Artificial intelligence promises us to predict virtually anything: future crime scenes, future criminals, the next big management star. However, in the many stages of developing algorithmic decision making systems, different types of mistakes can occur. Thus, bots might take wrong decisions, they might even discriminate some groups. Only testing these systems will ensure that they decide well and fair. In her talk, Prof. Zweig will first explain how machines actually learn before showing some of the mistakes that can happen. The conclusion is that testing will be of utmost importance to ensure that bots decide best possible.
1) How do computers actually learn?
2) How do we decide whether they have learnt well enough? Testing the bot
3) Who will teach them right from wrong? Regulation of AI
Prof. Dr. Zweig has studied biochemistry and bioinformatics. Today she is a professor of computer science at the TU Kaiserslautern, where she heads the Algorithm Accountability Lab and organizes the field of study called "Socioinformatics". Among others, she was awarded a national teaching award ("Ars legendi-Fakultätenpreis in den Ingenieurswissenschaften und Informatik", 2017) and the Comunicator-price by the German Research Foundation and the Stifterverband (2019). With her Start-Up "Trusted AI GmbH", she consults parties, churches, companies, national medial authorities, and politics on the ethical development and embedding of algorithmic decision making systems.