Skip to main content

Why Dyson is investing in AI and robotics to make better vacuum cleaners

Why Dyson is investing in AI and robotics to make better vacuum cleaners

Share this story

R&D at Dyson’s UK campus.
R&D at Dyson’s UK campus.
Photo: Dyson

When Dyson unveiled its new hair dryer last year, the UK firm emphasized over and over again the amount of research it had done to develop the product. A thousand miles of “virgin hair” was purchased for testing the $400 Supersonic, four years were dedicated to development, and total R&D spending topped $72 million. All that to make a device that, in our estimation, only performed slightly better than the much-cheaper competition.

But this has always been the Dyson way: invest in research, and make (expensive) products that out-perform the competition. And you can’t say it doesn’t work.

Sales have doubled over the past six years

Dyson says total company sales have roughly doubled over the past six years, and will reportedly hit £2 billion ($2.43 billion) in 2016. That growth is being fueled by markets in the Asian-Pacific region (as of 2015, the company’s biggest source of profit), and over the past couple of years Dyson has expanded its product line. The company has taken the motor technology it perfected in its vacuums, and repurposed it for bladeless fans, heaters, humidifiers, purifiers, the Supersonic, and its new robotic vacuum cleaner, the 360 Eye.

But where does the company go from here? As one former employee told the Financial Times: “There is no point in cleaning air further than we are currently cleaning it.”

The answer, says Dyson, is yet more investment, with an increased focus on robotics and artificial intelligence. Exactly what this means in terms of new products isn’t clear (Dyson keep its cards close to its chest), but in the past year, the company has announced a string of initiatives to underpin this effort, including a new UK campus which will double its workforce to 7,000 over the next five years, and a £330 million research facility in Singapore focusing on “connected technology and intelligent machines.”

To get a better idea of what Dyson is planning next, we sat down with Mike Aldred, the company’s head of robotics.

The following interview had been edited and condensed for clarity.

Mike Aldred with a deconstructed 360 Eye robot vacuum cleaner.
Mike Aldred with a deconstructed 360 Eye robot vacuum cleaner.
Photo by James Vincent / The Verge

So, Mike, what exactly do you look after at Dyson?

In essence I’m focusing on new tech and technology that’s coming along. 

There are three groups within Dyson: Research, New Product Innovation, and New Product Development. The researchers are the equivalent to your academics, looking at technologies we could used in future products; the NPI get it to the Tomorrow’s World stage [a famous BBC TV show about technology] — the demonstration video that works one in 10 times, but proves the point; and then the NPD team are the one that take that and make it robust and reliable. 

My brain is more about today, whereas the guys in Research are doing the “we should do a cheese-powered lawn mower” sort of thing. And I’m sitting there going, “Wow really?” and they’re telling us, “Cheese is the future!”

The 360 Eye has been Dyson’s first big foray into robotics. Where do you see robots finding a place in homes in the future? 

Firstly, I would say there’s a long way to go with vacuum cleaning. We’re not saying this is the perfect end product. My end goal has always been that I’d love for people not to be able to tell me what their robot looks like. So they come home and the floors are clean and they don’t know what it looks like because it just happened when they’re not there. 

How do you go from simply avoiding obstacles, to knowing what they are?

For robotics in general, there are two big things. One of them is visual understanding: understanding the context of the environment you’re in. At the moment, robotics is mostly about avoidance. It’s about covering the floor space while avoiding everything else. What we want to be moving toward is interaction. Because if you can interact, if you can understand the room, and its objects, we can start doing things like manipulation. And manipulation opens up lots of other functions. It means we can do more intelligent cleaning, like dusting. So you don’t want to dust, for example, your Ming vase, but you do want to dust your coffee table. But to be able to do that you need to be able to differentiate between the Ming vase and the coffee table. 

And then the second one is machine learning and AI. For me, the way I would describe it from a customer’s viewpoint, is when you get this product, it’s a fantastic product and it will clean your floors. But it will do the same job every time, and if you have a frustration, it will maintain the frustration. So, wouldn’t it be great that when it came out the box [and] had a frustration, that would eventually disappear over time because it learned to stop doing that. 

Dyson 360 Eye Robotic Vacuum
Dyson’s 360 Eye robotic vacuum.
Photo by Ben Popper / The Verge

What’s a good example of this sort of AI-powered problem-solving?

For example, a machine would recognize where your high-dirt areas are and choose to schedule a clean there twice, rather than once. These are the sort of areas you can imagine moving toward. It’s adaptive behavior, but without having to put the onus on the customer. 

What machine learning and AI will enable is for a product to learn for itself how to adapt, without the customer having to understand the product. It may need them to explain what they find frustrating in order to change, but they don’t have to understand the mechanism by which that change occurs. Too many people are worried about the mechanisms of machine learning, and nothing about “How is my customer going to benefit from this?”

But what about the physical side of things, like actually cleaning the Ming vase. Is that something Dyson is thinking about? 

In terms of the vision side of it, it’s one of our main things that we do. In terms of the physical, we’ve got a room with 2,000 mechanical engineers who are all at the top of the game. So the resource pool available for the mechanical side is actually more established in the company.

As far as we know, Google is still trying to sell Boston Dynamics because the company couldn’t create a viable consumer product fast enough for Google’s liking. Do you think people expect too much from robotics companies too fast? 

I started working [on the 360 Eye] in 1998. And at that point we thought, naively, that it was going to take us three years. And after three years we’d produced [prototype Dyson robo vacuum] DC06. It had 70 sensors, it had three processors, it had 54 battery cells, and weighed about 15 kilos. It was completely unmanufacturable and very expensive. We hadn’t produced a product, we’d produced a demo. So we know how much work these things take. 

The DC06 robot vacuum cleaner was never sold to the public.
The DC06 robot vacuum cleaner was never sold to the public.
Photo: Dyson

I think what’s disappointing — and if I take my Dyson hat off and put my roboticist hat on here — is that so many companies are just not investing in technology. They’re not investing in innovation. What they’re investing in is their top line within the next two years. And if you only worry about the money you make in the next two years, you’re never going to innovate. I think it’s fair to say, that nobody at Dyson set out to develop a hair dryer. It’s having an open mind set to innovation that enables you to spark those things.

Are companies like Dyson expected to iterate faster than they used to, because of the pace of technology change?

Yes. [But] I think connectivity enables it. A more connected product means your innovation can come from something other than just mechanical iteration. It can come from over-the-air updates. The other thing I would say is, we don’t necessarily follow that trend of having to iterate that fast. We will produce a product when that product’s right for market. What we don’t do is play to the timescale. We will produce it when it’s ready, and when it’s better. We don’t just knock something out every three months for the sake of it. 

Dyson Supersonic hair dryer
Dyson’s $400 Supersonic hair dryer.
Photo by Vjeran Pavic / The Verge

Is Dyson interested in making products for the smart home, expanding your lineup into a wider, connected ecosystem?

We definitely are looking at how connecting products brings benefit to a customer. But we’re not jumping in with both feet at a point where it’s a relatively immature domain.

We have connected purifiers, connected robots, and you could imagine how pretty much all of our product range could be connected. But the first question we ask ourselves is: how do we avoid gimmickry? How do we put engineering ergs where they’re most valuably spent, rather than just strafing people with apps? We are in that process where we’re taking a step back and saying, “Where can we genuinely provide benefit?” We’re not just going to connect things for the sake of connecting.

And this is my opinion, not Dyson’s, but I don’t think there is much genuine benefit from connected stuff like — no, I won’t say their name! Because they’re in that gimmickry phase. We had some products that people tried, and they found it fantastic for that first week and then you ask them three weeks later and it’s not being used. 

What about voice integration? Has Dyson thought about connecting products to Alexa or Google Home?

It’s the sort of thing we’re looking at. But we’re trying to answer for ourselves, is that something people would use? If it is, these are areas where we’re ready to go. We have products that are connected, and we understand the infrastructural requirements to do a connected product. We have the back-end systems to support it. But just because we have the ability to go heavy into the connected… we’ll do it when it’s right, and we’ll do it when it’s providing genuine benefit. 

So you have these components. You have your motors, your batteries, and your computer vision. Where else can you combine this tech to be useful in the home?

The question I would put back to you is: what do you get most frustrated doing in your home you don’t want to do?

The genuine thing that we’re looking to provide people is time. The utopian goal for me is [solving] the things in my home that personally frustrate me: ironing, dusting, and washing up. I’m not saying we’re doing them, but those are the sort of targets we set ourselves at Dyson. We don’t come out and go “I’ve got a really cool motor, I’ve got a really cool vision system, how can I use it?” You start with problems, saying, “What frustrates me.”