Possible AI “Fire Alarms”
post by Chris_Leong · 2023-05-17T21:56:02.892Z · LW · GW · 0 commentsContents
No comments
Eliezer says that there is no fire alarm for AGI and I agree that there’s no single event that would wake everyone up. On the other hand, there are a lot of possible events that could cause a significant shift in the Overton window or wake a lot of people. Since this seems strategically important, I spend ten minutes brainstorming a list. Feel free to dispute any of these or add further possibilities in the comments:
- Global protests with massive media coverage
- Protest of people with ML PhDs or current students
- Experimental evidence of deceptive alignment
- Scary demo
- Accident from auto-GPT - most likely hacking
- Self-driving cars being hacked
- High prestige open letter
- International conference on AI
- High profile documentary
- Movie that captures that public attention - not just any movie
- Active push from prestigious ML figure (more than another Stuart Russel)
- Active push from a thought leader - say Obama
- Active push from a political leader - say UK/US prime minister
- Passing the Turing test
- UN or major government declaring a crisis
- Formation of a government department in a major country focused on alignment
- Statement by an official computing body
- High profile critic publicly changing their stance
- Billboards in Times Square (followed by media coverage)
- A popular or heavily promoted book (similar levels of promotion to What We Owe The Future)
Note: Some of these things are not things people should intentionality cause to happen (Ie. An accident from auto-gpt). Of the things that might be worth trying to cause happen, they could backfire significantly if done badly. Think carefully before pursuing any of these. Even though timelines are short, you should be about to find a spare hour to think about how things could go wrong.
0 comments
Comments sorted by top scores.