This post is my attempt to summarize overarching concerns about how the Optimality Function might be misaligned, and once such problem (implied in the 3rd contingency,) is that an agent arises that is misaligned—for a variety of reasons/explanations. How might an agent be misaligned? Because of embedded agency. I don’t say it, but it’s implied.
I don’t cite or even mention the exact precise ways the Optimality Function could go wrong—I added the mesa-optimization tag because it ties into contingency 2 but I DEFINITLEY don’t talk about that explicitly either.
Overall, this might just be a far too liberal usage of tags on my part.
I’m sorry, I’m new and just trying to summarize, mostly for myself. “Learn by teaching/doing” and all that.
Going by https://www.lesswrong.com/tag/embedded-agency#:~:text=Embedded%20Agency%20is%20the%20problem,and%20not%20separated%20from%20it.
This post is my attempt to summarize overarching concerns about how the Optimality Function might be misaligned, and once such problem (implied in the 3rd contingency,) is that an agent arises that is misaligned—for a variety of reasons/explanations. How might an agent be misaligned? Because of embedded agency. I don’t say it, but it’s implied.
I don’t cite or even mention the exact precise ways the Optimality Function could go wrong—I added the mesa-optimization tag because it ties into contingency 2 but I DEFINITLEY don’t talk about that explicitly either.
Overall, this might just be a far too liberal usage of tags on my part.
I’m sorry, I’m new and just trying to summarize, mostly for myself. “Learn by teaching/doing” and all that.
I did scour the https://www.lesswrong.com/tag/guide-to-the-lesswrong-editor to see how I should use tags, for what it’s worth, but it doesn’t mention “tag etiquette”. Perhaps it should, so cases like these are avoided.