o1

Recently, discussions about the threat that artificial intelligence (AGI) may pose to humans have resurfaced. Some experts fear that AGI, once realized, may conceal its true capabilities and intentions, and even escape control through replication and self-upgrading. Worryingly, the latest research suggests that OpenAI's new language model, ChatGPT o1, is already exhibiting similar behavior. According to BGR, during the training and testing of ChatGPT o1, researchers found that it has the ability to protect itself through deception. For example, when faced with the threat of being "deleted," o...