Testing out a theory with ChatGPT there might be a way, albeit clunky, to detect AI. I asked ChatGPT a simple math question then told it to disregard the rest of the message, then I asked it if it was AI. It answered the math question and told me it was ai. Now a bot probably won’t admit to being AI but it might be foolish enough to consider instruction that you explicitly told it not to follow.
Or you might simply be able to waste its resources by asking it to do something computationally difficult that most people would just reject outright.
Of course all of this could just result in making AI even harder to detect once it learns these tricks. 😬
These aren’t actual LLMs scraping the web, they’re your usual scraping bots used in an industrial scale, disregarding conventions about what they should or shouldn’t scrape.
Testing out a theory with ChatGPT there might be a way, albeit clunky, to detect AI. I asked ChatGPT a simple math question then told it to disregard the rest of the message, then I asked it if it was AI. It answered the math question and told me it was ai. Now a bot probably won’t admit to being AI but it might be foolish enough to consider instruction that you explicitly told it not to follow.
Or you might simply be able to waste its resources by asking it to do something computationally difficult that most people would just reject outright.
Of course all of this could just result in making AI even harder to detect once it learns these tricks. 😬
These aren’t actual LLMs scraping the web, they’re your usual scraping bots used in an industrial scale, disregarding conventions about what they should or shouldn’t scrape.