WEBVTT - Amazon Robotics Chief Technologist Tye Brady Talks Physical AI

0:00:02.520 --> 0:00:07.000
<v Speaker 1>Bloomberg Audio Studios, podcasts, radio news.

0:00:07.760 --> 0:00:09.680
<v Speaker 2>Hey, one of the things that we love to talk

0:00:09.680 --> 0:00:15.319
<v Speaker 2>about robotics, technology, automation. Bloomberg Weekend recently reported app that

0:00:15.440 --> 0:00:17.759
<v Speaker 2>humanoid robots are coming as soon as they learn to

0:00:18.040 --> 0:00:19.800
<v Speaker 2>fold clothes. Count me in.

0:00:19.960 --> 0:00:21.720
<v Speaker 1>Okay, okay.

0:00:22.480 --> 0:00:27.160
<v Speaker 2>Why they wrote this? Apparently the team attended a recent

0:00:27.200 --> 0:00:30.200
<v Speaker 2>Silicon Valley summit and there they saw small robots roaming

0:00:30.240 --> 0:00:35.480
<v Speaker 2>and pouring lattes while evangelists hailed new AI techniques as transformative,

0:00:35.760 --> 0:00:38.159
<v Speaker 2>but full size prototypes you're pretty scarce.

0:00:38.240 --> 0:00:38.480
<v Speaker 3>Okay.

0:00:38.479 --> 0:00:39.760
<v Speaker 1>I want to see what typ Brady has to say

0:00:39.760 --> 0:00:42.880
<v Speaker 1>about this. He's chief technologists at Amazon Robotics. Tye is

0:00:42.920 --> 0:00:45.239
<v Speaker 1>also founding partner of Mass Robotics. It's a not for

0:00:45.360 --> 0:00:48.720
<v Speaker 1>profit organization that's become the world's largest robotics innovation center

0:00:48.760 --> 0:00:50.880
<v Speaker 1>and serves on a number of boards promote stem learning

0:00:50.880 --> 0:00:53.840
<v Speaker 1>and advancement. Ty good to have you back on the program.

0:00:54.280 --> 0:00:56.440
<v Speaker 1>Happy holidays. I want to get to this idea of

0:00:56.520 --> 0:01:00.000
<v Speaker 1>humanoid robots, because it doesn't matter if a humanoid robat

0:01:00.480 --> 0:01:02.440
<v Speaker 1>can fold laundry for a lot of the purposes that

0:01:02.840 --> 0:01:06.640
<v Speaker 1>you have for Amazon, How are robots apart from the

0:01:06.720 --> 0:01:09.399
<v Speaker 1>Kiva robots used right now in Amazon warehouses.

0:01:10.440 --> 0:01:12.520
<v Speaker 3>Well, thanks Tim, thanks for having me on, Carol, it'd

0:01:12.560 --> 0:01:15.760
<v Speaker 3>been nice to hear you as well. It's always a

0:01:15.800 --> 0:01:17.800
<v Speaker 3>pleasure to be on the show and talk all things

0:01:17.920 --> 0:01:21.440
<v Speaker 3>robotics and tech with you. Indeed, the age of physical

0:01:21.480 --> 0:01:24.679
<v Speaker 3>ais here, Amazon is building our physical aias systems to

0:01:24.680 --> 0:01:28.360
<v Speaker 3>help better the customer experience, expanding our selection, lowering cost,

0:01:28.440 --> 0:01:31.800
<v Speaker 3>fueling the faster delivery times that we know our customers love.

0:01:31.880 --> 0:01:36.319
<v Speaker 3>But given your question, think of that very practical everyday

0:01:36.400 --> 0:01:39.160
<v Speaker 3>experience where we're doing this at a very high reliability

0:01:39.400 --> 0:01:42.280
<v Speaker 3>and scalingness, where we're shipping millions and millions of products

0:01:42.440 --> 0:01:45.160
<v Speaker 3>every day here for the holiday season. It's really exciting

0:01:45.200 --> 0:01:48.840
<v Speaker 3>to see. It's very real, it's very practical, and it's

0:01:49.640 --> 0:01:52.360
<v Speaker 3>something that we're very proud of in what we're doing

0:01:52.360 --> 0:01:53.000
<v Speaker 3>with our employees.

0:01:53.040 --> 0:01:54.560
<v Speaker 2>All Right, we want to talk about this, you know,

0:01:54.680 --> 0:01:56.280
<v Speaker 2>we do, and we will. But I do want to

0:01:56.280 --> 0:01:58.160
<v Speaker 2>pick your brains because I always feel like I learned

0:01:58.200 --> 0:02:00.000
<v Speaker 2>so much, you know when you join us.

0:02:00.120 --> 0:02:00.960
<v Speaker 1>Bring it?

0:02:01.000 --> 0:02:04.120
<v Speaker 2>Is that so these humanoid robots, I mean Elon's working

0:02:04.120 --> 0:02:07.360
<v Speaker 2>on on other people. Is it years away? What's your thought,

0:02:07.400 --> 0:02:10.080
<v Speaker 2>your expertise you know this world.

0:02:10.120 --> 0:02:12.120
<v Speaker 3>I do know this world. I will tell you that

0:02:12.200 --> 0:02:15.480
<v Speaker 3>we usually start with what problem are we trying to solve?

0:02:15.760 --> 0:02:17.880
<v Speaker 3>What's the problem we're trying to solve. And once we

0:02:17.919 --> 0:02:21.079
<v Speaker 3>figure out the problem, then we actually go to functions.

0:02:21.120 --> 0:02:25.440
<v Speaker 3>What functions should the robotics should the physical AI systems do? Right, So,

0:02:26.000 --> 0:02:29.800
<v Speaker 3>from the functionality it should it fold laundry? Should it

0:02:29.840 --> 0:02:33.760
<v Speaker 3>do your dishes? Then we derive form and you kind

0:02:33.760 --> 0:02:36.280
<v Speaker 3>of get the KRT ahead of the horse when you

0:02:36.320 --> 0:02:39.200
<v Speaker 3>start with form first, and then see how can you

0:02:39.240 --> 0:02:41.880
<v Speaker 3>apply this technology? And we don't do technology for technology's

0:02:41.880 --> 0:02:44.360
<v Speaker 3>sake inside of Amazon. What we do is we solve problems.

0:02:44.400 --> 0:02:47.400
<v Speaker 3>We solve problems, every day problems at an incredible scale

0:02:47.400 --> 0:02:51.440
<v Speaker 3>that actually advances the state of the art in robotics. Right, So,

0:02:51.800 --> 0:02:55.320
<v Speaker 3>if you want to a robot to washer dishes, well,

0:02:55.400 --> 0:02:58.359
<v Speaker 3>congratulations you have that. You don't necessarily need a humanoid

0:02:58.440 --> 0:03:00.560
<v Speaker 3>form to wash your dishes. As a matter of fact,

0:03:00.560 --> 0:03:03.160
<v Speaker 3>it'd be kind of comical to see a robot pick

0:03:03.240 --> 0:03:06.000
<v Speaker 3>up the dish, pick up a sponge, you know, put

0:03:06.040 --> 0:03:07.880
<v Speaker 3>the soap on it, scrub the dish, and put it

0:03:07.919 --> 0:03:10.200
<v Speaker 3>in the drying rack. When you have this really practical

0:03:10.280 --> 0:03:13.880
<v Speaker 3>robot in your kitchen today called the dishwasher, which is amazing.

0:03:13.919 --> 0:03:17.400
<v Speaker 3>It blends right into the woodwork. So we think about

0:03:17.560 --> 0:03:20.760
<v Speaker 3>function first and then allow the form to follow.

0:03:20.919 --> 0:03:22.799
<v Speaker 1>Okay, so we want to talk specifics here about what's

0:03:22.800 --> 0:03:25.040
<v Speaker 1>happening in the warehouses blue Jay and a Luna and

0:03:25.080 --> 0:03:29.720
<v Speaker 1>more your systems blue Jay Robotics, Luna agentic AI delivery Glasses,

0:03:29.760 --> 0:03:32.799
<v Speaker 1>Amazon's warehouses. These are well known for robotics in the

0:03:32.880 --> 0:03:35.520
<v Speaker 1>use of tech and AI. How does all of that

0:03:35.560 --> 0:03:40.600
<v Speaker 1>continue to evolve and specifically impact productivity side by side

0:03:40.880 --> 0:03:41.520
<v Speaker 1>with human work?

0:03:41.600 --> 0:03:45.760
<v Speaker 3>Absolutely? Yeah. So the word I want to use is supercharge.

0:03:45.760 --> 0:03:49.480
<v Speaker 3>We're supercharging the world's largest fleet of robotics out there

0:03:49.920 --> 0:03:52.960
<v Speaker 3>with AI, using AI systems to better the reasoning and

0:03:53.000 --> 0:03:54.560
<v Speaker 3>to better assist our employees.

0:03:54.640 --> 0:03:54.760
<v Speaker 2>Right.

0:03:54.760 --> 0:03:56.880
<v Speaker 3>We want to eliminate the menial, the mundane, and the

0:03:56.920 --> 0:03:59.720
<v Speaker 3>repetitive through the use of robotics as a tool set

0:03:59.760 --> 0:04:03.320
<v Speaker 3>for employees. And we're entering this era where foundation models

0:04:03.600 --> 0:04:07.680
<v Speaker 3>are making a robot smarter, more affordable, more adaptable, more conversational,

0:04:07.840 --> 0:04:10.840
<v Speaker 3>and really reshaping work as we know it, not just

0:04:10.880 --> 0:04:13.640
<v Speaker 3>an e commerce but actually across the industries and create

0:04:13.760 --> 0:04:14.520
<v Speaker 3>new kinds of jobs.

0:04:15.280 --> 0:04:17.280
<v Speaker 2>I'm jumping in because right now we're showing for those

0:04:17.279 --> 0:04:20.120
<v Speaker 2>who are watching on TV and YouTube and our streaming service,

0:04:20.440 --> 0:04:24.839
<v Speaker 2>what looks like your special delivery glasses. As we continue

0:04:24.880 --> 0:04:26.880
<v Speaker 2>to show this, just tell us what they're all about.

0:04:26.880 --> 0:04:29.480
<v Speaker 2>They look like normal glasses. But what do they enable

0:04:30.240 --> 0:04:31.840
<v Speaker 2>a driver to do more easily?

0:04:31.920 --> 0:04:35.120
<v Speaker 3>Yeah, well, it's the fundamental princess that we have is

0:04:35.160 --> 0:04:37.080
<v Speaker 3>to use them as a tool set. So allow the

0:04:37.160 --> 0:04:39.680
<v Speaker 3>driver to understand where the delivery should be, allow them

0:04:39.720 --> 0:04:41.440
<v Speaker 3>to be kind of hands free in order to take

0:04:41.480 --> 0:04:44.040
<v Speaker 3>the picture of the package at your door, Allow them

0:04:44.080 --> 0:04:48.200
<v Speaker 3>to understand what's the best route to the customer's door,

0:04:48.400 --> 0:04:50.840
<v Speaker 3>allow them to see their delivery notes of where they

0:04:50.839 --> 0:04:55.760
<v Speaker 3>should place that the delivery at the right spot that's

0:04:55.839 --> 0:04:58.080
<v Speaker 3>just right for our customer. I mean, we're customer obsessed.

0:04:58.080 --> 0:05:00.240
<v Speaker 3>So any tool that we can give our deliver every

0:05:00.680 --> 0:05:03.800
<v Speaker 3>employees or or frontline employees, that's what that's how we

0:05:03.839 --> 0:05:05.160
<v Speaker 3>want to use technology.

0:05:05.200 --> 0:05:06.760
<v Speaker 1>Are they are they rolled out?

0:05:07.880 --> 0:05:10.800
<v Speaker 3>I think that we're still in the early stages of deployment.

0:05:10.839 --> 0:05:13.719
<v Speaker 1>So when would they be fully rolled out across the network?

0:05:14.400 --> 0:05:16.000
<v Speaker 3>Hard to tell. Would it be.

0:05:15.960 --> 0:05:18.480
<v Speaker 1>Within a year, maybe within twenty twenty six.

0:05:18.640 --> 0:05:20.960
<v Speaker 3>Yeah, we're pretty We're we are a pretty Once we

0:05:21.000 --> 0:05:23.680
<v Speaker 3>get to the alpha and beta deployment, it's typically within

0:05:23.720 --> 0:05:25.920
<v Speaker 3>it about a year that you'll see this starting to

0:05:26.000 --> 0:05:28.040
<v Speaker 3>roll out, and we take a very measured approach. When

0:05:28.080 --> 0:05:30.680
<v Speaker 3>things roll out, we test it. We make sure it's

0:05:30.839 --> 0:05:33.839
<v Speaker 3>great for our employees. Does it add value for our employees,

0:05:34.120 --> 0:05:37.000
<v Speaker 3>does it create a safer environment for employees? And then

0:05:37.320 --> 0:05:40.520
<v Speaker 3>we'll test that at a smaller scale, and then we

0:05:40.600 --> 0:05:43.159
<v Speaker 3>go into the you know, the millions and millions of scale.

0:05:43.320 --> 0:05:44.599
<v Speaker 2>Hey, I want to go back to Blue Jay in

0:05:44.600 --> 0:05:46.880
<v Speaker 2>a Luna timber on arapp. Blue Jay your next generation

0:05:47.000 --> 0:05:50.760
<v Speaker 2>robotics systems system. Excuse me, and a Luna is an

0:05:50.760 --> 0:05:54.280
<v Speaker 2>agentic AI system, So talk to us about that. How

0:05:54.360 --> 0:05:58.320
<v Speaker 2>much are they deployed today across your fulfillment fulfillment network?

0:05:58.320 --> 0:06:00.880
<v Speaker 2>And I'm curious about milestones we should all be watching

0:06:00.880 --> 0:06:03.120
<v Speaker 2>out for in the next twelve to twenty four months time.

0:06:03.960 --> 0:06:07.359
<v Speaker 3>Sure. So at Luna we're actually using kind of in

0:06:07.400 --> 0:06:12.080
<v Speaker 3>a monitor in a monitor only manner right now during

0:06:12.160 --> 0:06:16.159
<v Speaker 3>peak right we actually just just went through the peak

0:06:16.160 --> 0:06:19.960
<v Speaker 3>of peaks where we have our largest outbound happening. Just

0:06:20.040 --> 0:06:23.160
<v Speaker 3>actually yesterday, And what a Luna does is it gives

0:06:23.160 --> 0:06:27.279
<v Speaker 3>operators a complete real time view to help guide their

0:06:27.320 --> 0:06:30.000
<v Speaker 3>every move right. So think of many many dashboards now

0:06:30.120 --> 0:06:35.280
<v Speaker 3>can be consolidated into human consumable texts, human consumable recommendations

0:06:35.279 --> 0:06:38.320
<v Speaker 3>of how to actually deploy the network and how to

0:06:38.360 --> 0:06:42.400
<v Speaker 3>deploy various robotics inside that building in a way that

0:06:42.440 --> 0:06:45.599
<v Speaker 3>it gains these more efficiencies. So we're seeing that that

0:06:45.720 --> 0:06:49.520
<v Speaker 3>is rolling out today, and then we're really excited about

0:06:49.520 --> 0:06:53.000
<v Speaker 3>blue Jay as well. You can think of this really

0:06:53.080 --> 0:06:56.440
<v Speaker 3>for our sub same day network. That's a growing network

0:06:56.480 --> 0:06:59.320
<v Speaker 3>they have inside of Amazon where we're getting the goods

0:06:59.400 --> 0:07:02.000
<v Speaker 3>right to the customer and a matter of hours. What

0:07:02.040 --> 0:07:05.720
<v Speaker 3>blue Jay does is it really indexes on a smaller footprint.

0:07:05.960 --> 0:07:08.720
<v Speaker 3>You can think of this as three assembly lines kind

0:07:08.720 --> 0:07:11.680
<v Speaker 3>of combined into one, where we have robotic arms picking

0:07:12.040 --> 0:07:15.440
<v Speaker 3>orders for our customers out of our containerized storage as

0:07:15.480 --> 0:07:17.720
<v Speaker 3>systems and allowing those to be delivered right to the

0:07:17.720 --> 0:07:19.920
<v Speaker 3>customer's door. That's really an exciting time.

0:07:20.680 --> 0:07:23.280
<v Speaker 1>So if we think about this time right now, the

0:07:23.280 --> 0:07:26.160
<v Speaker 1>holiday peak where people are trying to get stuff as

0:07:26.240 --> 0:07:29.880
<v Speaker 1>quick as possible. Specifically, where does the tech that you

0:07:30.000 --> 0:07:34.280
<v Speaker 1>just outlined have the biggest impact. Is it smoothing labor variability,

0:07:34.320 --> 0:07:37.800
<v Speaker 1>preventing bottlenecks? It doesn't improve on time delivery metrics. Is

0:07:37.800 --> 0:07:39.640
<v Speaker 1>it all the above? Is it something I didn't even mention?

0:07:40.400 --> 0:07:43.240
<v Speaker 3>Yeah, it is all the above plus or what it's

0:07:43.280 --> 0:07:46.720
<v Speaker 3>allowing us to do is really exception handling. Right when

0:07:46.760 --> 0:07:49.320
<v Speaker 3>we think about robotics, and you think robotics and physical

0:07:49.320 --> 0:07:53.080
<v Speaker 3>AI at scale, when you ship millions and millions every day,

0:07:53.600 --> 0:07:56.840
<v Speaker 3>just one percent of exception handling can kind of each

0:07:56.920 --> 0:07:59.440
<v Speaker 3>eat your lunch. So any tools that we have to

0:07:59.520 --> 0:08:02.120
<v Speaker 3>help us with the exception handling that maybe one of

0:08:02.160 --> 0:08:05.360
<v Speaker 3>the systems is not picking up a particular object right,

0:08:05.400 --> 0:08:07.560
<v Speaker 3>so that learning needs to be propagated through our other

0:08:07.640 --> 0:08:10.800
<v Speaker 3>manipulation systems. Or maybe a mobility system is having some

0:08:10.840 --> 0:08:13.920
<v Speaker 3>bottlenecks because of the way the inventory is stored inside

0:08:13.920 --> 0:08:16.880
<v Speaker 3>of the building. People have really good skills of critical

0:08:16.920 --> 0:08:21.560
<v Speaker 3>reasoning at critical thinking and using common sense. So tools

0:08:21.560 --> 0:08:25.600
<v Speaker 3>that allow them to understand the situation better, perceive their

0:08:25.680 --> 0:08:28.280
<v Speaker 3>environment better through the use of robotics, and then have

0:08:29.280 --> 0:08:32.040
<v Speaker 3>truly physical agents to help with that in concert with

0:08:32.080 --> 0:08:34.280
<v Speaker 3>our employees. Is that allow us to deal with all

0:08:34.280 --> 0:08:38.520
<v Speaker 3>the exceptions that we see every day, especially at peak.

0:08:38.640 --> 0:08:40.640
<v Speaker 2>Hey, Ty, just got about a minute left here. And

0:08:40.679 --> 0:08:42.640
<v Speaker 2>I know we've talked about this with you before, and

0:08:42.679 --> 0:08:44.160
<v Speaker 2>I think it's a fair question, and you guys have

0:08:44.200 --> 0:08:46.800
<v Speaker 2>addressed it about the impact on your workforce. And I

0:08:46.840 --> 0:08:50.640
<v Speaker 2>wonder if it's smarter to think about maybe not that

0:08:50.679 --> 0:08:52.560
<v Speaker 2>what you guys are doing in terms of automation or

0:08:52.600 --> 0:08:55.720
<v Speaker 2>other companies for that matter, that it means doing away

0:08:55.760 --> 0:08:57.960
<v Speaker 2>with workers, but maybe it means that you will have

0:08:58.000 --> 0:09:01.839
<v Speaker 2>to hire fewer workers in the future. Is that fair?

0:09:03.240 --> 0:09:06.520
<v Speaker 3>Well, first of all, Carly, any question you ask is fair,

0:09:06.600 --> 0:09:08.520
<v Speaker 3>So I think it is spared always asks those questions,

0:09:08.559 --> 0:09:13.000
<v Speaker 3>and we should always kind of have the mindset of

0:09:13.080 --> 0:09:15.560
<v Speaker 3>people first, in which we do inside of Amazon. So

0:09:15.600 --> 0:09:17.840
<v Speaker 3>what are we doing specifically for our employees? And I'm

0:09:17.880 --> 0:09:20.440
<v Speaker 3>really proud of this. We're building better machines for them,

0:09:20.520 --> 0:09:23.240
<v Speaker 3>We're listening to the iterating our designs. We're rolling out

0:09:23.360 --> 0:09:25.000
<v Speaker 3>a tool set from them to use that every day

0:09:25.000 --> 0:09:28.120
<v Speaker 3>at scale a scale it's almost unimaginable when it comes

0:09:28.200 --> 0:09:31.600
<v Speaker 3>to what we've done inside of robotics, we're upscaling them.

0:09:31.600 --> 0:09:33.920
<v Speaker 3>We have a two point five billion dollar pledge that

0:09:34.000 --> 0:09:37.160
<v Speaker 3>we call Future Ready to prepare fifty million people for

0:09:37.200 --> 0:09:40.079
<v Speaker 3>how technology will impact and change not only the nature

0:09:40.120 --> 0:09:43.400
<v Speaker 3>of work, but also what their everyday lives. And we're

0:09:43.400 --> 0:09:47.520
<v Speaker 3>also creating a safer environment with our robotics, eliminating the repetitive,

0:09:47.520 --> 0:09:51.160
<v Speaker 3>the menial, the mundane, allowing machines to do the heavy

0:09:51.160 --> 0:09:54.720
<v Speaker 3>lifting and the repetitive actions, and allowing people to work

0:09:54.760 --> 0:09:57.840
<v Speaker 3>with them in order to use this beautiful thing here