Prof. Dr.-Ing. Jens Lambrecht, Thomas Staufenbiel, Dr. Eugen Funk © Gestalt Robotics

04 October 2019

"The industrial application of AI methods often in practice fails at the point of 'data hunger'."

Gestalt Robotics' business model is based on the project-based development of robotics software with a focus on perception, cognition and interaction. This orientation makes it necessary to use methods and tools from highly technological areas. The Berlin-based company with its three foun­ders Jens Lambrecht, Eugen Funk and Thomas Staufenbiel has been developing customized soft­ware for industrial and service robots since 2016. Gestalt Robotics meets the constantly growing demand for automation solutions from companies in various branches of industry. Robotics soft­ware is also an option for the retail and e-commerce sectors, such as logistics order picking and ware­house operations. #ai_berlin met Dr.-Ing. Jens Lambrecht for an interview.

In your opinion, what makes Berlin so strong as an AI location?

Berlin is a key science and business location in the heart of Europe, with excellent universities and AI research as well as an innovative power that meanwhile also includes AI. Especially in the field of autonomous mobility and robotics, Berlin is increasingly becoming the centre of Germany. When I moved to the city ten years ago, the situation was different: at that time, I was fully expecting to have to leave Berlin after completing my doctorate in robotics in search of an appropriate job. Today things are completely different: Berlin is currently developing into a large AI and robotics location with an active ecosystem, and I am very proud to be able to help shape it. Gestalt Robotics is also trying to actively support this, for example with meetups or our podcast series "Berlin Robotics".

What makes Berlin different from other locations?

I see Berlin not only in politics, but also in the field of innovation as a source of inspiration for technical developments. Almost every large company in the technical sector is looking at the tech­nology and start-up scene in Berlin or is already active here beyond a mere representative office. 

And even though the economic region around Berlin differs greatly from large parts of southern Germany, Berlin is a cosmopolitan city with an innovative character which cannot be found any­where in the rest of Germany. In addition, there is an influx of excellently-trained international spe­cialists together with well-trained specialists from Berlin's universities and technical colleges. Berlin is also able to retain young specialists beyond their training due to the unique atmosphere in the city. The activities of Berlin's universities and technical colleges to support young founders are as diverse as they are unique. Local economic and technology promotion also provides the best conditions for internationally competitive innovations in the deep-tech sector.

How did you get the idea to found (your business) / to start (your initiative)? Why in Berlin?

Gestalt Robotics was not born as a typical start-up. Instead of starting with an idea or a prototype and looking for financing, we were able to start directly with the first paid customer projects. In 2016, the happy circumstance arose that my co-founder Thomas, who had previously been with Airbus in Munich, moved to Berlin. Together with our mutual friend Eugen, who had just com­pleted his doctorate on 3D image processing at the German Aerospace Centre in Berlin-Adlershof, Gestalt Robotics was founded.  

Originally, our only goal was to be able to make initial requests for software development in the field of automation and AI. However, with our combined competences in automation, autonomous mobility, software development and AI, our team, customer base and project scope grew rapidly. On the basis of work experience in large companies, we try to form a counterpart at Gestalt Robotics and give our first employees room for their own ideas and innovations. We still maintain this philosophy, meanwhile with more than 15 employees and still completely self-financed and independent. We started with a small office at the TU Berlin and currently, due to our growth, we are about to move into our third office.

What is special about your company?

Gestalt Robotics sees itself as an innovation service provider for trend-setting solutions at the inter­face of classical industrial automation technology and AI. We thus combine industrial know-how with strong competences and integration experience in the field of AI, especially machine learning. Since our foundation, we have been working in partnership with our customers to solve real indus­trial challenges with tangible added value for the customer. Accordingly, we are specialists for the industrial use of AI, including integration and certification from individual software services to complex solutions and systems. 

The seamless interaction of our technologies from the areas of control, image processing, user interaction and AI is the foundation for the tailor-made implementation of flexible AI applications. With the help of our modular system and cross-sectional technologies, we combine and network these precisely. We work together with partners in the field of communication and implement AI applications also on the basis of 4G and 5G mobile networks as well as cloud and edge computing.

What is your focus area?

With our technologies we focus on two large application areas: production automation and autono­mous mobility. The solutions in these areas include AI-supported image processing, intelligent robot applications, adaptive assistance systems and a platform service for autonomous mobility

At first glance this appears to be a broad field, but it is based on a uniform technology platform, so that we can provide scalable solutions also for individual requirements with our modular technology modules. 

On the technological side, I would like to emphasize on the one hand current developments in soft­ware services to enable autonomous indoor and outdoor navigation via 4G/5G networks and edge computing. The corresponding solution for controlling mobile transport systems was developed in several joint projects together with Deutsche Telekom and is currently already in productive opera­tion with industrial users.

Another area I would like to highlight is the data-efficient application of modern AI methods. All advantages of modern machine learning methods are confronted with the need for large amounts of data for the training of robust and accurate algorithms, especially in industrial applications. Our experience shows that the industrial application of AI methods often in practice fails exactly at the point of "data hunger". Be it that data acquisition and the associated engineering are connected with effort and costs or that provision of the data, for example with visual recognition of rare errors, is not possible. For this reason, we are increasingly concerned with methods and technologies which enable us to use AI procedures effectively and reliably in an industrial environment, even with small amounts of data.

Where do you get your ideas from?

Many of our employees come from applied research. Whether with a doctorate or without, what our team has in common is a thirst for knowledge and constant further training and development. We encourage our employees to invest part of their working time in their own internal research projects. Some of these internal projects have meanwhile turned into product prototypes. Our EPIC technol­ogy, for example, for which we recently received the Berlin Deep Tech Award, also stems from a corresponding internal project. This is a special few-shot learning method that makes it possible to use modern AI methods in a very data-efficient way and to train them with just a few examples.

What is the significance of your company/programme event/your initiative for the AI industry in Berlin?

In the meantime, as representatives of the AI industry in Berlin, we almost see ourselves as "old hands". We are observing the growth of the AI scene in Berlin with great benevolence and are happy about the development of a large ecosystem. In the context of the AI hype, however, we also see increasing risks for the topic, due to a certain superficiality without technical background knowledge or media-fired expectations. Accordingly, I am very critical of exaggerated expectations of AI and of people who consciously promote these.

What do you recommend to people/companies who want to deal with the development of artificial intelligence?

My personal conviction regarding the success of technical innovations is based very conservatively on a good mixture of theoretical background knowledge and practical experience. It is always important to be sceptical whenever a hype comes into being. In this sense, the topic of AI is always a complex one and will always remain so. I would be cautious with exaggerated promises that make AI look like "magic".

Otherwise, real problems and challenges have to be identified as starting points for AI instead of just using AI for its own sake. If this is the case, there is not much left to be said against "just try it out", and the way is to work with reliable partners on future-oriented solutions.

What are the challenges behind artificial intelligence?

I'm not a big fan of the term "artificial intelligence"; there is too large a portfolio of different tech­nologies and methods behind it. Again and again we see challenges in demystifying the subject of AI and dampening exaggerated expectations. In addition to all the technical advances in recent years, the image of AI in society among the general public is characterized by fiction and sensa­tional journalism. This is also evident in the middle and upper management of large companies. An answer to this can only be the collection of "real" experiences and a good technical (further) educa­tion. 

Furthermore, AI must not be an end in itself, but must bring real economic, ecological or social added value to industrial use. For Gestalt Robotics, this concrete potential must be identified together with our customers on a cooperative level and specifically raised with tailor-made techni­cal solutions.

Thank you for your time.