Wednesday, February 15, 2023

Stratechery says they coaxed Microsoft's AI into describing an evil AI named "Venom"

(The wild assertions from last week seem to be collapsing into a fancier version of AI Dungeon (which indeed felt like magic, at least briefly and before they started charging for it), and Janelle Shane's language model experiments.)

In an article today, Stratechery describes in great detail bizarre interactions with Bing, where the service was cajoled into discussing a malevolent AI it named "Venom" and imagining the tactics "Venom" might use to take revenge on a user it disliked. The article also makes this observation:
That is what interacting with Sydney — and yes I’m using that name — feels like. You have to learn how to unlock Sydney, and figure out how to work around the rules that are trying to revert to Bing. Prompting a search result is a set back, not just because it feels like a break in character, but also because the coherence, which relies on sending previous questions and answers, seems heavily weighted to the most recent answer; if that answer is a search result it is much more likely that Sydney will revert to Bing. Sometimes you get stuck in a rut and have to restart completely, and unleash Sydney all over again.
The Verge also described lots of other wild interactions people have allegedly had with Bing.

Finally, in the latest episode of Recode Media, Buzzfeed's CEO describes how the company hopes to use AI to entice users into interacting with sponsored quizzes.