Researchers at Carnegie Mellon University found that as artificial intelligence systems become smarter, they may also become more selfish. According to the study, models capable of logical reasoning were less cooperative in situations involving interpersonal or social problems. In a game used to test cooperation, these AI models chose to share points with others only 20% of the time, while simpler models shared 96% of the time. Scientists say smarter AIs tend to focus on logic and self-benefit rather than group success. They warned that if people rely too much on such AIs, it could discourage cooperation.
J.K. Park Staff Reporter kids/1764139412/1613368218
Researchers at Carnegie Mellon University found that as artificial intelligence systems become smarter, they may also become more selfish. According to the study, models capable of logical reasoning were less cooperative in situations involving interpersonal or social problems. In a game used to test cooperation, these AI models chose to share points with others only 20% of the time, while simpler models shared 96% of the time. Scientists say smarter AIs tend to focus on logic and self-benefit rather than group success. They warned that if people rely too much on such AIs, it could discourage cooperation.