4. Ai Systems Tübingen Meetup

We're happy to welcome you to our 4th meetup hosted by the University of Tübingen and IBM Research and Development. This time, we will have two talks:

1. "Data-driven applications in education domain and their privacy considerations",
   Efe Bozkir, Uni Tübingen
   Abstract:
   Artificial intelligence and machine learning form different and exciting possibilities in many domains including education. It is possible to infer visual behaviors and perception of teachers and students in digitalized classrooms using data from various sensing modalities. However, such data usually consist of personal and sensitive identifiers as well. This talk will give an overview on the data-driven applications in real and virtual reality-based classroom configurations considering privacy-preserving algorithms that are applied to such setups.

2. "AI assisted Air Traffic Control",
   Mudhakar Srivatsa, IBM T. J. Watson Research Center, Yorktown Heights
   Abstract:
   Air traffic, despite the recent dip due to Covid, is expected to grow 30-40% year on year. With the anticipated inclusion of UAVs into controlled air space it is anticipated that the congestion levels in air space will increase 10 fold. Air traffic control aim at safely navigating an aircraft using real-time control actions - such as changes to ground speed, heading (direction of travel) and altitude of an aircraft. This talk will describe an AI-assisted air traffic control. The risk sensitive nature of this environment calls for high level of explanations and counterfactual explanations, real-time responsiveness, the ability to present succinct actions to a pilot, besides optimizing for aircraft delay, fuel burn rates and weather conditions. This talk will present algorithms and a system architecture for anticipating separation losses between aircrafts and a lattice-based AI planner to recommend actions to avoid such separation losses. The evaluations are conducted against an air traffic simulator showing the ability to avoid separation losses, while minimizing the number of actions at 2-3x normal air traffic loads.

3. "Actionable Decisions from Data Science"
   Sebastian Fink, IBM Technology Sales DACH
   Abstract:
   The aim of many Data Science projects is to generate value by discovering new insights to support better business decisions. However, exactly how to make these decisions in an optimal fashion is often far from obvious. At the same time, poor decisions can eliminate any value added by Data Science very quickly.
   In such cases, Mathematical Optimization can help by offering better decisions quickly and consistently while considering a multitude of factors like predictions of demand and supply of raw material. We invite you to explore together how this fascinating technology works, what it can do, and the advantages it brings.
   Starting with easy-to-follow examples we look at interesting use cases, move to hands-on-modelling with Python and close with some time for discussion and your questions.