alt Hacker NewsUhhh you know that the paperclip problem stems from ai just following a task, not understanding what it is doing? Not from being misalignment.
I would go out a limb and say that current a could create a paperclip problem, given powerful enough tools.
The argument is that it's misaligned because it only values one thing: more paperclips, while human values are much more varied and complex.
Debatable whether it truly understands what it's doing or not, but the argument usually assumes that it does know what it's doing at least in that it's able to imagine outcomes and create plans to reach its singular goal, making it a very simple toy example of a misaligned system.