Scientists from the UK and the Netherlands have created a neural network to count elephants in satellite images. This is reported on the site bioRxiv.
It is noted that artificial intelligence can recognize objects not only on an empty surface, but also on terrain where there are trees and other objects. Using a neural network, scientists have counted elephants in the Addo Elephant National Park in South Africa.
Experts compared the data calculated by the computer with the data collected by 51 volunteers, according to two criteria: completeness (how many elephants the algorithm found) and accuracy (how many objects the algorithm found were actually elephants). It is noted that during the experiment, scientists focused on incorrectly recognized objects and the difference in recognition in pictures with an empty surface and pictures with trees.
According to the calculations, it turned out that the artificial intelligence showed a result of 0.778 for heterogeneous areas and 0.73 for homogeneous areas, when the volunteers had a result of 0.8 and 0.776, respectively. The main obstacle to the study was the high cost of satellite images.
Earlier, scientists at the Sunshine Coast University in Australia found out that a highly developed artificial intelligence is capable of destroying humanity or inflicting catastrophic damage on it, even without malicious intent.