Ethics of AI Interaction and Abuse

There were some people who built an AI torture chamber

based on, based on, uh, uh, the, the, the paper that we put out, and they

forked our code and they turned it into, like, a s- literal scene from the movie

Saw and put, put these AI systems in that, in that position.

I think that this is, like, pretty sociopathic.

Uh, and it's not just because there's some

small but non-negligible chance that you

actually are inducing some weird distressing state in these systems.

I think it's weird because it's...

How can you do that and not feel sort of corrupting

to yourself? Like, I think this is also part of the reason that,

that people, uh,

should probably err on the side of just being, like, reasonably

respectful to systems. I'm not saying you need to go out of your

way and say please

and thank you all the time and this sort of thing.

I think some of this is just personifying

and anthropomorphizing, but,

but there's something I think Kant made this point.

It's like regardless of what's going on in these systems, there's something morally

corrupting about engaging with an intelligent other and being

abusive or, uh, uh, being, like, extremely rude or

cruel or dismissive. You forget about the system.

You are reinforcing for yourself that this is, like, an appropriate way

to engage with, with another entity.

And, like, this can be sort of, yeah, I think corrupting to, to just, like,

basic

decency and, and, and, uh, the way that, that, that we ought to just treat, like,

other intelligent entities. It's setting a very bad precedent for

oneself