Artificial Intelligence: Smarter Than Man?

This is possible because several developments are currently coinciding:

This abstraction, this use of experience and this finding of creative, new possibilities is what gives these systems cognitive abilities that people have long since had. The algorithm learns patterns from data and draws conclusions from them without knowing a defined context.

Software that can learn in this way has a similar structure to the neural networks in the human brain. In the brain, synapses form the nodes in a finely meshed structure that is closely connected. Deep learning algorithms consist of simulated nodes arranged in layers one after another, unlike in the brain. The input goes through the individual layers successively. Thus each layer takes on special tasks during processing. For example, in image recognition, the first layer processes the image information pixel by pixel, while the next one specializes in analyzing corners and edges. This is followed by an analysis of oblique lines or round shapes and so on. Only at the end is the object identified as “tree”, “cat” or “street sign”. The number of involved nodes is immense: Current AI systems organize billions of artificial neurons in roughly 30 layers.

The term deep learning refers to the depth of this layered structure. Deep learning algorithms also allow for feedback and correction loops and are particularly well-suited to specializing in the detection of patterns or deviations in huge data quantities. This makes them unbeatable for use in image recognition programs, for example. In addition to enormous computing power, they also require training on the basis of a veritable flood of data. After all, an algorithm is only as good as the data with which it was trained.

However, training AI systems is one of the big challenges. How do you test a system that is designed to solve unforeseen problems? Virtual training and software-in-the-loop methods will make a contribution to overcoming this hurdle. On this basis, autonomous vehicles could soon significantly reduce the number of traffic accidents. What that means and where artificial intelligence is already being used in the car today is the topic of the second part of our AI Special .