Building Robots That Work for People: Amazon’s Playbook for Scaled Automation

Transcript

Tye Brady (00:01)

You put people to center the robotics universe always, right? Always you put people at the center, like your employees, your frontline employees, the women and men who designed these systems. make it for people. It's not people against machines. It's people and machines working together. And the onus is on us in order to make it responsible, accountable, and ethical for human consumption.

Brian Heater (00:26)

Hey, everyone. Welcome to yet another episode of automated. My name is Brian Heater. I am the managing editor at a three. This week we have a great conversation with Amazon's chief technologist and generally affable gentleman, Ty Brady. We basically go down to his resume and discuss how he helped Amazon amass a one million robot army. We talk jobs, humanoid, AGI, moonshots, spaceships, Star Wars and all that fun stuff.

If you're enjoying the show, please like, subscribe, maybe tell a few friends. Now sit back, relax and watch me and Ty Brady talk for about 45 minutes or so. So I have to say Chief Technologist is a pretty good title, but for my money, it's tough to principal spacecraft engineer.

Tye Brady (01:26)

How about just geeks anonymous? No, I'm not sure about that. I'll tell you. ? I love verbatises. I do love verbatises. I like technologists. ? I like all things nerdy. So whatever title it is, as long as I'm building something, I feel like I'm pretty good.

Brian Heater (01:47)

Obviously, undergrad, you studied aerospace, you entered that world, you stayed in that world for a really long time. You saw yourself really getting into and staying in aerospace and space exploration.

Tye Brady (02:01)

Well, I'll tell you how it all started. Here we go, Brian. Here's how it all started. All right. Let's go. It's in the 70s, late 70s. I loved Star Wars as a kid. Like space and robots was my jam. It was like, isn't that amazing? Just the power of storytelling. I'll share another story. So I love R2D2. R2D2 is definitely my bot. But I'm going to share another very personal story is that

Brian Heater (02:13)

Very familiar story. Literally everybody in this industry has.

Tye Brady (02:30)

When just at that moment in 1977, that moment when Star Wars came out, I'm like, that's really cool. ? My family's not really the tech family at all. Is that I had one teacher, one single teacher who said to our class, our fifth grade class, is anyone interested in computers? I just saw Star Wars. I'm like, hey, that seems pretty cool. I'm interested. So I raised my hand and another kid raises his hand and it turns out this teacher was on the the board for a technical high school in the county. I grew up in a kind of a one stoplight town and this technical high school, had a deck PDP eight and I didn't know anything about computers and I thought he surely knew about computers. He's asking about it and we show up there because it's like the seventies. We show up there ? in his car and there's a computer boxed that hasn't been used in six months and so we looked to our teacher and like, hey, what's going on? He was like, I don't know, let's bust out the manual and start to figure this stuff out. So over the next three months, I kind of figured out how to do a little bit of programming, how to write my name. And I was hooked. I love computers. And then the 80s comes around and we have the personal computer revolution. I'm like, oh my gosh, I can actually, I had a TI-994A. Shout out to all my homies out there that know that that's the first 16-bit computer.

Really learned to program. fell in love with computing, all things software. I could go to any local arcade and kind of rip off that game maniacally and program at my house. I love that. Then I started doing radio control airplanes as well in the early 80s. And I said, wouldn't it be an amazing, awesome world if I can combine my love for radio control and computers all into one thing.

And lo and behold, that's the field of robotics. it's one person. Just one person can make a huge difference in someone's life.

Brian Heater (04:34)

It's wild, you know, especially now looking at how competitive things are, like, sometimes it's just about being the right place at the right time and literally being the one of two people raising your hand. And that's all it takes.

Tye Brady (04:47)

Yeah, that's all it takes. And there's definitely just some luck involved. There's a passion and a drive for sure. I came to Boston, sight unseen. This was the hub of the computer universe. Like digital was here. Apollo was here. Wang was here. Mercury Computer was here. It was all in Boston. And I wanted to be a part of that. I just wanted to be a part of that. Now comes the miracle number two is that I couldn't afford college. the goodness of some anonymous donors.

Folks that were given back to Boston universities allowed me to go to school on a scholarship. And I came to Boston knowing how to program with, and I loved aerospace. I wanted to put robots in space. That was just kind of my dream. And I got into aerospace and it turns out that ? combining ? my skills in software and my love of computers with more of a mechanical degree actually was really helpful in the realm of robotics.

So I got involved actually with the lab at BU, because one of the professors said, hey, you're pretty good at programming. Jeez, did you really do this yourself? I'm like, yeah, I love this. And so he invited me part of the lab that he was running. It was a graduate student lab. was only undergraduate in it. And in that lab, these are some legends. They were working on Voyager, like Veeja, the Voyager. They did a Faraday Cup for Voyager. And it was incredible just to be around these luminaries, these icons of space travel. Like it was just, you know, your heart actually beats a little bit faster. And there's two professors at MIT and they kind of, they knew how much I love to build and how much of the passion I had for robotics and for space. And so they invited me when I graduated to be a full-time employee at MIT. And that was a dream come true.

Brian Heater (06:38)

There is this special kind of, I think, fearlessness that you have when you're younger and when you don't know how much you don't know when you're an undergrad. if you're like, sure, like, of course I'm going, I'll join these like graduate students in these legends. And yeah, no problem. I can absolutely compete with these folks.

Tye Brady (06:58)

Yeah, it's funny. Compete's actually not the word. It's just like, could I help build? Could I be a part of something amazing? And my time at MIT, and by the way, I applied to MIT when I undergrad, going to look for schools, and summarily rejected. Like, my gosh, I know computers so well and rejected. So then when I went to MIT as a staff research engineer there, with that awesome title that I grew into. I learned to be a builder and I had some incredible mentors around me, like people that really cared, people that would give me, give time to me and share their craft and share, you know, actually what's, how to build things from the ground up, how to work with a team, how to do some really cutting edge technologies. And that was transformational in my life. It really gave me the confidence.

That what I was doing could actually make a giant difference. We built spacecraft. We built a spacecraft in the mid 90s called HETI, where we did, you know, talk about one hat or two hats. I probably wore 10 different hats, whether it's hardware, software, thermal, dynamic systems, astrophysics, like you name it. We were all kind of in on it to build this spacecraft. And what it really gave me is it just gave me confidence and it's still with me. I will tell you, it's still with me, the generosity of people and the kindness that one can offer, particularly around education, of what an impact that can be for any young girl, young boy out there with their own aspirations. know, mean, Star Wars is 1977, a new hope. How appropriately named, a new hope.

Brian Heater (08:44)

So you've really got a crash course. mean, you know, not only were you there and not only were you getting your masters, but you were also an employee at the time. So you were working with the people who are really doing the work.

Tye Brady (08:56)

Yeah, not only that, I've seen some really incredible things in my career. The birth of the personal computer, like I said, that was a really big deal. But when I came in 1990 to MIT, the internet literally just started. I know it's weird and hard to think about, but the internet just began. So now, we can really share our learnings and our love of computing and building things and building these robotic systems with each other. We can share software, can share techniques, and it was a lot easier than just one-on-one type of interactions. It was a many-to-one and a one-to-many type of interaction. And to be at a place like MIT at the birth of internet was really, really incredible. ? So that had a profound impact as far as my learning would go. by the way, you're at MIT, right? So a lot of folks would...

You meet different universities and folks, industry folks from the outside. And it was really an incredible time. what I learned most and probably the deepest thing that I learned is just to be good to yourself and have confidence. I took a class every semester I was there. You can take classes for free. And finally, a professor said, didn't I see you a couple of years ago? I'm like, yeah. goes, how many of graduated yet? I'm like, no, I'm just taking classes because I love it. He goes, listen, how many classes have you taken? I think I said 10.

He goes, why don't we just get you a degree? So I'm a trickulator in 99, got my masters in.

Yeah, it was kind of funny.

Brian Heater (10:32)

Yeah, it's really interesting hearing you talk about that process because looking at it from the outside, you know, taking a kind of ? bird's eye view of your CV. I was going to ask you whether the plan was to kind of stick around academics, but now it seems pretty clear to me that that wasn't necessarily that you just wanted to be in and around rockets in and around airspace. You weren't you weren't trying to become like a tenured professor. ?

Tye Brady (10:59)

No, I really would. And I still have this desire and love. I still just want to build things. Kind of a startup mentality, even an insurgent mindset at times of, how can we really use technology for the benefit of people? How can we build things that help people do their jobs better? How can we build things to make their workplace safer? That's been with me. It's still with me. It is funny. we build machines and really index and I have I'm super proud of the technologies like it's some a lot of some worlds first actually I'm really proud of that but it's you know when I reflect back it's it's always about the people I can I can share with you the heartbreaks and I can share with you the the accomplishments and you know how we picked each other up when we needed it like I'd like to say that kindness is a parachute for failure but it's a force multiplier for success and when people are kind to each other especially in times of failure, you can fall back on each other and it softens that blow. And we all have failures. We all should learn from our failures, but that doesn't mean it doesn't hurt. It hurts. But when you surround yourself with kind people and you actively make kindness an action word, ? it makes a huge difference, right? The successes are even higher and the lows are a lot softer.

Brian Heater (12:23)

Yeah, and the lesson that I learned early on in covering robotics, and this was actually, I think, around the first TechCrunch robotics event at MIT that you were at, I learned how, at that time, and even, you know, obviously to a slightly lesser extent, but even now, how small the industry is, how in some ways kind of insular it is in that you really are working with the same people over and over again. And that you really do end up connecting with those people that you went to school with or that taught you at MIT 20, 30, 40 years ago. So just yet another reason to be kind to people.

Tye Brady (13:02)

Yeah, I mean, how amazing is that though? Like I know you had Mark Rayburn on and I knew Mark back in the 90s and he's an amazing person and know, ? Colin Angle's part of the community as well and Cynthia Brazil and Russ T. Drake and like these are all, they're friends. These are friends. We're really close to that community and I think that we see this incredible technology in physical AI that we're developing.

And that there's so much growth that can happen in that. We don't compete. That's where I got a little allergic to that word. We don't really compete with each other. What we do is we try to lift each other up because it can have such a profound impact on people's lives in making the workplace better and making the home better and allowing us to... Actually, if you want to get super meta, I'm in this game because it allows us to be more human. It allows us to connect with each other even more closely.

Right? Our technology allows that. Of course, you have to be mindful of how you build that technology so that people use it in the right ways and with the right protections and responsible and accountable for sure. But the goal, I think the goal that we all have is to allow us to be more intelligent, more capable, more human, more connected to one another.

Brian Heater (14:17)

There's competing and competing isn't necessarily like always a bad word. There is healthy competition, but the nature of that competition and we're kind of getting to this bit of a transition period for you as we're again going down your CV. ? The nature of competition does tend to start to change a little bit as you move into the commercial space.

Tye Brady (14:41)

It does. mean, I was a competitive cyclist for almost 20 years and it is funny that, you know, I mean, I definitely know what competition can do. And I think that there's even something when you compete against somebody that you both have shared kind of the trials and tribulations, you both have gone through that pain. I think that there's mutual respect there.

Brian Heater (15:03)

shared -- a very unique experience that only the two or very few of you have experienced.

Tye Brady (15:08)

That's right. Yeah. I think that's absolutely right. And especially when you compete with a team, like your teammates are part of your journey. Like I have from my cycling team back in the 90s and the early 2000s, we're still lifelong friends. my kids hang out with their kids and, you know, I mean, there's a real bond there. So I think that bond, to bring it back to robotics, is there in the robotics community. I mean, we...know each other for sure. We couldn't believe when Amazon actually bought Kiva Systems in 2012. It was not that long ago, but in 2012, $785 million. I can tell you every single robotist has said, oh my gosh, oh my goodness, are you kidding me? A robotics company actually sold for something with the units of millions? Right? I mean, it was kind of unheard of. But boy, Amazon knew what they were doing. Jeff knew what he was doing there because it was really not the robotics. It was the idea of a goods to person fulfillment system that he saw. And from that investment, here we are with Amazon Robotics, the world's largest manufacturer of industrial mobile robots with more than a million robots out there that we've manufactured in Massachusetts.

Brian Heater (16:19)

I'm really trying to wrap my brain around the nature of how time works in and around the robotics field because, yeah, 2012, especially in relation to that story, does feel like an incredibly long amount of time ago.

Tye Brady (16:34)

That's right. I joined in 2015, so I actually just had my 10-year anniversary inside of Amazon. It's been an incredible journey from just a couple of buildings where our systems were operating at to now we have more than 300 robotics-enabled facilities. are in not only world-class leaders in mobility, but also in manipulation and in storage and sortation and perception as well. I'm really proud of the work that many, many women and men that have done ? truly are pioneers.

doing their technology development through project context. And our context is very straightforward. Like we deliver goods, we want to have the world's largest selection of goods at a low cost delivered right to the door. And that context, that framing allows us to accelerate technology development in just the most basic of ways.

Brian Heater (17:22)

Prior to Amazon, you spent a number of years at Draper Laboratory. And it strikes me that ? in a sense, and I don't know if this is deliberate, probably not, but that really was kind of a bit of a transitional space between pure research and really starting to kind of commercialize or at least really productize technology.

Tye Brady (17:46)

Indeed, Brian. And I love that you've done your homework there. There was a little time in between that's kind of actually funny. I did graduate from MIT and it happened to be dot com. It was 1999 and 2000. if you wanted to do OK, if you wanted to maybe make a little bit of scratch in dot com, you really need to have two attributes. The first was, did you have a degree from MIT? And I did. I'm like, OK, I have a child on the way. I'm going to need a house. So I got one of those.

And the second attribute was, were you breathing? And I was breathing. So I did have a little bit of consulting time in between. Spacecast. That's right. That's right. Very good. Then ? dotcom starts to wane, let's say. ? And I joined Draper Laboratory. It was a great experience for me. But there was, ? if the prior decade was about building, this was about leadership.

Brian Heater (18:27)

Solutions, Yeah.

Tye Brady (18:45)

I led teams to do some really cutting edge novel work. flew the first MEMS gyro integrated into a star camera, the first CMOS star camera where we put those together. Got the patent for that, flew it in space. It was a tremendous success. First time the world's ever seen that. And then we had a team really focus on lunar landing, precision lunar landing, the ability to land anywhere on the moon to whether inside of 100 meters under any lightning condition.

And we did a lot of fun stuff, inclusive of doing the really one of the first vertical takeoff, vertical landing rockets way before, you know, what you see out there today. This is in 2010, 2011, where we'd have a rocket kind of take off and land, mimicking how it would land on the moon. Some really great stuff that we did using terrain relative navigation and actually using the same software that you would actually use on the moon. I led a team to do the autonomy, guidance, navigation, and control on that system, working really closely with NASA. And while we never got to the moon while I was at Draper, a friend of mine, a coworker inside of NASA, Tim Crane, ? or Dr. Tim Crane, I should say, ? he started Intuitive Machines and was the first commercial company to actually land on the moon using a lot of the techniques and the technologies that we discovered the decade prior. So I'm really proud of that as well. But it's really a time of leadership and you know how to lead teams and how to ? lead technology in a way that the world hasn't seen. And I really did enjoy my time there. In 2015, I was approached by Amazon saying, really want to expand our portfolio. We want to just be more than just ? Kiva. We want to get into manipulation. want to get into different sorts of ? systems that can store goods and sort goods. ? And I love that. I love kind of that mindset of invent and simplify and think big and learn and be curious. That's my mindset. And those are also leadership principles inside of Amazon. So I knew it was a fit and joined Amazon. And, you know, here we go. So quite a quite a resume of robots. I also co-founded Master of robotics in between.

I forgot to tell you that.

Brian Heater (21:07)

Right before I do as an aside. I had I had to ask, you know, like what what that conversation was like that got you to leave, you know, not not Draper necessarily, but this this very specific career path, this this trajectory, this, you know, this idea that you had had in your head since you were a kid and saw Star Wars and wanted to build rockets to, you know, again, primarily focusing on robots for warehouse and logistics and fulfillment centers. That's quite a shift.

Tye Brady (21:45)

It does seem that way. And I will tell you that, you know, in my career, I've been I've been very blessed, like I said, with a great teammates and a great support system. And I was, of course, approached by some of the big tech companies during that time. And I'm like, hey, I'm happy what I'm doing. I'm good. I like what we're doing. But when Amazon came, it was a little bit different. And I'll tell you why. My journey is kind of from rockets to robots, for sure.

But when Amazon came to town and we had the initial conversation, it was about application, right? It wasn't just in the digital world. It was about true application. And if it's something that I learned at MIT was that that forcing function of constraints and making something work in the real world. And also I had a giant advantage of now this actually ties e-commerce and aerospace together. Is that when you're at the scale of Amazon, and you ship literally billions and billions of packages, like literally, then precision matters. Like you can't be, you know, nine out of 10 good. You have to be 99.99. And that's like what we think in aerospace terms. High reliability, high quality. So I knew I had an advantage there, but I was really impressed with the fact that they have purpose. They have purpose, right?

And if you're listening, if you're listening to this podcast, no matter what field that you're in, you need three things. And I credit Pam Reeves for really sharing this with me. You need purpose, need strategy, and you need operations. Purpose, strategy, and operations, right? The purpose for Amazon is really straightforward. Have the world's largest selection of goods at a low cost delivered right to the customer's door. That's our purpose.

And that has shown that that works. And when you do it well, you spin the flywheel, as Jeff calls it, and it continues to grow jobs and grow opportunity. Our strategy is actually really straightforward. Our strategy, of course, is in our leadership principles that Amazon has. But you even hear Andy talk about it. We tend to operate like the world's largest startup. And I like that. I like that scrappy mentality. I like the experimentalist mentality. I like the let's go build ? type of thing, ? build things mentality.

And in robotics, one of the things that is very obvious, I think is obvious, is that what we're trying to do is have a world-class capability in the basics of robotics. Not the hype of robotics, but the basics of robotics. Can we move things? Can we manipulate things? Can we store things? Can we sort things? Can we perceive the world around us? And can we pack things? Just the basics. You might call them primitives.

In order to really achieve a world class capability in this, you need that project context, the purpose that we have. So that's our strategy. We want to have a world class capability in just the basics. ? And if we can eliminate the menial, the mundane, and the repetitive with the basics, that's going to be better for our employees. That's going to be better for people. So now our operations turns into how can we enable people with the best tool set possible out there? How can we create a safer environment for our folks with robotics? And how can we have our folks be more productive using that tool set? That is a really hard problem. The onus isn't on the operator. The onus is on us, the roboticists, in order to make our tools simple, not complex, in order to make our tools actually have utility and to have the high reliability in the real world. That gains us a tremendous advantage.

Because when we do that, that spins that flywheel back to we're getting better at the basics. Now, if you want that, that was a long way to get to the where, why I get really excited and why I even took this job. You have to be able to think all the way through that, that, that life cycle. And I do believe that what Amazon is doing in robotics is truly pioneering work. And I do believe that by mastering the fundamentals of robotics through application will allow us not only to transform e-commerce is what we have seen. We've created hundreds of thousands of jobs. We've created a safer environment for our employees. have 1.5 million plus employees out there that we use our robotics for. But I believe that the fundamentals of robotics will actually start to transcend different sectors, whether it's in agriculture, or it's in fishing, or it's in automotive, or it's in healthcare. That the work that we're doing, this collaborative robotics philosophy that we have.

we actually start to impact these other areas. And when we do that, that's actually creating a very positive change for the world. It's allowing us to be more intelligent, more capable, and more human to each other, and still with the greatest value of human-to-human connection, with the greatest value of a person having a great job in a safe environment and allowing them to do productive things. That excites me. And there's no other company in the world that can do that.

Brian Heater (26:57)

But you do understand why people always question whether the technology can and will make human jobs less human and whether they will displace jobs. I these are very real concerns. how do you have those conversations with people and how has that conversation evolved over time?

Tye Brady (27:13)

Amen.

You have to put your you have to put your money where your mouth is. First of all, like I'm really proud that Amazon has committed more than one point two billion dollars in upscaling more than 70000 employees. You have to be a responsible company will put money into upscaling because the workforce will change. There's no doubt about that. Jobs will.

Brian Heater (27:44)

Let's break that down real quick. up upskilling, obviously, you know, it varies from person to person from place to place from job to job. like, what's what's a concrete example of upskilling in the real?

Tye Brady (27:57)

Okay, so first of all, I just say that no other company in the history of the United States for sure has created more jobs than Amazon. Like just created jobs through their productivity gains and by treating our employees with the respect that they deserve and giving them the tool sets through robotics that the world has never seen. So let me give you a concrete example. So one is that we have our Career Choice Program where any of our employees from day one when they start that we will bring the classroom actually right to a fulfillment center. And if they're interested in science and tech or actually just about any field, we'll pay for that schooling. It's very near and dear to my heart. If the employee wants to pursue it, we'll do that for them. I like that. Another one that actually is very close to us is our robotic apprenticeship program where...

If somebody is working in on the front lines and they say, hey, that's a really cool proteus robot that we have over there. I love what you're doing with Hercules or that sparrow. It's amazing. We will teach folks actually how to be if they're interested in being an apprentice, we will give them an apprenticeship program and they go through the apprenticeship program. They'll get about 40 % more pay and they're doing something that they love. So we're seeing a real big uptick in that. Like these are concrete real world examples what employers can do in order to to help. ?

That's with our employees. We also have to have a mindset towards our future generation, right? So our Amazon future engineers, we put millions of, 50 million plus into scholarships for high school kids to go to college. I love that. That's very relatable to me. We're sponsor of mass robotics. I know that would be a giant shocker, but I'm really proud of that, right? Because it's the world's largest nonprofit for incubation of robotic startups in the world.

Like that's really, really incredible.

Brian Heater (29:47)

You mentioned ? being the world's largest startup and it strikes me that there's a few ways in which that's kind of manifesting itself. ? Obviously there's a lot of the of the day one research that's happening, ? the stuff that's happening behind the scenes. There's the industrial innovation fund, which is actually like funding a lot of startups. ? then something that really just sort of came to light recently that I think is probably close to maybe some of the work at Draper and close to maybe closer to what TRI is doing, maybe close to what Mark is doing over at RAI is ? the work at Amazon Frontier AI and Robotics, Amazon FAR, which like that's a direct product of your hiring Peter Abil and the rest of the variant team.

Tye Brady (30:37)

Yeah, yeah, yeah, it's a super some I don't know if you've seen it's a really, really exciting early stage research that that we're doing over there. I think it's like, you know, while we are applied, absolutely, we are all about application. But it's also, I think a good thing to think about moonshots and like, what what is the art of possible? Look, we are experimental set heart. I have my own lab and we do really cool things to see what is possible. What's cutting edge? What's hype? What's not hype?

It's kind of the fun part of robotics for sure. We're learning. We're totally learning. And we want to build a Pruso concept. And the whole point is, can we ultimately build a tool that would assist our people on the front line? Where it's going, I couldn't tell you. It's too early stage, right? But ? it's definitely, I'd say, really exciting.

Brian Heater (31:29)

So, you know, we have to get to this at some point in the scale of what's hype and what isn't hype, where does bipedal humanoid robots fall?

Tye Brady (31:44)

So there's, let me start with a quote first. It's Roy Amaro. He's a technologist from ? SRI and he has a quote that we tend to overestimate the benefits of technology in the short run and underestimate the benefits of technology in the long run. And I think that there's something there. I really do. Right? you know, are we peak hype right now? I'm not sure. I think we're definitely in the overestimation part. Like I can tell you it's really easy to make a YouTube video.

It is. I can cut, I can edit, we can make something. ? It's really hard to have your robots operational every day, shipping billions of packages and creating a safer environment for our employees. that's what we're not, that's only what we can do. That's what we are doing. ? And that high reliability, real world applications actually, what I think is driving the field. Don't get me wrong. I love for...

especially the next generation to get inspired about the potential of robotics and what robotics can do. I love that for sure. But when it comes to bipedal humanoids, which is what your question, like, absolutely. Am I interested in the bipedal nature of mobility? Of course. It has two legs or four legs or six legs or one or it's wheeled. Like we're interested in anything that moves for sure. Right. Like we want to understand what, when's there a use case?

what's the problem that we're trying to solve with robotics and it's good to have different modalities in your toolbox. So the bipedal part is totally interesting. ? I think it's just, again, early stages.

Brian Heater (33:25)

Yeah, yeah. So if I'm understanding, you know, it's that these will be useful, but right now it's just a very practical problem. And this is speaking as somebody, again, who is involved in helping ? scale and deploy a million robots, is actually like getting these things and making them reliable and getting them out in large numbers into the real world.

Tye Brady (33:50)

Yeah, I think the foundation models are like really incredible. So again, I want to get specific with my language here. Like there's large language models, right, that uses language. I almost call it conversational AI, right? So it's language in, language out. ? Summarize this document for me, merge these two things. Yeah, you know, that's a language. Like that's ? conversational. Okay. In robotics, you're not teaching it

Brian Heater (34:08)

ChatGPT.

Tye Brady (34:18)

through language. mean, you can use language in order to task it, but it's really learning through prior data and prior actions in the unstructured world, right? Or even in the structured world. So we're building up that foundation model. There's another, you know, while we're just on AI. So I actually put it in three buckets, if you're interested. still doubt it? the first bucket is conversational is what it said. The second one is actually creative AI.

Brian Heater (34:39)

I love to hear about that.

Tye Brady (34:47)

And this is AI that allows a person to be more creative with their own ideas and the assistant is helping them. Right. So, for example, inside of Amazon, we can ask a generative AI system to lay out a building for us, given this amount of input, given this amount of output, given the economics that are around this area. What size building and how should that building be laid out? Sure, we can. And what we'll find like.

almost exclusively is that that building can't actually be built. That's missing something. what it will do is it'll inspire the actual solutions designers, the architects are actually laying out the buildings and go, ? hey, there's a choke point over there. I didn't really realize why that's working so well. It allows a person to be more creative in their job. then it's human output given AI input, right?

I actually even have a side story. I do like to paint every now and then. I'm a terrible painter, ? but I enjoy it. I like to kind of find the zen. ? I have a Boston Terrier and I asked a generator to say, you generate for me a Boston Terrier in a kayak, ? thinking pensively about the starry night ? in ? the style of Vincent van Gogh? ?

Right. And voila, out comes this prompt and I see about five different images. And what it did for me is it's like, I didn't really think about that pose or I didn't think about even that. It's something as crazy as what I just just told you. But it invoked a creative side of me. And I decided to take one of the prompts and one of the outputs from the prompt and paint it myself. Right. So that's my own creative. OK, it can inspire you. The third bucket is operational.

Right, so the operational AI is when it's embedded actually into the system using generative AI in order to make the task more productive or create a safer environment. great example of that is our deep fleet where we have, again, 10 plus years of an algorithm in order to optimize the movement of two or three or 4,000 robots at a time inside one of our facilities. And deep fleet within a year using generative AI end to end allows us to do to be about 10 % more efficient in the movement of those drives. Like real, like, whoa. And now it's operational, right? Our generative AI, in order to help our damage detection algorithms, we generate packages that have dents on it and scratches and marks, and we use that as a test. Our generative AI, in order to optimize the trajectories of our robotic arms, like we're using that now. So many, many examples of generative AI that are really changing the game for us. ? If you can get to, and they're not, no, no.

None of the three buckets are better than the others. They just have different purposes and functions, right? I think you should be at least in one or two or three. If you're doing anything productive, you should probably in one, two or three.

Brian Heater (37:51)

Yeah, and but that's also where it's important to, again, start setting realistic realistic expectations, you know, from somebody who doesn't understand the nature of how these things work. Well, you know, it can write a story. So, of course, you know, we're close to, you know, AGI, you know, of course, it can tell a robot how to do this. And then, you know, that that compounded with, as you said, the the ease with which somebody can create a flashy YouTube video.

I feel like we're maybe not doing ourselves any favors as far as setting those realistic expectations. At Amazon, as a company who again has a very strong investment in real world robots doing real world things, what can you do to help temper expectations?

Tye Brady (38:39)

Well, first it's a mindset and I've had this mindset for quite a while. I'm going to just quote a book by Jim Collins. It's called Built to Last. If you haven't read that book, it's a fascinating read. It's about how companies can survive for 50 or more years and what are the common characteristics in those companies, right, for longevity. And there's a really great...mental model that is put forward. The quote goes, be the clockmaker, not the time teller. So be the clockmaker, not the time teller. Meaning, get into it. Dive head first into what it is that you are doing, not just use it as a service, but at least in robotics, that's what we're doing. We're getting knee deep, arm deep.

completely immersed in building responsible, accountable, and ethical physical AI systems in order to help people be more productive. And it works. Like, for example, our Shreveport building that we just announced this past spring, 2,500 new jobs, brand new jobs down there, 10 times the amount of robotics we've ever done. We can process orders 25 % faster in that building. There's 30% more demand for more skilled workers. Our apprenticeship program is taking off there. This works. It's creating jobs. It's allowing people to focus on what matters at work. And it's really changing the game when it comes to our customer obsession in order to get the right good at the right time, right to the customer's door.

Brian Heater (40:17)

This is maybe like a too big, overly broad question, but when you say, what do mean when you say ethical AI?

Tye Brady (40:26)

That means that you always have people at it. You put people to center the robotics universe. Always. Right. Always you put people at the center like your employees, your frontline employees, the women and men who designed these systems. You make it for people. It's not people against machines. It's people and machines working together. And the onus is on us in order to make it responsible, accountable and ethical for human consumption.

Brian Heater (40:54)

My understanding is, after talking to a few people, that obviously the big first really publicized work that Amazon was doing in terms of exploration with the humanoid form factor in industrial warehouse setting was the 2023 partnerships with Agility. But the company was exploring some sort bipedal form factors prior to that. That was something that you had considered.

actually whether or not it would make sense to fit that into the processes that you had in place.

Tye Brady (41:30)

Well, I remember in the 90s, speaking of Mark Raibert, like in the locomotion lab, it is just truly about how to walk through difficult terrain. It's fascinating. It's a fascinating subject for sure. ? I know in the early days of agility as well that they had that in mind as well. I think they've shown a pretty famous video of almost a chicken, of how a chicken just kind of can make its way across.

variable terrain kind of with the physical intelligence, right, the physical intelligence that is new. I, again, I applaud those type of efforts of anything that we can do to advance the state of mobility. I applaud.

Brian Heater (42:11)

It strikes me that you will explore technology. I think maybe around that time it was again like early days of agility. were showing off Cassie. Explore the potential, realize maybe this isn't the time, we're not quite ready, then you don't count it out. Then you realize that the time will come and then we will try again to integrate this into the system.

Tye Brady (42:34)

Yeah, it's such a good thought because remember, it's a full spectrum when it comes to robotics, right? There's early technology readiness levels or TRL was what we used to call it in aerospace where there's ideation and there's proof of concept of like, is this possible? And I always remind the team, it's like, we don't do technology for technology sake, right? What problem are we trying to solve?

And even in the ideation phase, have to stay sharp and focused on what's the problem we're trying to solve in front of us. And if we see a problem that we want to solve, whether it's a new sortation or a new storage ? capability or a new manipulation capability, then that gives us the focus in order to advance our technologies. But it's not over then. Like there's so many steps in for it to get to the fulfillment center. Right. Yeah. It's commonly known as the Valley of Death. Right. So you have early stage ideation.

going on with even a proof of concept on this side. And on the far side, you have your operational users and they really don't want any change. They want stuff that works. Like they don't have time to mess around with like, need good tools, I need professional tools and they have to work, right? And they're resistant to change. So what happens in between? Well, if you're resistant to change, you're not gonna pull something through. And on the other side, if you have no idea what the application is, you're not gonna push something through.

But I think Amazon strikes a really great balance in between where we have product managers that understand what technologies that we do we need to evolve. They reach out to me and to my team and to ? robotics in general say, hey, if we can just solve this, that would be a really big win for us. So that puts it in the pipeline for us and helps pull it through that valley of death because there is application. But every time that we do that, mean, every single time that we do it, we may convince ourselves it's working great in the lab.

And we're like, my gosh, yeah, this is it. And then we'll go to the actual environment, TRL 7, we're in the actual environment, and we'll go to one quarter of one floor in one building and try it with real ? users, real frontline employees, and we will listen directly to their feedback. And we like to hear it exactly the way that they say it. Like, you don't get to change a single word about how they're saying what's good or what's bad about it.

And this is after it's gone through and we have 9,000 plus safety engineers, right? We have manufacturing engineers that are involved in this. There's a lot of process in order for it to even get to the floor. And then we will learn and we will iterate. And if it's not working, we'll iterate. And if we can't iterate it well enough for our employees, it's gone. It's like we go back to the drawing board because it's not amplifying human potential. It's not augmenting what the employee is capable of doing.

if it's a replacement strategy, other than the menial, the mundane and the repetitive, we want to make it better. Okay. So even, even if a one quarter of one floor, we'll go to one floor and then maybe we even go to one building and it's still not good enough. Right. When we start to roll something up for what we call general availability, that is really a big deal. Right. So there's so many steps in between of like what we're doing in the lab to actually, is it deployed on the front lines? And when it hits the front lines, it works. It works really well.

Brian Heater (45:51)

feel like maybe the most public facing and one of the most interesting examples of this is actually takes us outside of the warehouse floor is Scout, right? Because Scout to me, I'm looking at Scout and it's like, of course Amazon is exploring last mile delivery robot. makes perfect sense. ? Well publicized, know, at least at that point, you know.

from where I said it seems like those experiments didn't quite work out quite like Amazon was hoping. then obviously like folks like me in the media are like, right, this is gone, blah, blah, blah. That's what the stories are about. But an example like that, I can't imagine that that specific pilot didn't work. So you or whoever at Amazon is suddenly saying, all right, well, we're not.

doing last mile delivery robots anymore.

Tye Brady (46:50)

Well, you know, we also had Fire Phone as well. Right. I mean, sure. Right. Sure. That's a big one. Slightly. And Scalpel, look again, what problem you're trying to solve. One of the things that we saw in Scout that we we we didn't like is it would call attended delivery. When you get your packages today, those packages show up outside your door ready to go. And it's it's you know, you get your goods.

Brian Heater (46:56)

Different example, but yeah.

Tye Brady (47:18)

But if an intended delivery, if you have to rendezvous and schedule with a robot, that's adding friction. That's not customer obsession. That is not insisting on the highest standard for our customer. So we needed really to go back to the be vocally self-critical that we're not to what problem we're trying to solve. We weren't really on there. And we really needed to, ? I'd say, better that system.

As much as I love technology, I love a cool robot, I love that for sure, you got to think about what problem you're trying to solve. think Scout is a good example of like, hey, we need to go back to the drawing board.

Brian Heater (48:00)

Obviously for Amazon, robots really majorly starting in the warehouse and these fulfillment centers, really scaling that up, you 1 million units. ? You know, you've got your consumer ? work over here, but ? eventually, and you alluded to this before in terms of agriculture and other things, but eventually starting to see some of that work filter out into other things that Amazon does and other parts just of the world generally.

Tye Brady (48:29)

Absolutely. We started in what we call the structured fields, So structured fields with Hercules, with QR codes and robots and fences. Now we're moving towards semi-structured, like with Proteus, for example, fully safety certified, working around people, moving carts of goods to the dock doors at will, working amazingly beautiful. So from structured to semi-structured is something that we are mastering right now. So I wouldn't be surprised, you know, for us to continue to innovate in all ways for our customers, whether it's even in the less structured environment.

Brian Heater (49:04)

Ty, always a pleasure speaking with you. Great sweater. Hope to talk to you soon.

Tye Brady (49:09)

Looking sharp, Brian, always great to see you too. Take care, my friend.

Brian Heater (49:14)

wonderful conversation. Thank you so much to Ty for taking the time out of his schedule to set that up. Thanks to Andy at Amazon for helping set up that interview. Thanks to Maureen and VN, the humans and Frankie the dog for the production and editing this week. And thanks to you for sitting long enough to hear this outro. Please like and subscribe and don't forget to sign up for our weekly newsletter of the same name and we will see you next week for another episode of Automated.