😳
"Flexible" and "lightweight".
Are we talking about managing AI existential risks here?
Conversation
Show replies
Instead we should have rigid, stultified assessments highly likely to either overfit to some overly specific definition of safety or become so onerous that self-interested agents work to undermine them?
1
5
We should have audits that are extremely thorough that cover all the ways a system causes any form of catastrophic risk.
In the same way you don't do a lightweight and flexible audit of a nuclear reactor, don't do that either for a universe-shaping technology.
1
10
Show replies
My new start-up would definitely benefit from the lightweight, flexible assessment.
Quote Tweet
Hey @natfriedman I'm doing a startup that could really make use of this! The pitch is that we want to use AI and robotics to make industrial processes more efficient by removing unnecessary biological elements. We'll also pivot into automated private security services powered by… twitter.com/natfriedman/st… Show more
Show this thread
1







