Bosch Exec Dishes on the Company’s Renewed IoT Commitment
Can you tell me more about your background?
I’m really a tech guy — an electrical engineer with experience in signal processing. And I started at Bosch in 1992 doing research and then I stepped out, which is unusual, to create a startup — a fabless semiconductor company in Dresden in Germany.
Then, I re-entered Boch and for many years, I worked with multimedia infotainment systems — highly complex software systems — in development and general manager roles. Then, I moved into research again. I was president of corporate research for four years. And last year, I took the new board position.
Can you tell me more about your experience with multimedia?
I worked on the multimedia area when I started in the early days — working with digital audio casting and codecs and these things.
And when I reentered Bosch and began working on multimedia, I worked on infotainment systems. At that point in time, they were the most complex software systems in cars. We had huge development projects with more than 600 software developers
My task there was to find the right development methods, to find the right structure to execute those types of projects. And one thing that we did, which was quite radical at that point in time in the automotive space, was become the first company to introduce open source software in the automotive domain for this kind of system. That was more than a decade ago.
Open source has become a big trend recently.
It seems easy, but it’s not. We completely developed the approach and the process that you need for open source.If if you look in the details of the license regulations, it’s quite complex. But we see open source as a very valid and very interesting approach in order to use technology that was developed in communities. We successfully introduced infotainment systems based on open source systems for General Motors, for example, here in the U.S.
We also use IoT in our IoT stack solutions. In this space, our approach is to use as much as possible open software, open source software solutions and components, and integrate them in our solutions because we think this is a much faster way to introduce new technologies.
Speaking of other big trends, where do you think we are now in early 2019, in terms of AI adoption?
Most of the very impressive results we have seen [with AI], for example, voice recognition, image recognition, speech translation and things like that are based on machine learning. This is what you could call a narrow AI. It’s based on labeled data.
Today, I think our focus is to a large extent on machine learning. And in the next years, five to 10 years, I see an extension toward broader AI solutions. One of the major problems of machine learning is that you need a huge amount of labeled data. In many industries, it is a significant cost factor. So we need to develop AI that is able to learn without labeled data. This is one aspect of what I would call a real AI system, but that is really in development. Machine learning is quite established, but also there, the development is ongoing and we are very active in this.
How does Bosch’s industrial background shape the priorities for its research on AI?
An important element here is quality. You have to ensure that your products and systems have a certain quality level.
And this is an open research question in machine learning: how to develop the algorithms in such a way that they can give guarantees for their performance in the field. And this is very closely related to our approach. Our slogan is “invented for life.” As far as AI is concerned, that means: if you add AI technologies to products, we have to ensure that a high-quality level, explainability and robustness. This is one of the focus points of our people in the Bosch Center for Artificial Intelligence.
What sets apart Bosch’s IoT strategy from competitors that have more of an IT background?
If you see this whole space, there’s certainly a number of companies coming purely from the IT and the software side. And these companies are now trying to move into the physical domain because, in the end, you need products which are intelligent. And we are coming from the product side, the domain experience and also from the sensor side. And now we are extending our offering and our portfolio with an IoT stack and machine learning and AI. In the application of, for example, machine learning and AI, the domain experience is really the differentiating point. This is what you need in order to deliver high-quality exciting products for the customer.
What IoT application are you most excited about?
I think in the mobility space, autonomous driving, I would say, is the killer application for the IoT and for machine learning and artificial intelligence. So this is really the application that drives and needs this technology.
On the way toward full autonomous driving, we have driver assistance functions where we are very strong as a company. Also in the further development of these functionalities, we need these technologies: connectivity, Internet of Things and machine learning.
If we move to other domains, there is the [smart] home. I would say the home is a digital assistant which lives in various devices in the home. In the future, I see the whole space of home will have to become connected and the customers are expecting a seamless interaction. So, not with device per device, but with the home as a partner in the form of a digital assistant. And in order to do this, all the technologies we talked about are definitely needed.
The third area that we see, which is very important, is Industrial Internet or Industry 4.0 as we call it. Internet of things in this context means that things are, for example, connecting machines and machines. With sensors, they gather data and out of this data they can provide, for example, predictive abilities.
How do you see IoT changing product engineering?
IoT can be transformational for product development — for mechanical, electrical and software engineering.
The traditional internet, cloud services and also artificial intelligence play a huge role and they have influenced also the way we develop products. So, today we are moving in the direction of doing verification and qualification of products on the computer — in the virtual space and only at the last point in time you go really into hardware. So, this is one direction and here we also see that methods coming from machine learning and artificial intelligence are used in product development.
There’s also, on the classical product development side, a huge change ongoing in the way we develop products, in the qualification of products that you develop, manufacture and ship. In the future, products will be connected in the field and therefore they will also live in the field. The product development has [a] part in the field because customers expect that your products learn over time – that they get updated.
Product development no longer stops with the shipping of products. And that’s a radical change to the whole way we look at product development and product qualification. The product gets feedback and this feedback is used for updates. So you can have big loops for the development teams. You can have small loops. The product itself updates its software and its algorithms. But the important point is to understand that the product is not a dead piece in the field. It’s a living entity in the field. And this really changes the approach of developing, manufacturing and maintaining the product in the field.