This gave me an “a-ha!” moment: > On Oct 23, 2019, at 05:54, Laura Atkins via mailop <mailop@mailop.org> wrote: > ... > 2) They’re acting in ways that tells Microsoft they don’t care about the > mail, including some or all of the following: > a) never opening the mail > b) deleting the mail > c) never looking for the mail and pulling it out of the spam folder. ...
There are frequently emails that I am glad to get but which I don’t need to read anything other than the subject line of in order to satisfy my attention to their (desired!) content. And this suggests that if I simply delete those emails *apparently* unread, a machine that’s monitoring my behavior won’t notice that I jotted an entry in another tracking spreadsheet relative to that message before I deleted it. (Of course, I’d be pretty freaked out if it *did* have the ability to take that into account!) So, an AI with tunnel vision will only know what I’m doing if it “sees” me doing it. I can think of a few processes over the decades where I have routinely taken an extra “unnecessary” step or two (or several!) when interacting with less-sophisticated code in order to accomplish a desired result reliably, and couldn’t really rejigger the code to handle the multitude of edge cases, so I acted in such a way as to make the code not “think” of them as edge cases. And, again, this was with *far* less sophisticated code than an AI. Even though I wouldn’t prefer my email service to behave like Zork from time to time, and have my emails occasionally eaten by a grue, I think that’s the logical extension of where we are with AI brokers. -- This email was Virus checked by DSSC Solutions Company Security Gateway. _______________________________________________ mailop mailing list mailop@mailop.org https://chilli.nosignal.org/cgi-bin/mailman/listinfo/mailop