I thought persuasion was when you convince someone of something. It’s often an additional step after informing them of what you think is true.
In my experience, Persuasion is literally the easiest step.
I’m not convinced. Yud’s book came out months ago and it seems many important people have read it. I would think if the persuasion rate (of the top people) was not very low, things would look much different now.
Let me raise an inconvenient world for you: there are two types of actions in politics, fake and real. For fake actions such as signing a vague statement, attention/information might suffice. For real actions such as passing a law that will prevent lots of money from being made, persuasion of the underlying worldview is necessary.
Hindsight is 20/20! We do live in the world where we already have the statements, and still not the laws, so I don’t think it commits you to much!
But if you have a non-trivial understanding of “real” (ie, something beyond “it’s hard / hasn’t been done yet”), I’d love to hear more about it.
Ultimately, my goal for AI Policy org pipelines (ControlAI or not) is to build an incremental sequence of publicly verifiable political actions of increasing difficulty and importance, in a way that respects basic deontology.
I expect any non-trivial understanding of “real vs fake” would help me with designing such pipelines.
I thought persuasion was when you convince someone of something. It’s often an additional step after informing them of what you think is true.
I’m not convinced. Yud’s book came out months ago and it seems many important people have read it. I would think if the persuasion rate (of the top people) was not very low, things would look much different now.
Let me raise an inconvenient world for you: there are two types of actions in politics, fake and real. For fake actions such as signing a vague statement, attention/information might suffice. For real actions such as passing a law that will prevent lots of money from being made, persuasion of the underlying worldview is necessary.
I happen to think we do live in this world 🙂
Hindsight is 20/20! We do live in the world where we already have the statements, and still not the laws, so I don’t think it commits you to much!
But if you have a non-trivial understanding of “real” (ie, something beyond “it’s hard / hasn’t been done yet”), I’d love to hear more about it.
Ultimately, my goal for AI Policy org pipelines (ControlAI or not) is to build an incremental sequence of publicly verifiable political actions of increasing difficulty and importance, in a way that respects basic deontology.
I expect any non-trivial understanding of “real vs fake” would help me with designing such pipelines.