Name: A.I. robots are here, but are we ready for them?
Uploaded: 2021-11-09
Duration: 8 min 34 s
Description: Experts discuss the impact of these new technologies on our workplaces and on society.

Experts discuss the impact of these new technologies on our workplaces and on society. 

A.I. robots are here, but are we ready for them?

Conferences

Experts discuss the impact of these new technologies on your workplaces and on society. 

going to start with you James and I want you to just give us a set up the stage on maybe the impact of robots in the workforce. Maybe some of the facilities that are using them, like how has it changed the way people are working? Well, we've been tracking this for a while. And one thing that we see is that companies that invest in human and machine collaboration at the same rate as top performing firms in their sector have a revenue enhancement opportunity. We see that again and again, we've been tracking this for six or 7 years now when we look across industries, We see that uh that revenue potential opportunity to be about 38% on average. So really a significant uh revenue potential opportunity of course, most companies are not adopting Ai and robotics at the level of industry leaders. I think that's really important point. When we dig down into industries that are using robotics, we see, you know, some really interesting results sometimes coming out of time and motion studies, uh time and motion research. So, if you look in the automotive sector, uh we've seen companies like BMW for instance, bringing flexible and collaborative robotics into the automotive assembly, for instance, down in their spartanburg south Carolina plant And BMW has found that these human and machine teams, these robots and assembly worker teams are about 85% more productive and a lot a lot of that has to do with the fact that consumers these days have been demanding more customized vehicles and these flexible human robot teams are able to pull off and deliver that level of flexibility. Fantastic. And what that's setting the stage might be time to sort of dive into a little bit about what you all are doing. So I think in the second we should be able to get a video that's going to be playing a peter. Maybe you can just maybe give me a little bit of a flavor of what co variant is doing. Yeah, definitely. So called variant is a Silicon Valley based ai robotics companies. So what we build is what we call the coherent brain, which is the brain for robots. So robotics as the technologies have existed for decades. So what's being shown here is an A. B. B industrial robot that's jim, pick and place operations. If you go into a car manufacturing plant like James said in BMW for um, or other big automotive players, you will see a lot of robots, but what those robots are doing, they are doing precise pre programmed motion again and again, they don't really have the ability to perceive the world, see the differences and make different actions and take different decisions based on what they and that's the new path that we introduce A Bavarian essentially the Ai. So like how do you give this incredible mechanical electrical and control machine, the ability to react differently and to handle the flexibility um that customers require that the markets require today and explain to me what is it, how is it seeing, what is the noticing? Yeah. So um this is what the robot is doing is you can notice that the robot is actually doing a different motion every time and notice there is actually knocking an item off in order to pick it up. And so what's happening here is that there are cameras on top of the bean that is looking inside and then the cameras do the camera images is allowing the Ai brain to make sense of what the three D. Geometry that is seeing. What are the types of objects in front of it, reasoning about what is the best way to pick it up and learn from its own experiences. Like. So if I do this motion is a seat recall that if I do another thing isn't the seat, I also recall that and those become um the success and fairly signals that allow the robot brain to continue to learn and how is it, how long does it take it for to learn how to pick up? And I'm looking at so many different items there. I mean is it able to generalize a little bit? I mean it's definitely able to generalize significantly. So about half of the items that we are seeing. There are in fact items that the robot has never seen. Why is it able to tackle that? It's able to tackle that because the same Ai has seen millions of other types of items before. So the next new one is just marginally different from the millions of items that have seen before. And this is really the power of Ai. Right? So if you program the robot to do precisely the same motion if you introduce any change any difference in your environment or any customization that your customers need then you need to reprogram it. But instead if you have an Ai that can learn the underlying patterns that gives you the flexibility to handle new changes to adapt to new situations on the fly. Right? And now I want to shift over to you chris and talk about I. Robot. So in a second we're gonna start seeing video of some of the stuff you're working on. Maybe you can give me a run down a little bit on you know the type of things that I. Robot is working on. We all know the Roomba vacuum cleaner. I think that's probably everyone's you know that's like everyone's primary experience with robots. Give me a little flavor of what you're all working on. Sure. I mean I think um you know full to know about the Roomba of course which we launched in 2000 to a lot of people don't realize that we've been around since 1990 spent out of M. I. T. So we've been doing robots In all sorts of domains for for over 30 years. Um And then the journey that the home robots that were focused on these days that that we're really excited about. First off We there more than 35 million room is out there around the world. So that's a pretty large fleet to be thinking about the range of environments. Everyone's home is chaotic in its own unique way that we have to deal with. And how do we tackle those? How do you provide a robot that can go into the home? Figure out where it is like as in in the space of the home, know where it's cleaned, where it hasn't clean, how to respond to a request to go clean the kitchen and have you just know how to do that. So that's solving the wear problem. It's slam kind of a classical problem in robotics these days. What what the biggest opportunity and the most exciting thing out there is now taking it to the next level of the kind of the what? So it knows where it is in the home, how to get to the kitchen, how to go clean that particular area. But the, what is all about what's in front of me, what's around me? And I think this is where the Ai comes in, right? People aren't where slammer. It's like the legacy algorithms of how making robots move right? It's like generally how I would sort of paraphrase it, but what is ai how is this improving the capabilities of what uh the vacuum cleaners can do now. Yeah. So if you imagine that this robot, you're just seeing a robot picking up a lot of pet hair on the floor. Pet demographics is pretty big for us. One of the things that I love talking about sarcastically, but it's super important to our customers as pet waste. So we're applying machine learning deep learning to pet waste detection. Okay, so now you can imagine you have a robot in your applause got some sharing for that there. Yeah, we know your customer, we do. Uh So we those products in addition to many other things that knows where kind of the dining room table is or where your kitchen counters are. So you can actually tell the robot to go clean those areas. Uh but when it comes to pet waste, uh you want to avoid that typically. So I imagine that data set is dirty. You can imagine a lot of very smart deep learning engineers who spend days on end, you know, dealing with, you gotta you gotta training those models etcetera. So it's super important though if you imagine millions of robots around the world needing to robustly be able to detect, understand what's in front of them and then appropriately react to that avoid in particular. And it's a hard problem generalize that to the diversity of situations, but we put a lot of energy into it, our our machine learning or deep learning pipeline and things behind that we actually have what we call the poop guarantee. Pet owners official promise. Like if that we stand behind the technology legally binding if you if you have one of these robots and it doesn't successfully avoid, we'll replace the robot uh free of charge. Right? So this is a lot of problems in the carpet is not in the robot successfully avoided. What's that? I think the problem is in the carpet, not in the robot when you're successful.