
Mr Beyerer, what potential is there for the future? Let me give you a concrete example.
Jürgen Beyerer: Robots that learn through interaction are particularly suitable for acquiring skills for certain tasks that were previously carried out by humans at low cost. In some sectors in high-wage countries, such as Germany, it is no longer possible to produce economically and competitively with human labour alone. A good example of this is harvesting tasks in agriculture, which currently relies on a large number of harvest workers. However, the primary goal must not be to rationalise jobs, but rather to increasingly compensate for the labour shortage caused by our demographics. Think, for example, of robots that support carers and relieve them of tasks that can easily be automated. In addition, such adaptive robots can be used for tasks that are dangerous, unhealthy, inconvenient or too strenuous for humans. Corresponding examples include underwater work at great depths, sorting waste for recycling purposes or physically demanding manual tasks such as sanding in manual trades.
Safety plays a major role in social interaction. How can this be guaranteed?
Jürgen Beyerer: There are many aspects to safety when interacting with robots. On the one hand, it concerns the physical safety of humans, which can be guaranteed with different concepts - by limiting the momentum and kinetic energy of robotic movements, by low masses and speeds of robot extremities and by limiting forces through soft structures or force and torque sensors. Another aspect of safety concerns the protection of the privacy and data of interacting humans who are observed by the sensors of robots. Unauthorised evaluation and misuse of data must be prevented. To this end, the robots must be equipped with protection mechanisms from the manufacturer that have been verifiably tested and certified. And finally, the robots must also be cyber-secure so that they cannot be taken over or misused by third parties and the aforementioned security features cannot be undermined.
What challenges remain?
Jürgen Beyerer: There are several challenges for the use of interactive learning robots. One is the acceptance of such technologies in close proximity to humans. Fears could arise that their own work could be rationalised away, or that the robot could observe humans and the corresponding data could be misused for other purposes. There could also be concerns about whether the robot can fulfil its task well enough and reliably. And of course it depends very much on the actual task whether the use of robots can already be technically realised or whether they can provide the desired performance and at the same time be procured and operated economically.
Detailed expertise on the integration of robotics with Artificial Intelligence (AI) methods can be found in the white paper ‘AI in robotics. Flexible and customisable systems through interactive learning’ (in German) published by Plattform Lernende Systeme.
The interview is released for editorial use (provided the source is cited © Plattform Lernende Systeme).