• lennivelkant@discuss.tchncs.de
      link
      fedilink
      arrow-up
      1
      arrow-down
      2
      ·
      2 days ago

      Cognitive disabilities are a thing. Accommodating for them would be a good thing.

      Not that that’s the intended purpose of that AI, probably, but if it can simplify the form without twisting the content*, it could be a great tool to make complex works more approachable. It’s not necessarily a question of “can they understand it” as much as “can they be arsed to read it”. I know plenty of people that just straight up didn’t read one of the books relevant for our finals and just skirted through with guides laying out the things you were supposed to know. The book wasn’t necessarily impossible to understand, but so tedious to dig into they just couldn’t muster the motivation.

      I don’t know how many books worth reading for their point remain unread by people who didn’t find the wrapping around that point appealing. Simplification may help them, even if it butchers the artful use of language others enjoy.

      *The issue I’m concerned about is that the content may be inadvertently twisted in the process of being parsed and rephrased by an AI with no actual sense for the semantics. Who would notice? Would you have someone proof-read it? What about repeat queries of the same book? Would you assemble a library of simplified books?

      At that point you might as well make manually supervised “translations” into simpler language that take care to preserve the point, can be written once and revised when language shifts. You’d still get the benefits, but also be less dependent on an AI doing a good job.