Study explores Nervous systems of insects enthuse competent future artificial intelligence systems

Study explores Nervous systems of insects enthuse competent future artificial intelligence systems

The study explores the Nervous systems of insects enthuse competent future artificial intelligence systems

The thesis discusses the roles of the Fruit Fly nervous system in food search/results that are useful for the production and regulation of artificial intelligence. Zoologists at the University of Cologne have researched insect nervous systems to explore the concepts of biological brain computing and potential applications for machine learning and artificial intelligence. Specifically, they studied how insects learn to equate sensory knowledge in their surroundings with food rewards, and how they can later recover this information in order to solve complicated tasks such as locating food. The results indicate that translating sensory input into brain memory will motivate potential machine learning and artificial intelligence systems to solve complex tasks. The research was published in the journal PNAS.

Zoologists at the University of Cologne have researched insect nervous systems to explore the concepts of biological brain computing and potential applications for machine learning and artificial intelligence.

Living species have an extraordinary ability to deal with the challenges presented by complex and changing environments. They are able to generalize their knowledge in order to adjust their actions easily as the situation changes. Zoologists studied how the fruit fly nervous system regulates its actions when looking for food. Using a computer model, the equations in the fruit fly nervous system were replicated and analyzed in response to the scents originating from the food supply. ‘We initially trained our model of the fly brain in exactly the same way as insects are trained in experiments. In the simulation, we provided a particular scent with a reward and a second scent without reward. The model easily learns a stable representation of the awarded fragrance after a few perfume presentations and is then able to locate the source of this scent in a spatially complex and temporally dynamic environment,” said Dr. Hannes Rapp, a computer scientist who developed the model as part of his doctoral thesis at the UoC Institute of Zoology.

Thus the model created is capable of generalizing from its memories and incorporating what it has previously learned in a radically new and dynamic molecular odor landscape, whereas learning required only a very limited database of training samples. ‘We use the unique properties of biological information processing in nervous systems for our model explained Professor Martin Nawrot, senior author of the research. ‘These are in particular, the quick and concurrent processing of sensory inputs by means of brief nerve impulses, as well as the development of dispersed memory by the simultaneous adjustment of multiple synaptic connections during the learning process.’ The theoretical concepts underlying this paradigm can also be used for artificial intelligence and autonomous systems. They make it possible for an artificial agent to learn a lot more quickly and to adapt what he has learned in an evolving world.

Share This Post