That is already done, as a way to improve AI, in a technique called Generative Adversarial Neural Networks (GANN).
One network’s task is to produce fake content. The other network’s task is to detect it. These networks then evolve together, each getting better and better.
What that means is that if we develop a means to detect AI-generated content, then that mechanism becomes the stepping stone to better AI.
Unfortunately, no one's yet succeeded in making a workable checker (AI-written text and the writing of English-as-foreign-language speakers are awkwardly similar along many axes, apparently, so existing checkers have a lot of false positives).
Build a checker for AI-generated content too please!
That is already done, as a way to improve AI, in a technique called Generative Adversarial Neural Networks (GANN).
One network’s task is to produce fake content. The other network’s task is to detect it. These networks then evolve together, each getting better and better.
What that means is that if we develop a means to detect AI-generated content, then that mechanism becomes the stepping stone to better AI.
I'm reminding of the humorous attempts to build a bear proof trash can. Ends up a bear proof trash can is also human proof for some humans.
Unfortunately, no one's yet succeeded in making a workable checker (AI-written text and the writing of English-as-foreign-language speakers are awkwardly similar along many axes, apparently, so existing checkers have a lot of false positives).