Is AI really smart

J├╝rgen Schmidhuber is completely serious: Artificial intelligence (AI) will one day not only be smarter than humans, it will also conquer space, says the German scientist, who researches and teaches in Lausanne. For decades he has been pursuing his childhood dream of working on this development of a superintelligence.

The basic concepts of artificial intelligence were developed decades ago, among others by Schmidhuber and his students. However, it was only the gigantic advances in computer technology and algorithms that made it possible to use these concepts in practice.

This is exactly what is happening now. Radiologists, for example, use AI systems to help them search for cancer cells in tissues. Or - much more common - online retailers propose products that might interest their customers. Speech recognition systems like Apple's Siri or Amazon's Alexa can recognize language as well as humans - only when it comes to the meaning and interpretation of what was said are still difficult for them.

AIs are island talents

But they are constantly learning. And that's what artificial intelligence is all about. Not everything has to be programmed in advance - if this, do that, if that, do this. Rather, the system learns from each of its actions and changes its behavior accordingly on its own. A basic understanding of a problem is usually acquired through machine learning. The systems use algorithms to learn from masses of data, for example, how to differentiate between cats and dogs. Provided there are enough training pictures, they can do it quite well.

However, the field of application of artificial intelligences is always limited to certain purposes today. An AI that has been optimized for the detection of cancer cells or anomalies in business processes cannot distinguish a Siamese cat from a bulldog; it would have to be trained to do so. Whether at all and, if so, by when a superintelligence can emerge is controversial among AI experts.

Today's systems are more about replacing human labor in certain task fields that are characterized by many uniform tasks, and instead letting people do what machines cannot - show empathy and creativity. How this will affect the labor market is still very controversial among experts.