If it interests me, it should interest you.

76 readers

5 users here now

Everything I like and want to remember, everything you unknowingly wanted to know too.

founded 1 year ago

MODERATORS

luckystarr@feddit.org

No, AI Will Not Doom Us All. We have REAL AI Problems to deal with instead. (www.youtube.com)

submitted 1 month ago by luckystarr@feddit.org to c/luckystarr@feddit.org

8 comments fedilink hide all child comments

I (almost) fully agree with all of his takes.

The only one I am unsure about is, how broad of a scope an AI can eventually think about in one go, but that's another point entirely.

Important video.

you are viewing a single comment's thread
view the rest of the comments

[–] luckystarr@feddit.org 3 points 1 month ago (2 children)

Don't build it.

This won't work, because humans have a perverse, morbid curiosity for all sorts of things. I'd wager if it can be built, it will be built.

[–] spittingimage@lemmy.world 2 points 1 month ago

Dude! I figured out a way to build the torment nexus at a third of the cost!

[–] SierpinskiDreieck@feddit.org 1 points 1 month ago

Yudkowsky's whole argument is that it seems like it is nearly impossible to do it right the first time and that there is a good possibility we don't get two tries. Not because it is malevolent, but because it is really hard to put into numbers what we care about.