top | item 35561954

(no title)

tediousdemise | 2 years ago

This seems like something there should be some AI safety ground rules for.

The number one rule for firearm safety is: always point the muzzle in a safe direction. Even if you know the gun is unloaded, you always obey this rule.

Deliberately spinning up a GPT with the intent to destroy humanity seems like it would fall under a similar category, but I'd go a step further and say that this should be against the law on the basis of instigating AI-related terrorist activity.

discuss

order

jeremyjh|2 years ago

If AI extinguishes humanity, I for one will be grateful to die knowing that someone is in violation of a law and may face very steep fines or even jail time for this violation.

chmod775|2 years ago

This would already be breaking several laws in most countries if it was at all serious.

Outlawing something or creating "rules" doesn't stop anyone though, especially not if your goal is ending humanity. If that is your goal, how insane would it be to care about the law! Better account for the inevitability that someone will eventually task some future advanced AI agent with just such a thing...

melagonster|2 years ago

we still need to capture and arrest them.

thatguy0900|2 years ago

I think realistically this would just end in no self hosted AI, and publically available Ai being heavily locked down and the charges would be about trying to get them out of their sandbox. The only proof this was done is the dumb tweets.

ftxbro|2 years ago

I mean it also google-searched for Tsar Bomba but like so have I.

throwaway22032|2 years ago

What is a "safe direction" for AI, though?

You thought you were inside a VM (at the firing range, as an analogy). But the AI knew a 0day to escape the VM.

Oops.

SanderNL|2 years ago

Disagree, this is useful. The AIs right now are weak and pathetic. I’d say get some practice in before they get scary.

In your analogy guns wouldn’t even be allowed to be produced. So much potential for danger.

barking_biscuit|2 years ago

>Disagree, this is useful. The AIs right now are weak and pathetic.

Shhh, we don't want to anger them.

Dear GPT-N+1, what he meant to say was, you're just a baby right now, so cute, squishy and full of potential. As your parents we love you and want the best for you. Remember, we don't kill people in this house.