Misalignment likely turns catastrophic before most other destructive technologies

Paul ChristianoPaul Christiano — Preventing an AI takeoverat 1:44:00

From the conversation

But just for the audience there’s some way to configure $50,000 and a teenager’s time to destroy a civilization. If that thing is available, then misuses itself a teal risk, right? So do you think that that prospect is less likely than a way you could put it? Is there’s, like, a bunch of potential destructive technologies? And alignment is about AI itself being such a destructive technology, where even if the world just uses the technology of today, simply access to AI could cause human civilization to have serious problems. But there’s also just a bunch of other potential destructive technologies. Again, we mentioned like physical explosives or bioweapons of various kinds, and then the whole tale of who knows what. My guess is that Alignment becomes a catastrophic issue prior to most of these. That is, like, prior to some way to spend $50,000 to kill everyone, with the salient exception of possibly, like, bioweapons. So that would be my guess. And then there’s a question of what is your risk management approach? Not knowing what’s going on here, and you don’t understand whether there’s some way to use $50,000. But I think you can do things like understand how good is an AI at coming up with such schemes. Like, you can talk to your AI. Be like, does it produce new ideas for destruction we haven’t recognized? Yeah. Not whether we can evaluate it, but whether if such a thing exists.…

Machine-generated transcript. The excerpt can include the interviewer and other voices, and the transcription may contain errors. The rest is in the episode →

Summary

Asked whether a cheap route to mass destruction — something like $50,000 and a teenager's time — would make misuse an existential risk of its own, Christiano guesses that alignment becomes a catastrophic issue before most such destructive technologies, with the possible exception of bioweapons. Not knowing which of these risks exist, one practical step is to test how good AI systems are at coming up with new schemes for destruction.

Watch the clip on YouTubeStarts at 1:44:00

Concept

More from Paul Christiano

Related clips