The name David Mayer probably means nothing to you, but he created a controversy linked to the behavior of ChatGPT when he mentioned it. A look back at a case that shows the limits of AI.
The David Mayer case
It all started when users spotted that the chatbot didn’t want to answer the question “Who is David Mayer?” ». He begins an answer, then deletes it in the middle by specifying: “I can’t come up with an answer. »
The mystery provoked numerous comments and the press took up the matter. So much so that OpenAI had to explain. The tech company stressed that the issue has now been resolved. She adds:
There may be cases where ChatGPT does not provide certain information about individuals to protect their privacy. One of our tools flagged this name in error and prevented it from appearing in replies, which it shouldn’t have done.
Data protection experts have started to dig into this matter. As stated in the Wall Street JournalGuido Scorza, an Italian lawyer, tested the AI on himself and discovered the same problem. One of the other banned names is Jonathan Turley, the American law professor.
An AI flaw
Beyond these particular cases, this story highlights the extent to which issues related to privacy are major for these AIs. Moreover, several people have filed lawsuits on this subject against OpenAI.
To remedy this, the designers of these tools use techniques that are not innovative but, on the contrary, date back to the beginnings of computing. They actually impose rules on these chatbots which look like this example: If “such name” is mentioned, then do not provide a response and generate: “I am not able to produce an answer”. The same process is also used for certain taboo subjects that AI is not supposed to address.
Have you ever noticed these cases where ChatGPT and its rivals do not respond to your queries? Tell us in the comments.
What to remember:
- Internet users noticed that ChatGPT did not want to answer the question: “who is David Mayer?” »
- The bug has since been fixed and OpenAI says it was an error to protect people’s privacy
- This case illustrates the major problems faced by these AIs
