Models of systems in physics usually start with elementary processes. New work with a neural network shows how models can also be built by observing the system as a whole and deducing the underlying interactions.

This video shows the fly model reproducing a flight maneuver (spontaneous turning) of a real fly, executing commands to walk at a speed of 2 cm/s while turning left and right, and the model imitating a walking trajectory of the real fruit fly, which includes walking at different speeds, turning and briefly stopping. Credit: Vaxenburg et al.

By infusing a virtual fruit fly with artificial intelligence, Janelia and Google DeepMind scientists have created a computerized insect that can walk and fly just like the real thing.

The new virtual fly is the most realistic simulation of a fruit fly created to date. It combines a new anatomically accurate model of the fly’s outer skeleton, a fast physics simulator, and an artificial neural network trained on fly behaviors to mimic the actions of a real fly.

Their AI is able to recognize patterns in complex data sets and to formulate them in a physical theory. The development of a new theory is typically associated with the greats of physics. You might think of Isaac Newton or Albert Einstein, for example. Many Nobel Prizes have already been awarded for new theories. Researchers at Forschungszentrum Jülich have now programmed an artificial intelligence that has also mastered this feat. Their AI is able to recognize patterns in complex data sets and to formulate them in a physical theory.

In the following interview, Prof. Moritz Helias from Forschungszentrum Jülich’s Institute for Advanced Simulation (IAS-6) explains what the “Physics of AI” is all about and to what extent it differs from conventional approaches.