Nightmare Machine by MIT Media Lab | Hacker News Reader

Nightmare Machine by MIT Media Lab

nightmare.mit.edu

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.

On behalf of Saito Game Team, I most humbly apologize.

https://news.ycombinator.com/item?id=12765022

Is it a Windows machine?

I wonder if other animals find gore and distorted faces horrifying.

AI, "The Boogeyman" man of our time.

Yeah right. Just teach skynet how to scare humans...

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.

Should we be investing any time at all in developing AI that is meant to make people happy or have a positive impact on the way people feel?

Have you seen the movie "enter the void"?

Would your opinion on this change after watching that movie?

If we don't invent a nightmare machine then the Soviets surely will, and then where will we be?! A nightmare gap, that's where!

My opinion is we need to invest more funding in nightmare-generation research before it's too late.

People enjoy horror stories, films, games, etc. Is that having a negative or positive impact?

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.

Err - my comment was more of a joke. What a difference a ;) can make...

Who asked for this? This really seems antithetical to AI/human goal alignment.

Can you please elaborate further on "AI/human goal alignment"?

Humans make art like this, would you say those people are working against "human goals"?

Reply on news.ycombinator.com