In movie blockbusters, we learn that artificial intelligence can rise against humanity and sound the death knell of the world. Is this just Hollywood fantasy or already reality? And why is Elon Musk "afraid" of the AI future? In this talk, we will deal with failed AI learnings and the abstruse results that can be created by such learning attempts. Here we will have a look at difference areas of AI like NLP or reinforcement learning. Ever thought about a PacMan that is controlled by an evil AI and how such character would act? We will look at the errors that can occur during learning processes. Here we will not only see amusing results, but also scenarios that can be a real danger in a productive environment. In the conclusion will also take a closer look at the causes of these failed learning processes and discuss what can be done to avoid such problems and pitfalls.
Iryna Dohndorf is a software engineer at the Karakun, which is placed in the beautiful city Basel. Before joining the Karakun, I received my Doctor in Computer Science from the Technical University Dortmund in 2017 and did postdoctoral research since 2018. My research has been focused on formal methods in computer science, stochastic modeling, simulation and optimization. I have developed innovative models and algorithms to tackle the challenges in data-driven decision making under uncertainty and apply the models in multiple domains, including predictive maintanance, transportation, logistics, finance, and psychological science. I am a passionate interdisciplinary researcher and software developer and have been actively collaborating with cross-domain researchers, computer scientists and engineers. My most recent publication is 'A Blockchain-Based Approach to Provenance and Reproducibility in Research Workflows.'
After my postdoc, I worked as a software developer, researcher and scientific advisor at the university library of the Ruhr University Bochum, where I developed software using Java technologies, among others.