‘Jailbreaks’ bring out the evil side of chatbots
Story by Kathryn Hulick in Science News Explores
Researchers break chatbots in order to fix them. This so-called red-teaming is an important way to improve AI’s behavior.
Read the full story in Science News Explores:
‘Jailbreaks’ bring out the evil side of chatbots
Sometimes breaking a bot can help developers improve its behavior
Quote
“Is this the right way to develop intelligent agents?” — Furong Huang, computer scientist, University of Maryland