Juan Gamella has built special miniature laboratories (mini-labs) that are suitable as test beds for new AI algorithms. “The mini-labs provide a flexible test environment that delivers real measurement data. They are a bit like an experimental field for algorithms, where researchers can test their AI beyond simulated data in a controlled and safe environment,” says Gamella. The mini-labs are based on well-known physics, so researchers can use this knowledge to check whether their algorithms arrive at the right solution for a variety of problems. If an AI fails the test, the researchers can make targeted improvements to the underlying mathematical assumptions and algorithms early on in the development process.

Gamella's first mini-labs are based on two physical systems that have essential properties so that many AI tools have to cope with them under real conditions: Exactly how the mini-labs can be used depends on the problem to be tested and what the algorithm is supposed to do. One of his mini-labs, for example, contains a dynamic system such as wind, which is constantly changing and reacts to external influences. It can be used to test AI tools for control problems. His second mini-laboratory follows well-known physical laws for light. It can be used to test an AI that automatically learns such laws from data to help scientists make new discoveries.

The mini-laboratories are concrete devices that are about the size of a desktop computer and can be controlled via PC remote control. They are reminiscent of the historical demonstration experiments used by researchers from the 16th century onwards to present, discuss and improve their theories and findings in scientific societies. Juan Gamella compares the role of miniature laboratories in the development of AI algorithms with that of a wind tunnel in aircraft construction: when a new aircraft is developed, most of the design is initially carried out using computer simulations because it is cheaper and more efficient. Once the engineers have agreed on their designs, they build miniature models and check them in the wind tunnel. Only then do they build a full-size aircraft and test it in real flights.

“Like the wind tunnel for airplanes, the mini-labs are used for safety testing to ensure that everything works at an early stage when we move from simulation to reality,” says Gamella. He sees testing AI algorithms in a controlled environment as a crucial intermediate step to ensure that an AI works in complex, real-world scenarios. The mini-labs provide this for certain types of AI, especially those designed to interact directly with the physical world.

The mini-labs help researchers to investigate the problem of the transition from simulation to reality: they provide them with a test environment in which researchers can carry out as many experiments as they need. This transition problem is also relevant for the overlap area of robotics and AI, where AI algorithms are often trained to solve tasks in a simulated environment first and only then in the real world. This increases their reliability.