The third episode of our mini-series »Artificial Intelligence (AI) and Statistics« is all about »Shallow Learning«. In this episode, our podcast team – consisting of Sascha Feth, Jochen Fiedler and Esther Packullat – explains what lies behind this concept. Together, we shed light on how insufficient training data and historical distortions (keyword bias) influence the performance of AI systems. We also question whether errors in algorithms occurred before the development of AI-supported systems – and how such problems can be remedied.
It soon becomes clear: without high-quality data, no AI can develop its full potential. To make this tangible, we bring concrete examples into play:
In the next part of our mini-series, we take a look at »Unsupervised Learning«. Here, models learn independently and discover patterns just like archaeologists trying to decipher an ancient language.