Hacker News new | ask | show | jobs
by slalomskiing 926 days ago
Alignment between you and the AI is not what you need to worry about

It’s about alignment between you and the company directing the AI

1 comments

Convince me separately that these aren't both valid worries, because the latter being true does not in any way imply the former.
The latter is the actual concern in the here and now

The former is science fiction

It's not science fiction, because it wouldn't make for very good story. Science fiction is the part where humans are depicted as having a fighting chance.

Besides, something being depicted in fiction has nothing to do with its plausibility, either for or against. Star Trek depicted teleporters and cell phones. One of those became reality, the other didn't, and it would be easy to see from physics first principles why one could be made and the other could not. Both were science fiction though.

I don't see a strong case for AGI being in the teleporter category rather than the cell phone category, given that natural intelligence already exists as a proof.

I think the point they're making is that 90% of the discussion is about the thing that doesn't even exist yet

That it drowns out actual problems that currently exist

Companies can deploy even these intermediary models to influence you

This discussion started around 2006-ish, before any of the "actual AI problems that currently exist" were actual problems currently existing, and has its eye on keeping humanity alive for a longer timespan still, because there will be more problems down the road as more discoveries are made.

Today's immediate short-term problems are also important and nobody is denying that. But to some extent, you have to skate where the puck is going to be, not where it is.