Exclusive: iRobot’s Colin Angle on STEM, Autonomy, and Startup Advice

Image: iRobot

June 20, 2019      

iRobot has long established itself as one of the top consumer robotics companies in the world, with millions of cleaning robots vacuuming, mopping, and lawn mowing (later this year) at homes worldwide. A major player in the Boston-area robotics community, iRobot has spawned the creation of several other robotics startup companies in its almost 30-year history.

Root Robotics system Colin Angle interviewBut rather than sitting on these achievements, the company continues to expand its robotic reach. Today, the company announced the acquisition of Root Robotics, a startup that develops the Root coding robot (pictured, left) aimed at the STEM (Science, Technology, Engineering, Math) education market. Initially developed by a group within the Wyss Institute at Harvard University, Root is a two-wheeled, mobile robot that teaches coding and 21st-century problem-solving skills to children as young as four years old. The robot is available for purchase for $199 through this iRobot site.

iRobot said its acquisition helps diversify its educational robot product offerings and furthers “the company’s commitment to make robotic technology more accessible to students, parents, and educators.

Robotics Business Review spoke with Colin Angle, chairman and CEO of iRobot, about the acquisition and his goal of having the next generation learn coding, as well as topics covered during his recent keynote at the Amazon re:MARS conference in Las Vegas.

Colin Angle iRobot CEO interview

Colin Angle, CEO and chairman, iRobot.

Q: What was it about Root Robotics that made you want to acquire them?

Angle: As you might imagine, I’ve seen a lot of educational robots in the wild, and iRobot is also very passionate and focused on STEM education. What has me really excited about Root is its ability to teach coding. So bringing the convening and exciting power of robotics to the table and point it at helping people learn to code, which is a real shortfall in classroom-based education today.

We have an educational system that is a desert, as it relates to teaching computational thinking and coding. Our educational standards have not evolved particularly materially in decades. So when thinking about how we can do more with iRobot’s commitment to STEM education, the idea that we might be able to teach serious coding with robots was something we found to be incredibly powerful. And Root does that in some really unique and super-cool ways.

I’m excited about this because we have a huge problem in educating our young people with the skill sets they’re going to need to survive. A lot of times when people ask me career advice, what I say is, ‘It actually doesn’t matter what your child is interested in, teach them to code because you can make more money in an hour of coding than a week or two waitressing. The best thing you can do for your aspiring ballet dancer is to teach them to code in their free time.’ I think that iRobot’s acquisition of Root has put my money where my mouth is.

Q: What is unique about Root’s approach than other STEM-based robots or educational robots on the market?

Angle: The programming dimension of Root and what makes it really cool is that you can program across three levels, growing alongside users as their skills advance. At Level 1, it’s a purely icon-based language, ideal for beginners and non-readers coding Root’s inputs and behaviors. At Level 2, it infuses text into graphical blocks, providing greater precision with values, logic, variables, and flow control statements. At Level 3, it reaches a complete text block interfaced designed to teach the structure and syntax of Swift, the coding language of all iOS devices. [Editor’s note – iRobot said future plans for Level 3 include adding professional coding languages such as Python and JavaScript]

Colin Angle interview Root Robotics whiteboard

The Root coding robot can move along a whiteboard, providing more coding opportunities for students. Image: iRobot

You could free program of course, but it’s all backed up by a lesson plan that if you complete, you can code. So this is not a power-puff device – this is something that will teach, in a do-it-yourself fashion, an individual how to actually program with good, basic skills if you run it all the way from the beginning. So I found that to be really exciting.

You want the robot to be rich enough and operate in a sophisticated enough environment to allow a growing level of challenge and activity that will keep the programmer engaged for years, and have those challenges hard enough that it forces the programmer to write some pretty interesting code as they get better at it.

Q: Do you think that other systems don’t take into account the growth of the coder, that once they finish a project they either get bored with it, or it’s too hard and they stop?

Angle: With a large percentage of educational robots, no one gets past ‘Hello World,’ and then there’s the type of programmable robot that are in the category of ‘Minutes of Fun!,’ where once you play with your different block chains in a few different ways, you’re kind of done with what it can do. So trying to find that balance requires this flexibility between the levels of programming on Root.

Going beyond autonomy

Q: In your keynote at Amazon re:MARS, you spoke about how autonomy is not just intelligence. Do you think a lot of companies are building towards autonomy and then they just stop?

Angle: The purpose of the talk was to engender a conversation around the fact that designing intelligence, ignoring a rich connection to the people for whom the robots are serving, isn’t very smart. Thinking through how the different devices in the home, and how your home can collaborate to create meta-competence, is also something that is not widely thought through and investigated.

There’s a sprint by [Amazon] to connect up to 60,000 devices [to Alexa]. In six months they’ve doubled from 30,000 to 60,000. That’s incredibly fast adoption, but the richness of the interaction is lagging the completeness of their catalog. You heard them talking about some of the things they’re doing to increase the richness. I’m trying to raise the bar even higher, where you think about the true system intelligence required for the home to understand what’s going on, and to do the right thing.

That means a baseline of autonomy that just works. Layered on top of that is directability, collaboration, and then a whole new level on top of that – system-level goals. I think it’s a ladder of competence. If you don’t have good autonomy and you’re only directed, you end up making a remote control car, where you’re joysticking the thing around, or you’re giving it hyper-specific commands because the robot can’t be trusted to operate on its own.

If you have a rich autonomous platform, then you can add these higher-level instructions, like ‘Clean the living room’ or ‘Clean around the couch in the living room,’ and that is very additive to the experience. The directed room cleaning feature on the i7, and now the S9 and M6 very quickly became the most used digital feature we’ve ever done – even more than scheduling. It’s a well implemented game changer for how people use the robots. That is validating that if you have good autonomy and then you add this attention to directability on top.

Q: Are you happy with the autonomy that you have on all of your systems?

Angle: I think that it’s good, and we’ll continue to get better. One of the things we are able to do is learn, because we can remember where we are, and we can remember more about the situation that existed when the robot got stuck – and don’t do that again, or try a different strategy.

iRobot Roomba S9 Clean Base

Roomba S9 Clean Base. Image: iRobot

Another thing that the Clean Base did was change the definition of mission completion’s importance. So, mission completion is how many missions did your Roomba go before it got stuck. If every three or four missions you are picking up the robot anyways to empty it, going three or four missions without getting stuck was pretty good. If you got stuck every once in a while, then you’ve probably done a mission or two or three, you just empty the bin and put it back. That was part of Roomba ownership.

Now, you can go a year without touching it and run hundreds of missions. That’s a whole other level, because your expectation now is that you don’t touch your Roomba – turning it off or it getting stuck is death. It’s very different. Even going into a room where there’s a person watching television is a really dangerous thing for your robot, because someone might say, ‘Oh well, I don’t want that. It’s too loud.’ – and you’ve broken the paradigm. So the challenge of figuring out how to live for long periods of time in harmony with the people in the home suddenly becomes super important. The autonomy part of the robot just got 50 times harder with the advent of the Clean Base.

Personality vs. task performer

Q: What are your feelings about whether robots in the home should have personality, versus robots that just stay in the background? A keynoter at the show said 85% of Roombas have a name, and we’ve all seen videos of the cat sitting on the Roomba, or different skin options to add color to the Roomba.

Angle: The fact that you call your robot by name is not core to its mission. It actually benefits the mission, because if it does get stuck, you have some feelings for it, and it’s not, ‘Stupid robot, I’m going to return it.’ It’s more ‘Oh, how do I move this chair or table so the robot doesn’t get stuck there next time?’ So the empathy the person has for the robot helps the robot succeed at its task better, and creates a little bit more of a partnership. Even though you didn’t buy your robot to go spend time with them, your empathy for the robot’s mission is helpful and having the robot do its job better.

There are many things we could do with our design of Roomba to make it more character-like, of which we do none. In fact, we go the other direction, and try to make sure that when you look at the robot, it looks like a serious professional cleaning device that you can have faith is going to do a good job. The naming just happens regardless of what we do, and we’re happy and accept it because I think we net positive on that relationship.

Q: Do you feel pressured by either customers or other partners to add additional features that go beyond the cleaning parts of the robots? For example, adding a camera and creating a security function, or a screen for telepresence?

Angle: No, but occasionally we talk about it. For a successful consumer product, you should have clarity as to why you’re buying it. If it’s a security camera and vacuum robot, if I don’t want both, I won’t buy it. If I do want both, maybe I’ll consider it, if I think that I’m not compromising. And of course, you are, because if your camera is on the floor, you’re not going to be able to see most of the things you want to see. If you’re buying it as a vacuum and it’s also a security system, you might not be comfortable having cameras in your home, so you probably wouldn’t really want somehow thinking your camera was streaming video into the cloud. So it’s not a logical combination.

There are logical combinations that will come out that if I build a robot whose mission is to talk to you, it might also make a good security robot. You want to be able to have a camera up high enough both to look at you and to look onto the tops of tables and things like that, so you can actually use it as an effective roving camera. We’ll see some successes if someone can actually get it right and make it feel like a good security system or a good telepresence system. I think Ava Robotics is a great enterprise conferencing system, and when we made our choice to focus on the home, the guys that were working on it at iRobot left the company and it’s now Ava Robotics.

Q: Do you feel like iRobot has a voice in the discussion around whole-home automation?

Angle: I’ve been saying that spatial awareness is crucial for intelligence for a while. I would say that in the last six to 12 months, people are listening. As soon as you have 10 smart devices in your home, expecting the consumer to program them [is difficult]. Probably the answer is actually once you’re beyond two. Or maybe even one.

So the idea that a robot could go find them and help program them is increasingly of interest. Three years ago we had less than a million connected robots in the market, and at the end of this year we’ll have almost 10 million connected robots in the market place. So we went from being irrelevant to this part of home conversation to big and central to delivering on the promise. That’s a huge change.

iRobot’s legacy in robotics

Q: When we meet with startups in the Boston area, inevitably the founders have either worked at iRobot, or at Kiva Systems. Can you talk about this legacy that the company has built in the robotics community?

Angle: There are 25 companies in the Boston area founded by ex-iRobot employees. And I view that as a point of pride. At iRobot, we taught people the importance of a pragmatic approach and appreciation of the customer – how to actually go and take technology and turn them into these ideas, into physical products. Part of me hates it, but part of me loves it.

Colin Angle iRobot mission statement

Image: iRobot

Someone comes into my office and says, ‘Colin, I’m going to start a company.’ The first thing I tell them is it would be hypocritical of me to tell you not to. Now here’s what you need to know. I’ll do my best to give them a realistic view of what they’re signing up for. A number [of companies] that spun out of iRobot, we’ve actually invested in. We view it as part of our legacy and we keep growing and finding more great people.

I think as the CEO and founder, I have the luxury of being able to say what I said and ensuring that iRobot’s talent is not diluted by people frequently leaving. Our attrition rates are very low. We have probably more than a dozen people who left and came back, and there’s a little bit of ‘the grass is always greener on the other side,’ but we also create a great environment. People leave feeling supported. If things go well, great – and if they don’t, they’ll come back and they’ll carry that story with them, that ‘You guys don’t know how lucky you have it.’

Q: What advice would you give to startup companies that are looking to scale up? Several companies we cover, for example, are pretty successful at the startup stage, but then they run into some speed bumps as they’re scaling up.

Angle: Hire a great HR person before you are doing it because you have a giant problem and you need it. I think around 30 people, 20 to 30 people, you should be hiring an HR person. Most people don’t until they’re at 100, and they do it because they have a giant, colossal mess on their hands, and they realize that they are going to spend the next two years playing catch up. Many of them fail because the environment got toxic. It’s very avoidable, so that’s a big thing.

The other advice I usually give entrepreneurs is that your first idea – the idea you started your company – you’re either super lucky, or wrong. So having the flexibility of buying yourself enough runway to figure it out is critically important.

Stretch that bit of seed capital, whether it comes from your wallet or an investor, like it was the last dollar you’re going to get. When things are good, raise money to extend the runway, because things don’t always stay good, and if you ever need the money, you won’t be able to find it.

You can easily be penny wise and pound foolish with your capital strategy. Companies go under when they run out of money. Rarely do people say, ‘I’m bored, I think I’ll give up now.’ It’s, ‘I have no choice but to stop because I can’t make payroll.’

There’s a lot of companies that miss on the business model. They succeed on cool, and haven’t yet really identified a customer need of sufficient severity that people are going to open their checkbooks. It’s a tough game, building robots. It’s expensive. But it’s also super fun and cool, and if you do it well, you change the world.