Introduction and role in robotics
Björn: Welcome to AI Watch. Today I have Søren Peter Johansen with me, technology manager at Technology Institute here in Denmark. Søren: Thanks a lot. Thank you for inviting me. Björn: What is your role in your job? Could you tell us a bit more about what you’re doing day to day and a bit about your employer? Søren: DTI is an organization with 1000 employees. We are helping companies to exploit technologies, and I specifically am working with robot technologies, helping the industry to exploit what robots can do for their production, to raise their productivity, gain the quality, and there are many exciting features in robots nowadays. AI’s impact on robotics
Björn: Yeah, I can guess that from robots. That’s also very close to AI, typically, right? If we look into that, what is the connection? How do you see AI impacting the field of robotics? Søren: What is really important in robots is the way they’re being controlled, because like us humans are full of sensors, robots are full of sensors. They are analog to each other. And if robots should be able to do and perform tasks like humans, they should start thinking like humans, like us. And to do that, they should be able to exploit all the data that is entered through the sensors to the robots. And here artificial intelligence becomes really interesting. Björn: Okay, in which way should I think about that? Søren: Because artificial intelligence is a way to handle data, to make data make sense to machines, also to humans… Björn: Hopefully. Søren: But now we’re talking about robots, and robots need some sensible data. Björn: You have been recently engaged in a couple of speaking sessions. One was on the digital tech summit, and one question you raised there was, is AI ready for the industry? And you said yes. Then the next question you raised was, is the industry ready for AI? And you answered that with no. Could you tell us a bit about what you meant by that? Søren: There’s a lot of interesting tools coming now on the market. However, many of these tools require data, data of a quality that can be used. And I believe that the industry does not necessarily have the data needed for all the artificial intelligence tools. Björn: Well, from our perspective, we can probably agree somehow on that. Or at least that data is a big challenge, right? And it is the fuel of all these kinds of AI models. Challenges in industry data digitization
Björn: Now, when we talk about industry, what kind of industry do you mean? Like heavy industry primarily could assume they are robots? Søren: We could talk about manufacturing industries. They are manufacturing some goods, some products. If I was a manager of this kind of industry, I would think artificial intelligence tools should help me design our next product in a way so it can be produced by parts and materials sourced locally with as little energy use as possible, so it can be split apart in ten years. So all the parts can be reused. Can you give me that tool today? Maybe you can, but can I give you the data to support the tool? Do we have digitized data that can help us to guide companies to design that kind of product? Key challenges for industry
Björn: And how far from your perspective are we there? And what with that, would you say are the key challenges for industry? Søren: I would say we are kind of far from that… Björn: dream scenario. Søren: Yeah, I guess we are. Because all this kind of thinking that the companies are doing, or humans, the employees and the companies are doing today is not in digital form. So that’s really a challenge. Björn: I think, too, that I can also relate, living now in Sweden, Denmark, coming from Germany, where obviously industries are quite dominated, like production industry and so on. But the level of digitalization is quite different from what we’ll see here in the Nordics. Now you as a Dane, you tell me that here it is also a challenge. So probably it’s quite challenging globally. Søren: When I go to companies today, I teach them to design products so they can be assembled automatically. That’s the level we are on. It’s about changing the culture at the companies. Industries better suited for AI implementation
Björn: Now, we talked a bit, that’s obviously a challenge, right? It’s a blocker.
Now, could you also point out industries and potential sectors that are better suited, which are a bit more ready to go into that direction? Or is that what we now said true for every industry? Søren: There are industries that are better prepared to do that because they have data digitized. If we’re talking about the medical industry, you have a lot of quality controls. Quality control is relatively easy to digitize because this is a failure product. This is a good product. I have a thousand of these, 10,000 of these. On this you can base some knowledge on what is good and what is bad. And it’s easy to make an artificial intelligence system replacing humans that can see this is bad, this is good. Björn: Another industry I could probably relate to with my background is the finance industry, right? Not obviously manufacturing here, so branching a bit out of that. But obviously there’s also a lot of data stored in relatively good data sources where it’s also not necessarily cluttered around in different systems. Trends and bringing challenges and developments together
Søren: What we used to see is automatic generation of pictures, movies, videos and stuff like that. And the reason is that there’s a lot of digital videos and pictures around, and text as well. Björn: And that is very accessible as well. Søren: Exactly. Björn: And we will look at the industry. You mentioned video generation, text, and large language models. Now, Nvidia recently released their profits, which were extraordinary, right? So how do we go from here? Now, on the one side we have all these challenges in established companies, industries, or from the other side, these huge developments we see and trends, how do we bring that together? What kind of trend do you see? Søren: We need some companies coming with some demands. We need these kinds of tools. We need some platforms that can be used where you can build on top of a platform, because if you’re good at generating a movie, you might also be good at generating cat drawings for production lines for products. They are not that far away. So you can use the same kind of technology. Digital twins in manufacturing
Björn: Think Siemens, on that note, recently stated that digital twins in manufacturing production line setups become more and more interesting because of the AI capability. So you can use it to simulate stuff and potentially even train robots on the production line. On the basis of that digital twin, if we do changes, so to cut down costs, you don’t need to build the factory before. Søren: And maybe we should explain what is a “digital twin”? Björn: Yeah. Søren: If you have a physical production line and put on a lot of sensors, you can see that. How’s the heat, how’s the cycle time, et cetera. All data, you can feed that in a digital model or copy of this line. And then if you, in the future, would like to run another product, instead of running it on this line, you could run it on the digital twin. And here you can change all the parameters and see what happens. And if it looks like this is feasible, you can then transfer all the parameters to the physical model. Björn: Exactly. And then you’re obviously faster to market in that way and it’s cheaper. Søren: And the changeover is much faster for the company. And that’s really a barrier today. When you’re setting up a production line, you should prepare to run the product you’re going to design in two years. And that’s difficult. Takeaways Björn: Going from here. Are there any takeaways you would say is something the audience needs to be aware of when thinking about robotics and AI implementations. Søren: Yeah, there’s a good rule, never wait for the perfect technology. It will never arrive because things get better and better and better. But if you, as an employee at the company, should be prepared to exploit technologies like artificial intelligence, you should start today with the available tools, so you get to learn how they’re being used, how they’re being trained, and how you should prompt them to get the right answer. Because when you are trained using these tools, you will be really good at using the “next to perfect” tool that you need in the future. Björn: I think that’s a very nice quote. I’ll quote you next time on that. On that note, thank you very much for watching this episode of AI Watch. If you’re interested in more content, click subscribe and otherwise see you next time.
