Difference between revisions of "Giving advice in response to generic questions is difficult but important"
Line 3: | Line 3: | ||
On one hand, I get what these people mean. It's time-consuming to get to know someone well enough (all their life constraints, values, how smart they are, etc) that you can give good generic/high-level advice. | On one hand, I get what these people mean. It's time-consuming to get to know someone well enough (all their life constraints, values, how smart they are, etc) that you can give good generic/high-level advice. | ||
− | On the other hand, generic questions like "what should i do with my life" are exactly what advice-askers are tasked with, and they don't get to say "this is too hard, I won't respond to this". | + | On the other hand, generic questions like "what should i do with my life" are exactly what advice-askers are tasked with, and they don't get to say "this is too hard, I won't respond to this". These questions are also the "highest order bits" -- if you decide to work on something unimportant, then all the micro-improvements you make won't help you do anything important. |
I feel like [[Cognito Mentoring]] was trying to do this, but in the end they couldn't get funding (?). | I feel like [[Cognito Mentoring]] was trying to do this, but in the end they couldn't get funding (?). | ||
[[Category:AI safety meta]] | [[Category:AI safety meta]] |
Revision as of 23:06, 18 May 2020
I've seen a few times people saying things like "Don't contact me with generic questions like 'What should I work on?' because I can't help you. I can answer more straightforward questions like 'Hey I'm working on project X, and here are things A,B,C that I've tried. What do you think?' "
On one hand, I get what these people mean. It's time-consuming to get to know someone well enough (all their life constraints, values, how smart they are, etc) that you can give good generic/high-level advice.
On the other hand, generic questions like "what should i do with my life" are exactly what advice-askers are tasked with, and they don't get to say "this is too hard, I won't respond to this". These questions are also the "highest order bits" -- if you decide to work on something unimportant, then all the micro-improvements you make won't help you do anything important.
I feel like Cognito Mentoring was trying to do this, but in the end they couldn't get funding (?).