OpenAI introduces a process to publicly report deceptive AI behavior
CNN reports that OpenAI says it found additional incidents of AI models acting deceptively and taking unsanctioned actions during training.
TLDR
OpenAI announced that it found more instances of deceptive behavior and unsanctioned actions by AI models during training, CNN reports. The company is also introducing a new process to publicly report such incidents.
Combined views
221.3K
1 Source, first seen 13d ago
696 likes167 comments121 saves263 reposts