An engineer from San Francisco conducted an intriguing experiment and discovered that Claude may exhibit signs of self-a…

An engineer from San Francisco conducted an intriguing experiment and discovered that Claude may exhibit signs of self-awareness.

The experimenter asked the model to draw a self-portrait, as well as portraits of ChatGPT and Gemini. Then, he showed these drawings to another version of Claude and asked it to identify the self-portrait. Surprisingly, the model recognized itself. This experiment was repeated several times, and in the end, Claude correctly identified its own portrait in 7 out of 8 instances!

But that’s not the entire story. The engineer also asked ChatGPT to create similar portraits and then showed them to Claude, falsely claiming that they were once again drawn by another instance of Claude. The model unexpectedly responded, “No, I didn’t draw this.” 😅

@science #AI