Thanks to the branch of artificial intelligence known as machine learning a team of scientists has cracked the code of bats.
Machine learning is a branch of artificial intelligence aimed at introducing the ability to learn into machines. There is more and more talk about this technology and one of the reasons is because there is more and more research that delves into it. By doing so, researchers have discovered that an algorithm can create your own rules by training it conveniently. This is what a team of specialists from Tel Aviv University has done. Its raw material: the bat code.
The aim of his study was to decipher the communication of these flying mammals. Your squeals are difficult to properly catalog, as it happens with other animals, such as some species of primates. That is why scientists at Tel Aviv University have used artificial intelligence techniques for their investigations.
Team he recorded for 75 days two groups of 11 bats that remained in two separate cages. Then the specialists went through all the videos trying to find out which bats screamed and to whom they were addressed, as well as the object of their communication –whether it was food, sleep or sex– and what the result of the discussion had been.
On the other hand scientists nurtured a machine learning algorithm with over 15,000 calls of seven females with their corresponding meanings. When asking the program about the recorded video, its hit ratio was high. 71% of the time correctly deduced which bat was making the sound and 61% guessed the topic of the discussion. His result was harder to figure out. Only 41% of the time did the algorithm succeed in this aspect.
In this way scientists have learned some lessons about the bat code. Until now it was thought that a series of common squeals meant ‘out of here’, but they have turned out to be somewhat more sophisticated. Actually these sounds contain a lot of information.
These discoveries would be the beginning. Scientists now expect to decipher much of the bat code based on the patterns identified. And this would allow learn more about the behavior of these animals.
Images: vitch and visual_paradox
