1. 1
  1.  

  2. 4

    The definitely tells you something about either GPT-3 or Daniel Dennett. It reminds me of when the BBC’s Tomorrow’s World did a live Turing Test and about as many people thought a chat bot was a human as thought Craig Charles was human. Reading a 10-line extract from the transcript made it obvious that the chat bot wasn’t human, so it was far more of a reflection on Craig Charles than the state of AI research.

    1. 1

      Did you do the quiz? I went into it expecting that GPT-3 would make obvious mistakes, but it really didn’t. The only “tell” I noticed was the real Dan refers to people by their first names, whereas the robot Dennett has to use last names, since it’s trained on book excerpts. Other than that, the machine does a very good job.