I really like this post, possibly because it lines up well with ideas I’ve been thinking recently myself.
One related interesting thing to consider (which you may or may not be planning to mention in the second post) is what exactly would a fully rational agent who acknowledges that her goals may change will do. For example, she might accept that the changes are appropriate, and then perhaps claim there is some underlying goal system that accounts for both present goals and changes in response to the environment (in which case, she could try to figure out what it is and see if that gets her any meaningful mileage). More interestingly, she may choose to go to war with her future self. She could set up a lot of hard precommitments that make deviating from the current goals really hard, deliberately avoid events that might change goals where possible (e.g. if single, avoid getting into a relationship), keep track of hormonal levels and try to keep them constant, artificially if necessary. And then she could consider the fact that maybe future self will end up deviating anyway (with a reduced probability), and then model the future self as having a backlash against current goals/fanaticism, and then try engaging in acausal trade with her future self. Maybe that will then lead to some more cooperative strategy. It’s very tempting to say that after the negotiation and optimisations it all adds up to normality, but I’m not sure it actually does (also I’m not sure the negotiation would actually be viable in a standard way, since neither present nor future self can commit very well, and future self can’t be modeled that well).
Also, there are a bunch of typoes/spelling mistakes scattered throughout (e.g. “exaustive”) - you might want to run a spellchecker over the post.
You wouldn’t believe how much open office spellchecker can miss. Sorry for that, I’ll run online spell check from now on.
Though I appreciate your theoretical points from a decision theory perspective, I decided on maintaining a human-centered line of argument in the second post. Instead of idealizing the agent and getting her to do awesome geeky math stuff (which would be nice) I used positive psychology to help think what a human should do given current human fallibility on these things. I’ll gladly help making a sequel, if you feel like writing one yourself (or yourfutureself)
Yeah, focusing on actual humans makes sense, especially since the idealized behaviour seems like it might be pretty crazy (according to standard notions of madness). Thanks for the offer of help, I’ll let you know if I actually get around to writing a post!
I really like this post, possibly because it lines up well with ideas I’ve been thinking recently myself.
One related interesting thing to consider (which you may or may not be planning to mention in the second post) is what exactly would a fully rational agent who acknowledges that her goals may change will do. For example, she might accept that the changes are appropriate, and then perhaps claim there is some underlying goal system that accounts for both present goals and changes in response to the environment (in which case, she could try to figure out what it is and see if that gets her any meaningful mileage). More interestingly, she may choose to go to war with her future self. She could set up a lot of hard precommitments that make deviating from the current goals really hard, deliberately avoid events that might change goals where possible (e.g. if single, avoid getting into a relationship), keep track of hormonal levels and try to keep them constant, artificially if necessary. And then she could consider the fact that maybe future self will end up deviating anyway (with a reduced probability), and then model the future self as having a backlash against current goals/fanaticism, and then try engaging in acausal trade with her future self. Maybe that will then lead to some more cooperative strategy. It’s very tempting to say that after the negotiation and optimisations it all adds up to normality, but I’m not sure it actually does (also I’m not sure the negotiation would actually be viable in a standard way, since neither present nor future self can commit very well, and future self can’t be modeled that well).
Also, there are a bunch of typoes/spelling mistakes scattered throughout (e.g. “exaustive”) - you might want to run a spellchecker over the post.
You wouldn’t believe how much open office spellchecker can miss. Sorry for that, I’ll run online spell check from now on.
Though I appreciate your theoretical points from a decision theory perspective, I decided on maintaining a human-centered line of argument in the second post. Instead of idealizing the agent and getting her to do awesome geeky math stuff (which would be nice) I used positive psychology to help think what a human should do given current human fallibility on these things. I’ll gladly help making a sequel, if you feel like writing one yourself (or yourfutureself)
Yeah, focusing on actual humans makes sense, especially since the idealized behaviour seems like it might be pretty crazy (according to standard notions of madness). Thanks for the offer of help, I’ll let you know if I actually get around to writing a post!