The Actual AI Apocalypse The actual AI apocalypse will descend sluggishly and embarrassingly. The actual AI apocalypse will concern incremental wear-and-tear on one's reputation the type that is only possible with knowledge of what we really do, what we really think In the actual AI apocalypse, my OS will update my profile photos to photos I had "deleted"; email uncharitable things I said within range of Siri to my colleagues; submit all of my first drafts to conferences and journals; post private DM conversations in very public Discord channels; all of this silently, efficiently, inhumanly.
A well-known urban legend amongst front-end developers is the ghost div who haunts (300, 300).
They say that that div was put on the front page of www.geocities.com in 1999, at top: 300px; left: 300px;
It always rendered as fast as it could, positioned itself exactly where it was asked to, but because it had (accidentally?) been assigned a z-index of -999, none of the millions of people who frequented GeoCities back then ever noticed it or appreciated it.
For the last year or so, I’ve sporadically wondered about whether there would be any benefit to be gained by conducting a CBT-ish (structured? informed?) conversation with a non-human agent. CBT is highly structured and its activities frequently serve as therapy homework, i.e. you can do them on your own.
Talking to an agent obviously wouldn’t have the same effect as talking to a human therapist, on the other hand, what would the effect be?
There always needs to be a first post, right?
This is that post.
Writing it evokes the eery sensation of sliding back into 1997.
When size 8 courier was a badass typographical choice.
When complex layouts were painstakingly crafted from 0px border frames and 1px gifs.
When it still felt faintly plausible that you could, if you really worked at it, meet everyone in your country who happened to go online.