Abstract

AI and machine learning are on a path to becoming a commodity, at least according to the AI industry. Their automated model building platforms enable a lay person without AI knowledge to upload data and retrieve AI models. However, such platforms typically only automate structurally simple and well defined problems like regression, classification or clustering. In other words, they only help us, if we already know what the problem is.

Large industrial analytics problems on the other hand are often initiated by a high level question like “How can we reduce the number of faulty cars we produce?” Such problems require an analytical exploration in a large team of engineers, technicians and researchers working in different disciplines across different departments.

The seminar will take you through such a large analytics problem and discuss the practical challenges of solving it. The big question is how maths and AI may be able to help address the challenges and facilitate the analytical collaboration between groups of analysts and computers. Rather than provide all the answers, the idea of the seminar is to trigger a discussion of the problems and possible solutions.

The second part of the talk is concerned with education in data science. Given the possibilities of digital media, Martin questions the current static sequential format of teaching material and textbooks. His observations apply more generally to scientific publications as well as reports of analytical projects in industry which links back to collaborative analytics.

About the Speaker: Martin completed his Master in Maths and Computer Science followed by a PhD in AI from Karlsruhe University (KIT). After this Martin spent 15 years at BT Research Labs in Ipswich working on AI algorithms and their application. As Chief Analytics Researcher, he left BT in 2016 to become a Professor for Data Science at HTW Berlin. He is currently on a sabbatical at CCIMI .