Sep 22, 2026
Distinguished Guest: Prof. Matthew E. Taylor Visits Dresden
The Chair of Applied Statistics was pleased to host Prof. Matthew E. Taylor from the University of Alberta, Canada, following his arrival in Dresden on September 20, 2026.
Upon his arrival, Prof. Taylor was welcomed by Prof. Ostap Okhrin to discuss the itinerary for his visit. On the morning of September 21, Prof. Taylor visited the Chair of Applied Statistics, where he engaged in series of in-depth research discussions with Prof. Okhrin and the chair's PhD students.
A highlight of the visit was Prof. Taylor's presentation at the ScaDS.AI General Assembly , where he delivered an insightful keynote titled:
"Learning Well With Others: From Agent-Agent Advising to Human-AI Teams"
In his talk, Prof. Taylor addressed the challenge of learning from imperfect external knowledge, exploring topics such as transfer learning, curricula design, non-expert demonstrations, noisy feedback, and action advice. He introduced multi-agent advising frameworks capable of estimating advisor quality online and maintaining robustness against bad advice. Furthermore, he highlighted how these frameworks apply when humans receive guidance from imperfect AI tools emphasizing why human-AI teams often underperform without deliberate system design. Prof. Taylor concluded his keynote by identifying key open research problems for interactive learning in the era of foundation models.
The Chair of Applied Statistics and ScaDS.AI express their gratitude to Prof. Taylor for his visit and valuable contributions to the academic community.