Animals like fruit flies or zebrafish were previously considered relatively simple stimulus reaction machines that react to certain ambient stimuli with solid behaviors. Scientists at the Max Planck Institute for Biological Cybernetics in Tubingen, however, have found evidence that in the brains of these animals is the lovely internal state that controls the behavior. This could not only explain the amazing flexibility of biological brains, but also help the Ki research further, reported Technology Review in his December ie (can be ordered on the kiosk or online).
Biological brains are strongly ruck-coupled networks that allow it to develop very complex stimulus reaction schemes. However, the number of possible schemes is also astronomically rough. Drew Robson and Jennifer Li from the Max Planck Institute for Biological Cybernetics in Tubingen suspect that internal governance steer the behavior in certain corridors – and thus a way could be drastically reducing the complexity of the problem.
Changing activity
Robson and Li examine shots of the brain activity of fish larvae. With the help of fluorescent neurons, they could show that the patterns of the brain activity of the larvae differ systematically in exploration and hunting. In addition, they were able to identify a group of nerve cells in the center of the fish brain, which was active as long as the animals were hunting mode – even if there was just no prey in which. The larvae only hedged with the hunt when the cell group was no longer active and changed back in the exploration mode that comes with its own continuing activity pattern in the brain.
Ki-pioneer Jurgen Schmidhuber, Scientific Director of the Institute of Artificial Intelligence in Lugano sees it in parallel to the function of certain neuronal networks. "There are already similar neuronal networks that can switch between different activity standards back and forth", he says. "Young examples are the artificial video players for Dota of Openai and Fur Starcraft of Deepmind." In many millions of players, these algorithms learned to control various agents in strategy video games – partly at the level of people.
"The progress of the KI in strategy games like Starcraft is remarkable", Stop Robson. "But it is just as remarkable that human players are competitive, without having played millions of games." Evolution optimized biological neural networks for the survival over millions of years. "But not simply the number of neurons has grown, but also their care of voting and specialization", Says Robson. Internal conclusion could play a central role in this fine-tuning.