Colin Angle: iRobot CEO | Lex Fridman Podcast #39

00:00:00.000 | The following is a conversation with Colin Angle.

00:00:02.800 | He's the CEO and co-founder of iRobot,

00:00:05.840 | a robotics company that for 29 years

00:00:08.220 | has been creating robots that operate successfully

00:00:11.180 | in the real world.

00:00:12.600 | Not as a demo or on a scale of dozens,

00:00:15.640 | but on a scale of thousands and millions.

00:00:18.860 | As of this year, iRobot has sold more than

00:00:21.900 | 25 million robots to consumers,

00:00:25.700 | including the Roomba vacuum cleaning robot,

00:00:28.200 | the Bravo floor mopping robot,

00:00:30.000 | and soon the Terra lawn mowing robot.

00:00:34.000 | 29 million robots successfully operating autonomously

00:00:37.660 | in real people's homes,

00:00:39.620 | to me is an incredible accomplishment

00:00:42.080 | of science, engineering, logistics,

00:00:45.080 | and all kinds of general entrepreneurial innovation.

00:00:48.760 | Most robotics companies fail.

00:00:51.340 | iRobot has survived and succeeded for 29 years.

00:00:56.900 | I spent all day at iRobot,

00:00:58.860 | including a long tour and conversation with Colin

00:01:01.400 | about the history of iRobot,

00:01:03.520 | and then sat down for this podcast conversation

00:01:06.740 | that would have been much longer

00:01:08.540 | if I didn't spend all day learning about

00:01:10.780 | and playing with the various robots

00:01:12.320 | and the company's history.

00:01:13.960 | I'll release the video of the tour separately.

00:01:17.500 | Colin, iRobot, its founding team,

00:01:20.720 | its current team, and its mission

00:01:23.220 | has been and continues to be an inspiration to me

00:01:26.220 | and thousands of engineers who are working hard

00:01:28.880 | to create AI systems that help real people.

00:01:33.020 | This is the Artificial Intelligence Podcast.

00:01:35.640 | If you enjoy it, subscribe on YouTube,

00:01:38.020 | give it five stars on iTunes,

00:01:39.780 | support it on Patreon,

00:01:41.280 | or simply connect with me on Twitter

00:01:43.280 | at Lex Friedman, spelled F-R-I-D-M-A-N.

00:01:47.120 | And now, here's my conversation with Colin Engel.

00:01:51.080 | In his 1942 short story, "Runaround,"

00:01:55.120 | from his iRobot collection,

00:01:56.960 | Asimov proposed the three laws of robotics

00:02:01.920 | in order, don't harm humans, obey orders, protect yourself.

00:02:06.800 | So two questions.

00:02:07.680 | First, does the Roomba follow these three laws?

00:02:11.640 | And also, more seriously,

00:02:14.760 | what role do you hope to see robots take in modern society

00:02:18.240 | and in the future world?

00:02:21.520 | - So the three laws are very thought-provoking

00:02:25.760 | and require such a profound understanding

00:02:30.760 | of the world a robot lives in,

00:02:36.280 | the ramifications of its action,

00:02:38.400 | and its own sense of self,

00:02:40.040 | that it's not a relevant bar,

00:02:44.740 | at least it won't be a relevant bar for decades to come.

00:02:50.120 | And so if Roomba follows the three laws,

00:02:54.600 | and I believe it does,

00:02:56.980 | it is designed to help humans, not hurt them,

00:03:00.960 | it's designed to be inherently safe,

00:03:03.160 | and we designed it to last a long time,

00:03:06.040 | it's not through any AI or intent on the robot's part.

00:03:11.640 | It's because following the three laws

00:03:15.000 | is aligned with being a good robot product.

00:03:19.640 | So I guess it does,

00:03:23.120 | but not by explicit design.

00:03:27.240 | - So then the bigger picture,

00:03:28.800 | what role do you hope to see robotics, robots take

00:03:33.040 | in what's currently mostly a world of humans?

00:03:37.360 | - We need robots to help us continue

00:03:42.360 | to improve our standard of living.

00:03:46.280 | We need robots because the average age of humanity

00:03:51.280 | is increasing very quickly,

00:03:55.200 | and simply the number of people young enough

00:03:59.920 | and spry enough to care for the older,

00:04:02.720 | growing demographic is inadequate.

00:04:08.080 | And so what is the role of robots?

00:04:11.560 | Today the role is to make our lives a little easier,

00:04:15.080 | a little cleaner, maybe a little healthier,

00:04:17.500 | but in time, robots are gonna be the difference

00:04:22.280 | between real gut-wrenching declines

00:04:25.840 | in our ability to live independently

00:04:28.280 | and maintain our standard of living,

00:04:30.440 | and a future that is the bright one

00:04:35.060 | where we have more control over our lives,

00:04:37.640 | can spend more of our time focused on activities we choose,

00:04:44.680 | and I'm so honored and excited

00:04:48.080 | to be playing a role in that journey.

00:04:50.520 | - So you give me a tour,

00:04:51.840 | you showed me some of the long histories,

00:04:54.080 | now 29 years that iRobot has been at it,

00:04:57.280 | creating some incredible robots.

00:04:59.320 | You showed me PackBot,

00:05:01.280 | you showed me a bunch of other stuff

00:05:03.200 | that led up to Roomba, that led to Bravo and Terra.

00:05:08.200 | So let's skip that incredible history

00:05:14.080 | in the interest of time,

00:05:15.040 | 'cause we already talked about it,

00:05:16.120 | I'll show this incredible footage.

00:05:18.040 | You mentioned elderly, and robotics in society,

00:05:22.640 | I think the home is a fascinating place for robots to be.

00:05:26.260 | So where do you see robots in the home?

00:05:29.800 | Currently, I would say, once again,

00:05:31.640 | probably most homes in the world don't have a robot.

00:05:34.520 | So how do you see that changing?

00:05:36.160 | Where do you think is the big initial value add

00:05:39.840 | that robots can do?

00:05:41.920 | - So iRobot has sort of over the years,

00:05:45.000 | narrowed in on the home, the consumer's home,

00:05:49.400 | as the place where we want to innovate

00:05:53.160 | and deliver tools that will help a home

00:05:58.160 | be a more automatically maintained place,

00:06:04.280 | a healthier place, a safer place,

00:06:06.840 | and perhaps even a more efficient place to be.

00:06:11.520 | And today, vacuum, we mop, soon we'll be mowing your lawn,

00:06:16.520 | but where things are going is,

00:06:21.480 | when do we get to the point where the home,

00:06:27.080 | not just the robots that live in your home,

00:06:29.120 | but the home itself becomes part of a system

00:06:32.160 | that maintains itself and plays an active role

00:06:35.960 | in caring for and helping the people who live in that home.

00:06:40.760 | And I see everything that we're doing

00:06:43.200 | as steps along the path toward that future.

00:06:46.160 | - So what are the steps?

00:06:47.720 | So if we can summarize some of the history of Roomba,

00:06:51.760 | you've mentioned, and maybe you can elaborate on it,

00:06:55.520 | but you mentioned that the early days

00:06:57.240 | were really taking a robot from something that works

00:07:02.240 | either in the lab or something that works in the field

00:07:04.880 | that helps soldiers do the difficult work they do

00:07:10.200 | to actually be in the hands of consumers

00:07:12.640 | and tens of thousands, hundreds of thousands of robots

00:07:15.640 | that don't break down over how much people love them

00:07:18.480 | over months of very extensive use.

00:07:21.440 | So that was the big first step.

00:07:22.840 | And then the second big step was the ability

00:07:26.000 | to sense the environment, to build a map, to localize,

00:07:29.920 | to be able to build a picture of the home

00:07:32.600 | that the human can then attach labels to

00:07:34.640 | in terms of giving some semantic knowledge to the robot

00:07:39.120 | about its environment.

00:07:40.920 | Okay, so that's like a huge, two big, huge steps.

00:07:44.880 | Maybe you can comment on them,

00:07:47.560 | but also what is the next step

00:07:51.080 | of making a robot part of the home?

00:07:54.760 | - Sure.

00:07:55.600 | So the goal is to make a home that takes care of itself,

00:08:00.600 | takes care of the people in the home,

00:08:03.720 | and gives the user an experience of just living their life

00:08:07.860 | in the home is somehow doing the right thing,

00:08:10.880 | turning on and off lights when you leave,

00:08:14.160 | cleaning up the environment.

00:08:17.280 | And we went from robots that were great in the lab,

00:08:22.280 | but were both too expensive and not sufficiently capable

00:08:30.000 | to ever do an acceptable job of anything

00:08:33.800 | other than being a toy or a curio in your home

00:08:37.360 | to something that was both affordable

00:08:42.160 | and sufficiently effective to drive,

00:08:45.840 | be above threshold and drive purchase intent.

00:08:48.520 | Now we've disrupted the entire vacuuming industry.

00:08:54.400 | The number one selling vacuums, for example,

00:08:59.360 | in the US are Roombas, so not robot vacuums, but vacuums.

00:09:02.960 | And that's really crazy and weird.

00:09:04.640 | - Yes, we need to pause it.

00:09:06.440 | I mean, that's incredible.

00:09:08.080 | That's incredible that a robot is the number one selling

00:09:13.080 | thing that does something.

00:09:15.560 | - Yep.

00:09:16.400 | - Something as essential as vacuuming.

00:09:18.240 | - So we're--

00:09:19.080 | - Congratulations.

00:09:20.080 | - Thank you.

00:09:20.920 | It's still kind of fun to say,

00:09:22.440 | but just because this was a crazy idea

00:09:26.600 | that just started in a room here,

00:09:30.920 | we're like, do you think we can do this?

00:09:33.700 | So, hey, let's give it a try.

00:09:36.180 | But now the robots are starting

00:09:40.420 | to understand their environment.

00:09:42.860 | And if you think about the next step,

00:09:45.260 | there's two dimensions.

00:09:47.900 | I've been working so hard since the beginning of iRobot

00:09:53.060 | to make robots are autonomous,

00:09:55.100 | that they're smart enough and understand their task enough

00:09:59.140 | that they can just go do it without human involvement.

00:10:03.500 | Now what I'm really excited and working on

00:10:07.460 | is how do I make them less autonomous?

00:10:09.560 | Meaning that the robot is supposed to be your partner,

00:10:15.660 | not this automaton that just goes and does

00:10:18.300 | what a robot does.

00:10:20.220 | And so that if you tell it,

00:10:22.500 | hey, I just dropped some flour by the fridge in the kitchen.

00:10:27.140 | Can you deal with it?

00:10:28.980 | Wouldn't it be awesome if the right thing just happened

00:10:32.780 | based on that utterance?

00:10:35.260 | And to some extent that's less autonomous

00:10:37.980 | 'cause it's actually listening to you,

00:10:40.140 | understanding the context and intent of the sentence,

00:10:44.420 | mapping it against its understanding of the home it lives in

00:10:49.420 | and knowing what to do.

00:10:52.700 | And so that's an area of research.

00:10:56.380 | It's an area where we're starting to roll out features.

00:10:59.400 | You can now tell your robot to clean up the kitchen

00:11:02.900 | and it knows what the kitchen is and can do that.

00:11:05.900 | And that's sort of 1.0 of where we're going.

00:11:09.380 | The other cool thing is that we're starting to know

00:11:13.300 | where stuff is and why is that important?

00:11:16.060 | Well, robots are supposed to have arms, right?

00:11:21.060 | Data had an arm, Rosie had an arm,

00:11:24.260 | Robbie the robot had an arm.

00:11:25.300 | I mean, robots are,

00:11:26.700 | they are physical things that move around

00:11:28.800 | in an environment and they're supposed to do work.

00:11:31.280 | And if you think about it,

00:11:34.140 | if a robot doesn't know where anything is,

00:11:37.240 | why should it have an arm?

00:11:38.800 | But with this new dawn of home understanding

00:11:43.800 | that we're starting to go enjoy,

00:11:47.720 | I know where the kitchen is.

00:11:49.360 | I might in the future know where the refrigerator is.

00:11:52.080 | I might, if I had an arm, be able to find the handle,

00:11:55.320 | open it and even get myself a beer.

00:11:58.540 | Obviously that's one of the true dreams of robotics

00:12:01.980 | is to have robots bringing us a beer

00:12:03.580 | while we watch television.

00:12:05.340 | But I think that that new category of tasks

00:12:10.340 | where physical manipulation, robot arms,

00:12:14.260 | is just a potpourri of new opportunity and excitement.

00:12:19.260 | - And you see humans as a crucial part of that.

00:12:23.820 | So you kind of mentioned that,

00:12:26.320 | and I personally find that a really compelling idea.

00:12:29.000 | I think full autonomy can only take us so far,

00:12:34.000 | especially in the home.

00:12:35.360 | So you see humans as helping the robot understand

00:12:38.920 | or give deeper meaning to this spatial information.

00:12:42.580 | - Right, it's a partnership.

00:12:45.720 | The robot is supposed to operate

00:12:48.980 | according to descriptors that you would use

00:12:53.980 | to describe your own home.

00:12:55.640 | The robot is supposed to, in lieu of better direction,

00:13:02.040 | kind of go about its routine,

00:13:03.960 | which ought to be basically right

00:13:07.660 | and lead to a home maintained

00:13:12.340 | in a way that it's learned you like,

00:13:14.980 | but also be perpetually ready to take direction

00:13:19.980 | that would activate a different set of behaviors or actions

00:13:26.560 | to meet a current need

00:13:29.040 | to the extent it could actually perform that task.

00:13:32.420 | - So I gotta ask you,

00:13:33.640 | I think this is a fundamental and a fascinating question,

00:13:37.040 | because iRobot has been a successful company

00:13:39.860 | and a rare successful robotics company.

00:13:42.400 | So Anki, Jibo, Mayfield Robotics with their Robot Curry,

00:13:46.820 | Sci-Fi Works, Rethink Robotics,

00:13:49.260 | these are robotics companies that were founded

00:13:51.400 | and run by brilliant people.

00:13:53.000 | But all, very unfortunately, at least for us roboticists,

00:13:59.100 | all went out of business recently.

00:14:02.180 | So why do you think they didn't last longer?

00:14:05.180 | Why do you think it is so hard

00:14:07.020 | to keep a robotics company alive?

00:14:10.700 | - You know, I say this only partially in jest

00:14:14.180 | that back in the day before Roomba,

00:14:16.800 | you know, I was a high-tech entrepreneur building robots,

00:14:22.860 | but it wasn't until I became a vacuum cleaner salesman

00:14:26.460 | that we had any success.

00:14:28.240 | So, I mean, the point is technology alone

00:14:34.220 | doesn't equal a successful business.

00:14:37.720 | We need to go and find the compelling need

00:14:42.640 | where the robot that we're creating

00:14:45.920 | can deliver clearly more value

00:14:52.640 | to the end user than it costs.

00:14:55.440 | And this is not a marginal thing

00:14:59.040 | where you're looking at the scale and you're like,

00:15:00.400 | "Eh, it's close, maybe we can hold our breath

00:15:03.360 | "and make it work."

00:15:04.380 | It's clearly more value than the cost of the robot

00:15:09.380 | to bring in the store.

00:15:13.900 | And I think that the challenge has been

00:15:15.820 | finding those businesses

00:15:20.820 | where that's true in a sustainable fashion.

00:15:26.020 | You know, when you get into entertainment style things,

00:15:34.060 | you could be the cat's meow one year,

00:15:38.220 | but 85% of toys, regardless of their merit,

00:15:42.400 | fail to make it to their second season.

00:15:45.620 | It's just super hard to do so.

00:15:47.720 | And so that's just a tough business.

00:15:53.700 | And there's been a lot of experimentation

00:15:57.780 | around what is the right type of social companion?

00:16:02.620 | What is the right robot in the home

00:16:05.840 | that is doing something other than tasks people do

00:16:10.840 | every week that they'd rather not do?

00:16:16.380 | And I'm not sure we've got it all figured out right.

00:16:20.880 | And so that you get brilliant roboticists

00:16:22.960 | with super interesting robots

00:16:25.660 | that ultimately don't quite have

00:16:29.800 | that magical user experience,

00:16:32.860 | and thus that value benefit equation remains ambiguous.

00:16:37.860 | - So you as somebody who dreams of robots

00:16:44.240 | changing the world, what's your estimate?

00:16:46.900 | How big is the space of applications

00:16:53.200 | that fit the criteria that you just described

00:16:55.780 | where you can really demonstrate

00:16:58.060 | an obvious significant value

00:17:00.480 | over the alternative non-robotic solution?

00:17:04.620 | - Well, I think that we're just about none of the way

00:17:08.660 | to achieving the potential of robotics at home.

00:17:13.340 | But we have to do it in a really eyes wide open,

00:17:18.340 | honest fashion.

00:17:22.340 | - Another way to put that is the potential is infinite

00:17:25.380 | because we did take a few steps,

00:17:27.020 | but you're saying those steps are just very initial steps.

00:17:29.620 | So the Roomba is a hugely successful product,

00:17:32.540 | but you're saying that's just the very, very beginning.

00:17:34.380 | - That's just the very, very beginning.

00:17:36.500 | It's the foot in the door.

00:17:37.940 | And I think I was lucky that in the early days of robotics,

00:17:42.940 | people would ask me, "When are you gonna clean my floor?"

00:17:48.380 | It was something that I grew up saying,

00:17:52.240 | "I got all these really good ideas,

00:17:54.780 | but everyone seems to want their floor clean.

00:17:58.060 | And so maybe we should do that."

00:18:02.260 | - Yeah, your good ideas--

00:18:03.380 | - Earn the right to do the next thing after that.

00:18:05.820 | - So the good ideas have to match

00:18:07.900 | with the desire of the people,

00:18:10.140 | and then the actual cost has to,

00:18:12.580 | like the financial aspect has to all match together.

00:18:16.620 | - Yeah, during our partnership back a number of years ago

00:18:21.260 | with Johnson Wax, they would explain to me

00:18:24.100 | that they would go into homes

00:18:29.100 | and just watch how people lived

00:18:32.500 | and try to figure out what were they doing

00:18:35.540 | that they really didn't really like to do,

00:18:39.960 | but they had to do it frequently enough

00:18:42.420 | that it was top of mind and understood as a burden.

00:18:51.700 | Hey, let's make a product or come up with a solution

00:18:55.860 | to make that pain point less challenging.

00:19:00.860 | - And sometimes we do certain burdens so often as a society

00:19:07.060 | that we actually don't even realize,

00:19:09.420 | like it's actually hard to see that that burden

00:19:11.460 | is something that could be removed.

00:19:13.180 | So it does require just going into the home and staring at,

00:19:17.140 | wait, how do I actually live life?

00:19:19.560 | What are the pain points?

00:19:21.060 | - Yeah, and getting those insights is a lot harder

00:19:26.060 | than it would seem it should be in retrospect.

00:19:29.340 | - So how hard on that point,

00:19:33.100 | I mean, one of the big challenges of robotics

00:19:37.420 | is driving the cost down to something

00:19:42.220 | that consumers, people would afford.

00:19:45.660 | So people would be less likely to buy a Roomba

00:19:48.860 | if it costs $500,000, right?

00:19:52.140 | Which is probably sort of what a Roomba would cost

00:19:55.900 | several decades ago.

00:19:58.080 | So how do you drive, which I mentioned is very difficult,

00:20:02.260 | how do you drive the cost of a Roomba or a robot down

00:20:05.380 | such that people would want to buy it?

00:20:07.960 | - When I started building robots,

00:20:09.740 | the cost of the robot had a lot to do

00:20:12.260 | with the amount of time it took to build it.

00:20:15.520 | And so that we would build our robots out of aluminum,

00:20:18.420 | I would go spend my time in the machine shop

00:20:21.220 | on the milling machine, cutting out the parts and so forth.

00:20:26.220 | And then when we got into the toy industry,

00:20:29.780 | I realized that if we were building at scale,

00:20:34.580 | I could determine the cost of the robot

00:20:36.020 | instead of adding up all the hours to mill out the parts,

00:20:38.940 | but by weighing it.

00:20:40.400 | And that's liberating.

00:20:44.220 | You can say, wow, the world has just changed

00:20:49.220 | as I think about construction in a different way.

00:20:53.200 | The 3D CAD tools that are available to us today,

00:20:56.940 | the operating at scale where I can do tooling

00:21:01.740 | and injection mold an arbitrarily complicated part

00:21:06.740 | and the cost is going to be basically the weight

00:21:10.500 | of the plastic in that part is incredibly exciting

00:21:15.500 | and liberating and opens up all sorts of opportunities.

00:21:18.580 | And for the sensing part of it,

00:21:22.120 | where we are today is instead of trying to build skin,

00:21:28.140 | which is like really hard for a long time,

00:21:31.420 | I spent creating strategies and ideas

00:21:38.260 | around how could we duplicate the skin on the human body

00:21:42.740 | because it's such an amazing sensor.

00:21:45.440 | Instead of going down that path,

00:21:49.620 | why don't we focus on vision?

00:21:53.020 | And how many of the problems that face a robot

00:21:59.020 | trying to do real work could be solved

00:22:04.860 | with a cheap camera and a big ass computer.

00:22:08.460 | - Yeah.

00:22:09.300 | - And Moore's law continues to work.

00:22:12.460 | The cell phone industry, the mobile industry

00:22:16.540 | is giving us better and better tools

00:22:18.800 | that can run on these embedded computers.

00:22:21.140 | And I think we passed an important moment,

00:22:26.140 | maybe two years ago,

00:22:29.580 | where you could put machine vision capable processors

00:22:35.420 | on robots at consumer price points.

00:22:39.620 | And I was waiting for it to happen.

00:22:43.020 | We avoided putting lasers on our robots to do navigation

00:22:48.020 | and instead spent years researching

00:22:51.820 | how to do vision based navigation

00:22:54.620 | because you could just see

00:22:58.420 | where these technology trends were going.

00:23:01.620 | And between injection molded plastic

00:23:05.860 | and a camera with a computer capable of running

00:23:09.900 | machine learning and visual object recognition,

00:23:12.520 | I could build an incredibly affordable,

00:23:15.520 | incredibly capable robot and that's gonna be the future.

00:23:20.520 | - So you know, on that point with a small tangent,

00:23:23.380 | but I think an important one,

00:23:24.980 | another industry in which I would say

00:23:27.540 | the only other industry in which

00:23:29.780 | there is automation actually touching people's lives today

00:23:34.780 | is autonomous vehicles.

00:23:36.500 | What the vision you just described of using computer vision

00:23:42.340 | and using cheap camera sensors,

00:23:44.420 | there's a debate on that of LIDAR versus computer vision.

00:23:48.260 | And sort of, Elon Musk famously said

00:23:53.260 | that LIDAR is a crutch that really in camera,

00:23:57.180 | in the longterm camera only is the right solution,

00:24:00.840 | which echoes some of the ideas you're expressing.

00:24:03.500 | Of course, the domain in terms of its safety criticality

00:24:06.780 | is different, but what do you think about that approach

00:24:10.660 | in the autonomous vehicle space?

00:24:13.420 | And in general, do you see a connection

00:24:15.180 | between the incredible real world challenges

00:24:18.520 | you have to solve in the home with Roomba?

00:24:20.780 | And I saw a demonstration of some of them,

00:24:22.940 | corner cases, literally, and autonomous vehicles.

00:24:27.880 | - So there's absolutely a tremendous overlap

00:24:31.680 | between both the problems, you know,

00:24:35.460 | a robot vacuum and autonomous vehicle are trying to solve,

00:24:38.660 | and the tools and the types of sensors

00:24:41.860 | that are being applied in the pursuit of the solutions.

00:24:48.020 | In my world, my environment is actually much harder

00:24:53.020 | than the environment in automobile travels.

00:24:57.320 | We don't have roads, we have t-shirts, we have steps,

00:25:02.320 | we have a near infinite number of patterns and colors

00:25:07.420 | and surface textures on the floor.

00:25:10.180 | - Especially from a visual perspective.

00:25:12.540 | - Yeah, visually it's really tough.

00:25:14.740 | It's infinitely variable.

00:25:18.860 | - On the other hand, safety is way easier on the inside.

00:25:22.540 | My robots, they're not very heavy, they're not very fast.

00:25:27.540 | If they bump into your foot, you think it's funny.

00:25:31.480 | And, you know, and autonomous vehicles

00:25:36.940 | kind of have the inverse problem.

00:25:39.420 | And so that for me saying vision is the future,

00:25:45.420 | I can say that without reservation.

00:25:47.780 | For autonomous vehicles, I think I believe what

00:25:52.740 | Elon's saying about the future

00:25:56.900 | is ultimately going to be vision.

00:25:59.020 | Maybe if we put a cheap lighter on there

00:26:01.060 | as a backup sensor,

00:26:02.260 | it might not be the worst idea in the world.

00:26:03.820 | - So the stakes are much higher.

00:26:05.020 | - The stakes are much higher.

00:26:05.860 | - You have to be much more careful thinking through

00:26:08.220 | how far away that future is, right?

00:26:10.740 | - Right, but I think that the primary

00:26:14.220 | environmental understanding sensor

00:26:19.340 | is going to be a visual system.

00:26:21.780 | - Visual system.

00:26:23.020 | So on that point, well, let me ask,

00:26:25.580 | do you hope there's an iRobot robot

00:26:28.420 | in every home in the world one day?

00:26:30.920 | - I expect there to be at least one iRobot robot

00:26:34.900 | in every home.

00:26:37.740 | You know, we've sold 25 million robots.

00:26:41.180 | So we're in about 10% of US homes,

00:26:44.620 | which is a great start.

00:26:45.780 | But I think that when we think about the numbers of things

00:26:52.140 | that robots can do,

00:26:54.300 | today I can vacuum your floor, mop your floor,

00:26:58.580 | cut your lawn, or soon we'll be able to cut your lawn.

00:27:01.280 | But there are more things that we could do in the home.

00:27:06.700 | And I hope that we continue using the techniques

00:27:11.500 | I described around exploiting computer vision

00:27:14.460 | and low-cost manufacturing that we'll be able

00:27:18.660 | to create these solutions at affordable price points.

00:27:22.620 | - So let me ask, on that point of a robot in every home,

00:27:25.580 | that's my dream as well.

00:27:26.840 | I'd love to see that.

00:27:28.580 | You know, I think the possibilities there

00:27:31.300 | are indeed infinite positive possibilities.

00:27:34.500 | But in our current culture,

00:27:36.740 | no thanks to science fiction and so on,

00:27:40.500 | there's a serious kind of hesitation,

00:27:44.700 | anxiety, concern about robots,

00:27:47.340 | and also a concern about privacy.

00:27:50.040 | And it's a fascinating question to me,

00:27:54.080 | why that concern is amongst a certain group of people

00:27:59.540 | is as intense as it is.

00:28:02.800 | So you have to think about it,

00:28:04.220 | 'cause it's a serious concern,

00:28:05.460 | but I wonder how you address it best.

00:28:07.980 | So from a perspective of a vision sensor,

00:28:09.780 | so robots that move about the home and sense the world,

00:28:14.060 | how do you alleviate people's privacy concerns?

00:28:19.060 | How do you make sure that they can trust iRobot

00:28:22.820 | and the robots that they share their home with?

00:28:25.300 | - I think that's a great question.

00:28:28.120 | And we've really leaned way forward on this

00:28:33.740 | because given our vision as to the role the company

00:28:38.740 | intends to play in the home,

00:28:40.700 | really for us, make or break is,

00:28:45.460 | can our approach be trusted to protecting the data

00:28:50.460 | and the privacy of the people who have our robots?

00:28:53.500 | And so we've gone out publicly with a privacy manifesto

00:28:58.300 | stating we'll never sell your data.

00:29:00.400 | We've adopted GDPR, not just where GDPR is required,

00:29:05.400 | but globally.

00:29:07.560 | We have ensured that images don't leave the robot.

00:29:14.160 | So processing data from the visual sensors

00:29:22.060 | happens locally on the robot

00:29:23.660 | and only semantic knowledge of the home

00:29:29.480 | with the consumer's consent is sent up.

00:29:32.720 | We show you what we know and are trying to go

00:29:36.280 | use data as an enabler for the performance of the robots

00:29:43.440 | with the informed consent and understanding

00:29:49.960 | of the people who own those robots.

00:29:52.400 | And we take it very seriously.

00:29:56.840 | And ultimately we think that by showing a customer that,

00:30:01.840 | if you let us build a semantic map of your home

00:30:07.360 | and know where the rooms are,

00:30:08.960 | well then you can say, clean the kitchen.

00:30:11.720 | If you don't want the robot to do that,

00:30:13.680 | don't make the map, it'll do its best job

00:30:15.800 | cleaning your home, but it won't be able to do that.

00:30:18.640 | And if you ever want us to forget

00:30:20.280 | that we know that it's your kitchen,

00:30:22.080 | you can have confidence that we will do that for you.

00:30:26.680 | So we're trying to go and be a sort of a

00:30:30.840 | data 2.0 perspective company where we treat the data

00:30:37.320 | that the robots have of the consumer's home

00:30:40.800 | as if it were the consumer's data

00:30:43.200 | and that they have rights to it.

00:30:47.320 | So we think by being the good guys on this front,

00:30:50.920 | we can build the trust and thus be entrusted

00:30:55.080 | to enable robots to do more things that are thoughtful.

00:31:00.080 | - You think people's worries will diminish over time?

00:31:03.760 | As a society, broadly speaking,

00:31:06.800 | do you think you can win over trust,

00:31:09.280 | not just for the company,

00:31:10.600 | but just the comfort that people have

00:31:12.940 | with AI in their home, enriching their lives in some way?

00:31:17.040 | - I think we're in an interesting place today

00:31:19.560 | where it's less about winning them over

00:31:22.400 | and more about finding a way to talk about privacy

00:31:26.280 | in a way that more people can understand.

00:31:28.880 | I would tell you that today,

00:31:30.960 | when there's a privacy breach,

00:31:33.360 | people get very upset and then go to the store

00:31:37.080 | and buy the cheapest thing,

00:31:38.360 | paying no attention to whether or not the products

00:31:41.040 | that they're buying honor privacy standards or not.

00:31:44.680 | In fact, if I put on the package of my Roomba,

00:31:50.120 | the privacy commitments that we have,

00:31:53.680 | I would sell less than I would if I did nothing at all.

00:31:58.680 | And that needs to change.

00:32:00.440 | So it's not a question about earning trust.

00:32:02.920 | I think that's necessary but not sufficient.

00:32:05.040 | We need to figure out how to have a comfortable set

00:32:08.480 | of what is the grade A meat standard applied to privacy

00:32:13.480 | that customers can trust and understand

00:32:18.440 | and then use in their buying decisions.

00:32:20.520 | That will reward companies for good behavior

00:32:25.520 | and that will ultimately be how this moves forward.

00:32:29.880 | - And maybe be part of the conversation

00:32:32.680 | between regular people about what it means,

00:32:34.800 | what privacy means.

00:32:36.280 | If you have some standards, you can say,

00:32:38.400 | you can start talking about who's following them,

00:32:41.080 | who's not, have more.

00:32:42.680 | 'Cause most people are actually quite clueless

00:32:45.440 | about all aspects of artificial intelligence

00:32:47.320 | or data collection and so on.

00:32:48.440 | It would be nice to change that for people to understand

00:32:51.440 | the good that AI can do and it's not some system

00:32:55.160 | that's trying to steal all the most sensitive data.

00:32:58.240 | - Yep.

00:32:59.080 | - Do you think, do you dream of a Roomba

00:33:02.640 | with human level intelligence one day?

00:33:05.240 | So you've mentioned a very successful localization

00:33:10.240 | and mapping of the environment,

00:33:11.880 | being able to do some basic communication

00:33:14.340 | to say go clean the kitchen.

00:33:16.560 | Do you see in your maybe more bored moments,

00:33:21.360 | once you get the beer, to sit back with that beer

00:33:27.000 | and have a chat on a Friday night with a Roomba

00:33:30.760 | about how your day went?

00:33:32.080 | - So to your latter question, absolutely.

00:33:37.560 | To your former question as to whether a robot

00:33:40.760 | can have human level intelligence, not in my lifetime.

00:33:45.280 | - You can have, you--

00:33:46.120 | - I think you can have a great conversation,

00:33:48.440 | a meaningful conversation with a robot

00:33:52.000 | without it having anything that resembles

00:33:56.320 | human level intelligence.

00:33:57.760 | And I think that as long as you realize

00:34:02.780 | that conversation is not about the robot

00:34:05.280 | and making the robot feel good.

00:34:08.560 | That conversation is about you learning interesting things

00:34:14.840 | that make you feel like the conversation

00:34:18.360 | that you had with the robot is

00:34:21.200 | a pretty awesome way of learning something.

00:34:27.240 | And it could be about what kind of day your pet had.

00:34:30.880 | It could be about how can I make my home

00:34:35.140 | more energy efficient?

00:34:36.240 | It could be about if I'm thinking about climbing Mount Everest

00:34:40.880 | what should I know?

00:34:44.440 | And that's a very doable thing.

00:34:46.560 | But if I think that that conversation

00:34:51.560 | I'm gonna have with a robot is I'm gonna be rewarded

00:34:54.640 | by making the robot happy,

00:34:56.880 | well I could just put a button on the robot

00:34:58.760 | that you could push and the robot would smile

00:35:00.320 | and that sort of thing.

00:35:02.160 | So I think you need to think about the question

00:35:04.440 | in the right way and robots can be awesomely effective

00:35:11.600 | at helping people feel less isolated,

00:35:14.440 | learn more about the home that they live in

00:35:17.560 | and fill some of those lonely gaps

00:35:21.980 | that we wish we were engaged learning cool stuff

00:35:24.880 | about our world.

00:35:25.720 | - Last question.

00:35:27.500 | If you could hang out for a day with a robot

00:35:32.360 | from science fiction, movies, books

00:35:35.800 | and safely pick its brain for that day,

00:35:40.400 | who would you pick?

00:35:42.120 | - Data.

00:35:43.360 | - Data.

00:35:44.200 | - From Star Trek.

00:35:45.280 | I think that A, data's really smart.

00:35:49.640 | Data's been through a lot trying to go and save the galaxy

00:35:53.560 | and I'm really interested actually in emotion and robotics

00:35:58.560 | and I think he'd have a lot to say about that

00:36:03.160 | 'cause I believe actually that emotion plays

00:36:08.520 | an incredibly useful role in doing reasonable things

00:36:13.520 | in situations where we have imperfect understanding

00:36:16.880 | of what's going on.

00:36:18.040 | - In social situations when there's imperfect information.

00:36:20.800 | - In social situations, also in competitive

00:36:25.520 | or dangerous situations that we have emotion for a reason

00:36:30.520 | and so that ultimately my theory is that

00:36:37.080 | as robots get smarter and smarter,

00:36:38.560 | they're actually gonna get more emotional

00:36:40.560 | because you can't actually survive on pure logic

00:36:46.440 | because only a very tiny fraction of the situations

00:36:53.800 | we find ourselves in can be resolved reasonably with logic

00:36:57.240 | and so I think data would have a lot to say about that

00:36:59.520 | and so I could find out whether he agrees.

00:37:02.360 | - If you could ask data one question

00:37:04.800 | and you would get a deep honest answer to,

00:37:07.320 | what would you ask?

00:37:08.600 | - What's Captain Picard really like?

00:37:10.480 | (laughing)

00:37:12.600 | - Okay, I think that's the perfect way to end it.

00:37:14.320 | Colin, thank you so much for talking today.

00:37:16.000 | I really appreciate it.

00:37:16.960 | - My pleasure.

00:37:17.800 | (upbeat music)

00:37:20.380 | (upbeat music)

00:37:22.960 | (upbeat music)

00:37:25.540 | (upbeat music)

00:37:28.120 | (upbeat music)

00:37:30.700 | (upbeat music)

00:37:33.280 | [BLANK_AUDIO]

Colin Angle: iRobot CEO | Lex Fridman Podcast #39

Chapters