I like the tone of this review. That might be because it scans as positive about something I wrote! :D But I think it’s at least in part because it feels clear, even where it’s gesturing at points of improvement or further work. I imagine I’d enjoy more reviews written in this style.
I would be interested to see research done to test the claim. Does increased sympathetic nervous system activation cause decreased efficacy [at AI research]?
If folk can find ways of isolating testable claims from this post and testing them, I’m totally for that project.
The claim you name isn’t quite the right one though. I’m not saying that people being stressed will make them bad at AI research inherently. I’m saying that people being in delusion will make what they do at best irrelevant for solving the actual problem, on net. And that for structural reasons, one of the signs of delusion is having significant recurring sympathetic nervous system (SNS) activation in response to something that has nothing to do with immediate physical action.
The SNS part is easy to measure. Galvanic skin response, heart rate, blood pressure, pupil dilation… basically hooking them up to a lie detector. But you can just buy a GSR meter and mess with it.
I’m not at all sure how to address the questions of (a) identifying when something is unrelated to immediate physical action, especially given the daughter’s arm phenomenon; or (b) whether someone’s actions on net have a positive effect on solving the AI problem.
E.g., it now looks plausible that Eliezer’s net effect was to accelerate AI timelines while scaring people. I’m not saying that is his net effect! But I’m noting that AFAIK we don’t know it isn’t.
I think it would be extremely valuable to have some way of measuring the overall direction of some AI effort, even in retrospect. Independent of this post!
But I’ve got nuthin’. Which is what I think everyone else has too.
I’d love for someone to prove me wrong here.
A sequence or book compiled from the wisdom of many LessWrongers discussing their mental health struggles and discoveries would be extremely valuable to the community (and to me, personally)…
I’m glad you enjoyed my review! Real credit for the style goes to whoever wrote the blurb that pops up when reviewing posts; I structured my review off of that.
When it comes to “some way of measuring the overall direction of some [AI] effort,” conditional prediction markets could help. “Given I do X/Y, will Z happen?” Perhaps some people need to run a “Given I take a vacation, will AI kill everyone?” market in order to let themselves take a break.
What would be the next step to creating a LessWrong Mental Health book?
I like the tone of this review. That might be because it scans as positive about something I wrote! :D But I think it’s at least in part because it feels clear, even where it’s gesturing at points of improvement or further work. I imagine I’d enjoy more reviews written in this style.
If folk can find ways of isolating testable claims from this post and testing them, I’m totally for that project.
The claim you name isn’t quite the right one though. I’m not saying that people being stressed will make them bad at AI research inherently. I’m saying that people being in delusion will make what they do at best irrelevant for solving the actual problem, on net. And that for structural reasons, one of the signs of delusion is having significant recurring sympathetic nervous system (SNS) activation in response to something that has nothing to do with immediate physical action.
The SNS part is easy to measure. Galvanic skin response, heart rate, blood pressure, pupil dilation… basically hooking them up to a lie detector. But you can just buy a GSR meter and mess with it.
I’m not at all sure how to address the questions of (a) identifying when something is unrelated to immediate physical action, especially given the daughter’s arm phenomenon; or (b) whether someone’s actions on net have a positive effect on solving the AI problem.
E.g., it now looks plausible that Eliezer’s net effect was to accelerate AI timelines while scaring people. I’m not saying that is his net effect! But I’m noting that AFAIK we don’t know it isn’t.
I think it would be extremely valuable to have some way of measuring the overall direction of some AI effort, even in retrospect. Independent of this post!
But I’ve got nuthin’. Which is what I think everyone else has too.
I’d love for someone to prove me wrong here.
This is a beautiful idea. At least to me.
I’m glad you enjoyed my review! Real credit for the style goes to whoever wrote the blurb that pops up when reviewing posts; I structured my review off of that.
When it comes to “some way of measuring the overall direction of some [AI] effort,” conditional prediction markets could help. “Given I do X/Y, will Z happen?” Perhaps some people need to run a “Given I take a vacation, will AI kill everyone?” market in order to let themselves take a break.
What would be the next step to creating a LessWrong Mental Health book?