AI is an expandable definition, which involves learning structures that are able to detect patterns and apply the learned patterns to predict or transform something. When we talk about AI, we talk about deep neural networks or reinforcement learning systems that are capable of solving large, complex problems like object detection, object classification, or autonomous driving. These applications are ruled by deep neural networks with millions of parameters.

Data is the new oil, but „AI ist the new electricity“ - Andrew Ng. Artificial intelligence helps to mine valuable knowledge from data. Deep neural networks got a boost in popularity in 2011 when the neural network AlexNet solved the ImageNet competition (detection of 1000 classes of objects in images) with an error rate of 16%. Before deep neural networks ruled this competition the average error rate was way above 25%. Since then the error rate has been decreasing to less than 5%. The success of deep neural networks relies on the huge amount of data that is necessary to train the millions of parameters. Since the amount of data is continuously increasing, the range of applications for AI becomes wider and deeper. While AI has been a long time an instrument to solve a single very specific task, AI is more and more developing into a generalized approach for problem solving.