EAISI jubilee AI Symposium

TERESE HELLSTRÖM & VICTORIA BRUNO

Title | Computer aided pancreatic cancer diagnosis with user-centered design
To be shared

MARKO PETKOVIĆ

Title | Zeolite Simulation using Deep Learning
To be shared

Marko Petković is a doctoral candidate in the Materials Simulation & Modelling (MSM) group in the department of Applied Physics at TU/e as well as the Data Mining (DM) group in the department of Mathematics and Computer Science. Marko’s work focuses on creating AI solutions for modelling nanoporous materials, such as zeolites.  

CELINE BUDDING

Title | Explaining what large language models know 
Given the impressive performance of large language models (LLMs), there is an increased interest in not only studying their behavior, but also their underlying processing. For example, it has been proposed that LLMs do not just perform next-word prediction based on surface statistics, but that they in fact acquire something like knowledge (Meng et al., 2022). Yet, it remains unclear what kind of knowledge LLMs might acquire and when this can and should be attributed. Taking inspiration from earlier debates between symbolic and connectionist AI in the 1980s and 90s, I propose that tacit knowledge (Davies, 1990) provides a suitable way to conceptualize potential knowledge in LLMs. Tacit knowledge, in this context, refers to implicitly represented rules or structures that causally affect the system’s behavior. In particular, Davies’ account of tacit knowledge provides clear constraints that should be met in order to attribute this knowledge, and could thus help better evaluate what LLMs learn and how they work. 

About Celine Budding. I am a PhD candidate at the Philosophy & Ethics group at TU/e. In my PhD project, I investigate what large language models like ChatGPT "know" about language. Specifically, I take inspiration from philosophical debates in AI in the 80s and 90s and argue that large language models can be considered to have tacit knowledge of language. More generally, I am interested in what AI systems learn from the data, how we can evaluate and explain their performance, and how we can translate insights from various academic fields–machine learning, cognitive science, and philosophy–to effective regulation and policy for AI. To this end, I am also involved in standardization of AI through the NEN, the Dutch standardization organization. Before joining TU/e, I completed an MSc degree in computational neuroscience at the BCCN Berlin. → Back to program

TILMAN NOLS

Title | More than just a cool tool: Humans, AI and the Teams of tomorrow
While many companies are drawn to AI and robots to support their decision-making, many theorists believe that human-technology collaboration is further developing in the future of Industry 4.0. In light of advanced autonomous capabilities, AI will be able to transcend from a mere tool to a full-fledged teammate, that occupies formal roles and responsibilities. In these team scenarios, humans and AI will become increasingly interdependent, complement each other strengths and weaknesses, and rely on each other to achieve organizational goals.   
However, past research shows intricate difficulties that stand in the way of implementing the workforce of tomorrow: cognitive, affective, and behavioral barriers limit human-AI teams' ability to realize their synergy, including communication difficulties, opacity, restricted adaptability and trust issues. Accordingly, the present pitch will shortly review the theoretical underpinnings and challenges of human-AI teams and provide a conceptual framework to optimize collaboration opportunities across varying, complex environments. 

Tilman is a work and organizational psychologist with work experience in the semiconductor industry. He obtained his research masters from Maastricht University, Leuphana University Lüneburg and University of Valencia and is particularly interested in issues of innovation, change management and team science. In his PhD at the Human Performance Management Group, Tilman studies the future of tomorrow's workforce by exploring the collaboration opportunities between artificial-intelligence-powered technology and humans. In that regard, he specifically focuses on the mechanisms that enable humans and AI or robots to team up and coordinate their actions in complex environments. Doing so, he aims to facilitate a dynamic, holistic and multi-disciplinary approach to tackle his line of research. → Back to program  

JEROEN OVERDEVEST & XINJI WEI

Title | Mitigating Automotive Radar Interference using Model-based Deep Learning 
In advanced driver assistance systems, radar sensors are often regarded as the most reliable due to their robustness against adverse light and weather conditions. However, the proliferation of automotive radars in semi-autonomous vehicles has led to an increasing problem of radar-to-radar interference. Within the RAISE/RADAR project, we collaborate with NXP Semiconductors, and one of the challenges we aim to address is interference mitigation. Our objective is to improve interference removal by replacing classical signal processing algorithms with model-based deep learning techniques. Using the known underlying physical models, we leverage model-based, data-driven models to enhance the performance and reliability of radar systems in diverse driving scenarios.

Jeroen Overdevest received the B.Sc. and M.Sc. degrees in electrical engineering from the Delft University of Technology in 2015 and 2018, respectively. After pursuing the M.Sc. thesis at NXP Semiconductors, he joined the Algorithms & Software Innovation Group in NXP Semiconductors, Eindhoven, The Netherlands, in 2018. His research interests include signal processing, deep learning, and waveform design for mm-wave automotive radar.  In October 2021, Jeroen started his PhD with SPS within the RAISE (Robust AI for SafE radar signal processing) program, where he seeks model-based deep learning approaches for mitigating real-world radar artifacts. 

Xinyi Wei received the B.S. degree from Shanghai Maritime University, Shanghai, China, and the M.S. degree in electrical engineering from the Eindhoven University of Technology (TU/e), Eindhoven, The Netherlands, where she is currently working toward the Ph.D. degree with TU/e. Her research interests include automotive radar interference mitigation and deep learning. → Back to program