compare A Computer with the Clever Hans is not a compliment. The horse heard this name, made before the First world war in Germany, made headlines because it supposedly, arithmetic, and could spell. Later it turned out that the bill cunning Orlov only the human body interpret the language particularly well and so the expectations of the questioner was able to meet.
Sascha Zoske
sheet-makers in the Rhein-Main-Zeitung.
F. A. Z.
a electronic brain can be to Clever Hans, if it is for the wrong reasons, the correct answers. With this phenomenon have Kristian Kersting and his colleagues from the Technical University of Darmstadt. The computer scientists wanted to show how the trust in Artificial intelligence can be increased. That is what is lacking currently, often, because it is not apparent how the learning networks of their predictions.
The researchers explain the Problem in an example: A computer program to acquire the ability to recognize photos of a horse. It is lined with Thousands of horse images. Quickly, the Artificial intelligence seems to be to learn to identify the Ungulate also on previously unknown recordings. But then it turns out that the Software reaches on the basis of the external features of a horse to your judgment – but with the help of the Copyright notices in the images, the references to a horse page on the Internet.
