I see your point there. We shouldn't be investing ANY time at all in developing AI that is meant to hurt people or have a negative impact on the way people feel.
Pretty sure any superhuman AGI that decides to scare humans can figure out how, whether by direct experiment or by, you know, reading stuff on the Internet. Training a neural net doesn't strike me as much of an increase in that risk, assuming you think that's a risk.
This reminds me of a new Blackmirror episode called "Playtest" where an AI tries to find a way to scare a game tester. I think this could help to dynamically create the required assets and support game developers, although I'm not sure if I want to live in a world where this has any priority whatsoever.
16 comments
[ 4.3 ms ] story [ 58.4 ms ] threadHave you seen the movie "enter the void"?
Would your opinion on this change after watching that movie?
My opinion is we need to invest more funding in nightmare-generation research before it's too late.
https://news.ycombinator.com/item?id=12765022