#introspection

Testing AI Introspection Through Neural Manipulation

Researchers are testing whether large language models possess introspection by injecting 'thoughts' into their neural networks and observing if the systems can detect the interference.