How did scientists use AI to bring a virtual fly to life?

A new virtual fruit fly developed by scientists at HHMI's Janelia Research Campus & Google DeepMind is the most realistic simulation of a fruit fly created to date.
To develop the new virtual fly, Janelia Research Specialist Igor Siwanowicz first imaged various parts of an adult female fruit fly with a microscope & used computer software to construct an anatomically accurate virtual model of the outside body of the fly that includes the movements of the fly’s joints & appendages.
DeepMind researchers, including Senior Research Scientist Yuval Tassa, Josh Merel (now at Fauna Robotics), and Research Engineer Guido Novati, translated this virtual model into code that they fed into the MuJoCo simulator, a fast, open-source physics simulator created for robotics & biomechanics. The tool enables researchers to virtually simulate how something might move & interact in the real world.
To support the fruit fly model, the researchers made significant upgrades to the simulator, including adhesion actuators that model the forces generated by insect feet gripping a surface. The team was also lucky to have Novati design a new fluid-force model, which describes the forces acting on the fly as it moves through the air. The model can support all sorts of aerodynamic behaviors, including winged flight.
Next, Janelia Machine Learning Researcher Roman Vaxenburg built an artificial neural network and trained it on real fly behavior by feeding it information taken from videos recorded by fly behavior experts.
The fast speed of the MuJoCo simulator enabled the team to infuse the fly model with Vaxenburg’s neural network, which had learned how to move the fly in a realistic way to carry out an action. When the model is given a high-level command, like to walk forward at a certain pace, the neural network translates this command into low-level movements that an actual fly would use.
Next, the team hopes to create an even more realistic fly model by incorporating other parts of the fly’s anatomy and a realistic sensory system into their virtual insect. They also want to be able to use a real neural network, like the fruit fly ventral nerve cord connectome, to power the model.