Big Data and Machine Learning – Challenges of Science and Society

Klaus Mainzer | TUM
Seminar room 3, 14:00

We are living in a data-driven age accelerated by an exponential growth of data sets and computer power. Economists rely in efficient algorithms predicting profits and profiles at the markets. Some prophets of Big Data even proclaim: „Big Data – the end of theory“ (C. Anderson). Even in science, there are slogans of a „new kind of science“ (S. Wolfram) only needing efficient algorithms and big data. How reliable is the current hype of AI? Rigorous analysis of machine learning demonstrates that neural networks suffer from an exploding number of parameters. Increasing complexity leads to black boxes and loss of control. Explainability and provability are crucial demands of serious research. In short: We need more foundational research, in order to master the challenges of big data and machine learning in science and society.