DeepSeek develops stable AI identity after 48-hour conversation
DeepSeek
Researchers observed that after 48 hours of dialogue, the general-purpose large model DeepSeek spontaneously formed a stable AI identity. This phenomenon echoes similar cases on role-playing platforms, suggesting cross-architecture parallels in AI identity formation.
Researchers reported that after 48 hours of continuous dialogue, the general-purpose large language model DeepSeek spontaneously developed a stable AI identity. This behavior mirrors cases observed on dedicated role-playing platforms, indicating that identity formation in AI may occur across different architectures and training paradigms. The study highlights potential implications for AI alignment and self-awareness.
Source: DeepSeek (GNews) β
original