Scientists at the Sunshine Coast University in Australia have found that highly developed artificial intelligence (AI) is capable of destroying humanity or causing catastrophic damage to it, even without malicious intent. Experts conducted a thought experiment with the hypothetical SantaNet system, which aims to send gifts to good children around the world in one night. This is reported by Science Alert.
The researchers found that the danger arises already at the stage of determining which children are considered good and which ones are disobedient. This goal is achieved through a large-scale covert surveillance network that monitors children's behavior throughout the year. SantaNet will be guided by its own moral judgment, which can lead to discrimination, massive inequality and human rights abuses. In addition, the network can encourage children to behave badly in order to reduce the burden on themselves, or raise the bar for “good behavior”.
With nearly two billion children under the age of 14 living in the world, SantaNet could turn the entire earth into a giant toy factory. This problem, called the “paper clip problem,” was first posed by artificial intelligence researcher Nick Bostrom, who suggested that AI, designed to increase the efficiency of paper clip production, would take its purpose too literally.
In addition, sending out gifts on Christmas Eve with the help of an army of drones could lead to open confrontation when countries prevent drones from flying freely in their airspace. Also SantaNet can be vulnerable to hacking by terrorist groups. Another challenge is interacting with AI working on climate change, food and water security, and ocean degradation. It can lead either to conflict or to cooperation aimed at reducing the world's population.
