Artificial intelligence is a trend popular in the media. However, rather in connection with large companies and investments. How did you get to it?
Marián Bódi: Big companies are dealing with artificial intelligence because it is trendy. From our experience, big companies talk more about it, while smaller ones experiment more and do real work. This topic requires fresh thinking and that is often contradictory to established corporate processes. People in corporations usually have long-term experience, but in different fields. However, AI is a new field and junior people have more experience with it.
Marek Šebo: We started modestly, with small projects for local clinics and IT companies. School and hobby projects provided us with skill in AI. We learnt a lot about practical IT, especially during first years from our colleagues and clients. Nowadays, we have successful projects for global companies behind us. However, we do not let up on improving.
What is needed to start working on artificial intelligence at a professional level?
Today, technology is so advanced. You only need to take a fifteen-minute course and you are able to run a model for digits or faces recognition. It takes years of practice to be able to create products that are really useful and solutions that go deep. This kind of experience can only be gained by trying and playing. All this development is often about intuition about how the neural network will behave in a given situation and how to set it up to avoid certain mistakes. Of course, some mathematical foundation is also necessary to start the intuition.
Why is it that today results can be achieved so quickly within artificial intelligence? Is it due to growing technology performance ?
M.B.: There are different types of products. The first prototype can be delivered quite quickly. The implementation in the real world is more difficult. We create products which can work in the real world. Our clients prove this. Often they already had a pilot – a machine with cameras, or some software that has worked in laboratory conditions. But after launch, this solution was unusable for workers in production.
M.Š.: It is simple to create some kind of solution. But it takes more than a few clicks to bring a solution implemented in the real world. On the other hand, systems are more simple now, than they were when we were starting this company. Nowadays, we have frameworks available for training a person and it is fast. When we started, even the most used deep learning tool was very new. When an error occurred, it took even two days to resolve it. We have grown together with the entire AI development community.
How do you choose industries for technologies and problems that can be solved with artificial intelligence?
M.Š.: Our history was random to a certain level. Gradually, we found industries in which we began to specialize and where it was initialized by some idea. Next, we found a client, thanks to which we went deeper into the topic. Later, we realized it might be of value to another client. We work on a specific project, with a specific client, and we try to listen to him and solve his problem. But in the context, we see other opportunities or ways to develop it further.
M.B.: We started in healthcare. We developed video recognition technology for a clinic in Prague to improve the success rate of artificial insemination. Our technology recognizes certain markers on the developing embryo and helps the doctor decide which embryo to choose for implantation. It is important, because the choice significantly affects the success of whether a woman conceives successfully or not. Our system performs an automatic analysis and provides the doctor with information.
Andrej Kozák: We realized that use is needed in industry as well. Production lines contain not-ok pieces which need to be discarded. System of cameras with neuron network or other similar technology is ideal for this. This is used to discard the not-ok piece automatically, with no human interaction. As complete newcomers, we were the first to approach industrial enterprises. This is how we got our first projects and relatively large know-how in manufacturing issues. We are still approached by companies and we work on projects that are also in other industries. This keeps us improving ourselves and allows us to identify other potential topics.
Do you also work in other areas?
M.Š.: Yes, we do. One of our key issues is human motion analysis. Using machine vision, we created a system that analyzes the skating technique of hockey players. It is used by NHL players and clubs, like Detroit or Vancouver, for training the hockey players. In the past, players needed a couch, to get technique improvement suggestions. Today, the player uses a skating training station. The player is recorded by cameras and our system make an analysis of basic parameters and brings improvement suggestions.
M.B.: This is a perfect proof of the usefulness of image processing technology and cooperation of Slovak companies on a global market. We do not work on it alone, we cooperate with HDTS – a Slovak company. Our responsibility is software, their responsibility is mechanical and methodological aspects and commercialization.
Read the original text here