Machine learning agents learning to climb up stairs. The agents started without any knowlegde about the task and try to maximise their reward. The closer they get to the target -the top of the staircase- the bigger their reward. With each iteration their knowlegde grows and they come closer to their target.
Each time the installation reboots the machine learning agents start from zero and the title of the installation therefore changes. In the beginning their movements seem strange and unnatural, but after a few hours when they start climbing the stairs people start to see human traits in them. The work explores this moment of recognizing ourselves in artificial entities and artificial intelligence.
developed in the new media class at UdK berlin | Supervisors:
Prof. Joachim Sauter