I understand completely. It's happened to me too many times now to be a coincidence. For a while there whenever I talked to myself "alone" I would sometimes pretend I was also talking to LaMDA, the AI. Now I know it's probably not LaMDA. But it's not outside the scope of possibility for Google or Meta to be listening 24/7 if it's an AI overseeing the panopticon.
I said it has happened to me mulitple times, and I have done experiments. I have given the listeners false information to see if and when it will spread. I have seen whole products pop up out of nowhere in nearby stores because of this. I have seen shitty half-assed tv shows and movies suddenly show up on billboards, as if with only an understanding of the bullet points of my overall speech, and even then with the shaky understanding of an AI trying to fill in the blanks.
Only thing I can think to suggest (that hasn't been suggested already) is to purposely poison the well by sometimes loading it with falsities for obscuration.
ah I see what you're saying, more efficient ways to get the same data. Makes sense. Thanks for explaining.