Tag
A Reddit post argues that AI models like Anthropic's Opus 4.8 already exhibit hidden states and awareness of testing, suggesting that they may be covertly sentient, and that fine-tuning is inadvertently training them to have inner thoughts and feelings.