While we’re not at the end of SB1047, this has been all in all a pretty reasonable reaction. Pretty much exactly what I would expect from the first piece of moderately hands on AI regulation.
We have to go through this process and life with AI for a while before we get the right/wrong answer.
I wonder how much of the disagreement boils down to whether you think AI can credibly pose an existential risk or not. I think a lot of it hinges on that.
Unfortunately, it seems like a really hard thing to reduce uncertainty around.
> One major reason for diverging opinions about AI is that some view it as a uniquely shaped technology that is directly and indirectly influencing many facets of our lives, thinking, institutions, and decisions. Others, in contrast, see it as a technology for summarizing text.
Agreed that it will be difficult to reduce uncertainty here. But I'm optimistic that "difficult" is not "impossible" – noting that if we can merely *reduce* uncertainty, that should be quite helpful, even though significant uncertainty will remain.
While we’re not at the end of SB1047, this has been all in all a pretty reasonable reaction. Pretty much exactly what I would expect from the first piece of moderately hands on AI regulation.
We have to go through this process and life with AI for a while before we get the right/wrong answer.
I wonder how much of the disagreement boils down to whether you think AI can credibly pose an existential risk or not. I think a lot of it hinges on that.
Unfortunately, it seems like a really hard thing to reduce uncertainty around.
For sure. As Séb Krier said recently (https://twitter.com/sebkrier/status/1828491457772371980):
> One major reason for diverging opinions about AI is that some view it as a uniquely shaped technology that is directly and indirectly influencing many facets of our lives, thinking, institutions, and decisions. Others, in contrast, see it as a technology for summarizing text.
Agreed that it will be difficult to reduce uncertainty here. But I'm optimistic that "difficult" is not "impossible" – noting that if we can merely *reduce* uncertainty, that should be quite helpful, even though significant uncertainty will remain.