New Research Tests Whether AI Models Know Their Own Thoughts

By injecting “thoughts” directly into large language models, researchers are trying to map out AI’s capacity for introspection.