. 552 00:53:52.548 --> 00:53:53.808 They should check out the company. 553 00:53:53.888 --> 00:53:57.249 They should check out the models that they have available out there. 554 00:53:57.770 --> 00:54:04.011 They should also learn about Formic, which is deploying robots in real manufacturing logistics settings. 555 00:54:04.031 --> 00:54:06.092 They have hundreds of deployments across the country. 556 00:54:06.172 --> 00:54:10.453 Most of their customers had no automation before automating with Formic. 557 00:54:11.318 --> 00:54:15.680 And their goal is to be the largest employer of robots globally. 558 00:54:16.200 --> 00:54:22.562 We'd like to say the equivalent of U.S. robotics will be of iRobot, but not evil. 559 00:54:23.262 --> 00:54:27.264 And so, you know, those two companies, they should absolutely check out. 560 00:54:27.564 --> 00:54:29.464 Shaheen, thanks so much for coming on the show today. 561 00:54:29.504 --> 00:54:29.824 Loved it. 562 00:54:30.005 --> 00:54:30.505 Loved it, David. 563 00:54:30.525 --> 00:54:31.125 Thanks for having me.