Meet Patryk Straszewicz, a PhD student at CONNECT in the Tyndall Institute, a partnership between University College Cork (UCC), the Science Foundation of Ireland (SFI) and the Department of Enterprise Trade and Employment.
How did you get to this point in your life?
Even when I was in grade school, I became enamored by computers. I started my journey helping my IT teacher with the setups of Windows 3.11 on school workstations using MS-DOS (pre–Windows era Microsoft’s text-based operating system). Fast forward through my studies in Automatics and Robotics, I ended up programming HVAC (Heating, Ventilation, Air Conditioning) systems for high-speed trains at ‘PESA’, a Polish train manufacturing company. Later, I moved to Ireland to work for Boston Scientific, a medical device manufacturing company, where I programmed robots and devices on manufacturing lines. While working at the latter, inspired by a documentary ‘AlphaGo’, I started taking a lot of courses on Deep Neural Networks. I prepared a proof of concept for automated part inspection using, a type of image recognition AI algorithm called Convolutional Neural Networks (CNNs), which evolved into a larger project and led to the assembly of an AI team at Boston Scientific’s Galway plant. Convinced that AI would be a driving force in technological evolution, and thoroughly enjoying my engagement with it, I decided to devote my entire time to further research. Networks is a ripe field where AI can be applied to improve the quality of service so that is why I decided to become a part of CONNECT and start my PhD here.
How would you explain your research to someone who has no idea about your discipline?
Most of us are using WiFi hotspots and sometimes in crowded areas we encounter a lot of delays, and the internet connection seems surprisingly poor. We can compare this to being at a big party where everyone is trying to talk at once. It gets noisy and you have to repeat a lot and shout just for your friend to understand a simple sentence. In my research, I use something called “reinforcement learning,” which is a type of artificial intelligence that learns from experience. It’s like teaching a robot to navigate the party efficiently, learning over time which conversations are most important and then giving clues on when it’s the best time to talk or listen to avoid disruptions and where everybody will have a chance to express themselves. My goal is to teach the wireless network to manage all the devices trying to communicate. By doing this, everyone’s messages get through faster and clearer, improving what we call the “quality of service (QoS).” This means faster internet browsing, and smoother video streaming for everyone at the ‘party.’
What is the most challenging element of your work?
Because reinforcement learning works best with realistic patterns yet must take actions within the environment it aims to improve, running training on a real network would be highly disruptive and inefficient. This is why we are using network simulations. The quality of the trained AI model and how well it will improve the QoS partly depends on how similar the simulated digital twin of a network is compared to the real one that it was based on. There is a lot that goes into it, such as pathloss model that takes into account moving users and physical obstacles between them, but also the patterns in traffic sent. Not everything has to be totally realistic as AI is generalizing to some extent anyway, so the main challenge is finding a balance. It is important to figure out which parts of the simulation can be simplified, and which ones have to be more realistic for the model to perform the best in real scenarios while avoiding running the simulation training at a glacial pace.
What do you think could be the next defining trend in technology?
In the AI space till now what we could hear was: ‘Big data, big data, big data’ which typically meant large volumes of real data collected from various sources such as transactions, social media, sensors, and so on. However, this kind of data has its ultimate limit and it is called ‘now’. It is only the data collected up to the present moment in time. Of course, it will expand by every minute, but it won’t be fast enough for our data-hungry models. In the future, the simulations and synthetic data will give a sustained boost to AI. Apart from networks, good examples for this trend right now are simulated driving environments that can generate diverse traffic scenarios for training autonomous driving systems more safely and efficiently than real-world testing alone or synthetic patient data that help train medical AI systems, allowing for the development of personalized medicine and advanced diagnostic tools without compromising patient privacy. It will soon span into many different areas where AI is used until a point where the simulation-based, synthetic data volume will far outweigh the real data that we use mainly right now.
Is there a personal experience that changed how you saw the world?
Seeing the Matrix! And no, it is not entirely a joke. While when it came out you could label it as a sci-fi action flick with great underlying philosophies, right now the ideas expressed within are starting to get much closer to becoming a reality. Of course, I wouldn’t go as far as to say we live in a simulation but even based on the advent of simulations that I mentioned earlier we will most likely get to a point where we can simulate to some extent not only particular parts of our worlds but the whole social and non-social infrastructures including physics. It could give us the power to anticipate and solve problems that have not yet revealed themselves to us.
CONNECT is the world leading Science Foundation Ireland Research Centre for Future Networks and Communications. CONNECT is funded under the Science Foundation Ireland Research Centres Programme and is co-funded under the European Regional Development Fund. We engage with over 35 companies including large multinationals, SMEs and start-ups. CONNECT brings together world-class expertise from ten Irish academic institutes to create a one-stop-shop for telecommunications research, development and innovation.
ArticlesHumans of CONNECT
