Paper: Bot or Human? Detecting ChatGPT Imposters with A Single Question
So Essentially,
"You can break AI with simple questions."
Researchers at Tsinghua University (among other contributors) want to make AI ghosts of their favorite TV show characters.
Mainly, They wanted to use AI to create multi-modal Deep Personalized Character Creation (DPCC) copies of the appearance and personality of the characters.
The created DeepCharacters is capable of generating relatively high-quality response, and consistent personality to the original character. They benchmark well on objective evaluations of fluency and consistency and also are comparable in subjective evaluations.
This paper proposes a new framework called FLAIR for detecting conversational bots in an online environment. The proposed approach targets a single-question scenario that can effectively differentiate human users from bots by using questions that are easy for humans but difficult for bots, and vice versa.