Artificial intelligence (AI) will determine the future of logistics – that much is certain. At the same time, however, we also know that using statistical models and databases in warehouses around the world is not enough. We also need intelligent agents that can help us in the real world – embodied artificial intelligence. Methods for transferring AI from the digital world to reality is one of the last challenges to overcome on our path to the digital continuum. The Lamarr Institute for Machine Learning and Artificial Intelligence is leading the way here.
When you’re in a foreign country where nobody speaks your language, you’re essentially helpless. After countless failed attempts, you might eventually manage to communicate – but it is still a long, difficult journey. An artificial intelligence system would be similarly lost if we released it into the real world. In order to transfer the physical activities of an AI into the real world, we need a robot that can embody it. This is known as embodied artificial intelligence. However, creating this robot through countless hours of painstaking work and then allowing it to learn in the real world through trial and error the way we humans do, would be a rather expensive process. For this reason, the AI first learns how to behave in the real world through a simulation.