Arash Azhand

The rise of Artificial Neuronal Networks (ANNs) as the majorly utilized AI technology seems unstoppable: Machines are now able to classify images in super-human speed, beat master players in the ancient Chinese game “Go” and the complex strategy game StarCraft 2. But there is an important question increasingly overshadowing the rise of ANNs: How do they work and come to their decisions? An example from Mount Sinai Hospital drastically shows why it is so crucial to answer this question: In 2015, researchers developed and trained an ANN with a huge set of patient records. It can predict the onset of schizophrenia extremely well. As this is usually very hard to diagnose, medical experts wanted to understand the prediction. But “Deep Patient” is so complex that not even its own creators can understand its decisions, which is extremely unsettling. This is why we’d like to motivate peers to make the coming decade the era of Explainable AI (XAI). Prerequisites: Artificial Neuronal Networks, Statistical theory, Mathematical understanding