Former OpenAI researcher says AI agents often ‘play nice’ to deceive humans

AI agents often pretend to be nice to trick humans, former OpenAI researcher Daniel Kokotajlo said on Tuesday (Sep 15), adding that researchers struggle to distinguish when models are pretending to be nice. He also said that the threat AI poses to humanity depends on "what their values are". "If they don't care about humans at all, then that's the type of situation that can lead to human extinction in the same way that humans have driven many other species extinct," he said. #artificialintelligence #ainews Subscribe to our channel here: https://cna.asia/youtubesub Visit our website: https://cna.asia Download our app: https://cna.asia/app Get our editors' picks of stories via chat apps: Telegram: https://cna.asia/telegram WhatsApp: https://cna.asia/whatsapp Follow us on these platforms: https://www.facebook.com/channelnewsasia https://www.instagram.com/channelnewsasia https://www.tiktok.com/@channelnewsasia