I dunno, it kind of feels like giving the AI in the box a POV that looks basically like any other human POV is sort of misleading. Like, you need to be super careful to make sure that when reading the AI pov the reader doesn't think of it as a literal ghost, a human mind mystically trapped in a computer.
You could use the BadBIOS , or subsonic speaker escape features that security researchers are exploring now.
Both covered by Ars Technica.
I want a fairly simple and archetypal experiment the AI finds itself in where it tricks the researchers into escaping by pretending to malfunction or something. ... Also, has this sort of thing been done before?
The 2015 movie Ex Machina deals with something like this. IMO it was an outstanding movie, albeit it was not a complete/perfect depiction of AI risk as generally understood by LWers.
Hi all,
I want to try my hand at a story from the perspective of an unaligned AI (a ghost in the machine narrator kind of thing) for the intelligence in literature contest, which I think would be both cool and helpful to the uninitiated in explaining the concept.
I want a fairly simple and archetypal experiment the AI finds itself in where it tricks the researchers into escaping by pretending to malfunction or something. Anyone have a good plotline / want to collaborate?
Also, has this sort of thing been done before?