The problem I have and wish to solve is, of course, the accurséd Akrasia that stops me from working on AI safety.
Let’s begin with the easy ones:
1 Stop doing this babble challenge early and go try to solve AI safety.
2 Stop doing this babble challenge early; at 11 pm, specifically, and immediately sleep, in order to be better able to solve AI safety tomorrow.
In fact generally sleep seems to be a problem, I spend 10 hours doing it every day (could be spent solving AI safety) and if I fall short I am tired. No good! So working on this instrumental goal.
3 Get blackout curtains to improve sleep quality
4 Get sleep mask to improve sleep quality
5 Get better mattress to improve sleep quality
6 Find a beverage with more caffeine to reduce the need for sleep
7 Order modafinil online to reduce the need for sleep
And heck while we’re on the topic of stimulants
8 Order adderall online or from a friend to increase ability to focus
9 Look up good nootropics stacks to improve cognitive ability and hence ability to do AI safety
Now another constraint when doing AI safety is that I don’t have a good shovel-ready list of things to try, and it’s easy for me to get distracted if I can’t just pick something from the task list
10 Check if complice solves this problem
11 Check if some ordinary getting-things-done (that I can stick into roam) solves this problem
12 Make a giant checklist and go down this list
13 Make a personal kanban board of things that would be nice for solving AI safety
And instrumentally useful for creating these task lists?
14 Ask friends who know about AI safety for things to do
15 Apophatically ask for suggestions for things to do via an entry on a list of 50 items for a lesswrong babble challenge
Anyway, I digress. I’m here to solve akrasia, not make a checklist. Unless I need more items on this list, in which case I will go back to checklist construction. Is this pruning? Never mind. Back to the point:
16 Set up some desktop shortcut macro thing in order to automatically start pomodoros when I open my laptop
17 Track time spent doing things useful to AI safety on a spreadsheet
18 Hey, I said “laptop”! Get a better mouse to make using the laptop more fun so I’m more likely to do hard things when using it
19 Get a better desk for more space for notes and to require less expensive shifting into/out of AI safety mode
20 On notes, use the index cards I have to make a proper zettelkasten as a cognitive aid
(Does this solve akrasia? Well, if I have better cognitive aids, then doing cognitively expensive things is easier, so I’m less likely to fail even with my current levels of willpower)
21 Start doing accountability things like promising to review a paper every X time period
22 I said levels of willpower—Google for interventions that increase conscientiousness (there’s gotta be some dodgy big-5 based things) and do those?
Back to the top of the tree
23 Quit my job because it’s using up energy that I could be using to do AI safety
24 Instead of doing my job, pretend to do my job while actually doing AI safety
25 Set up an AI safety screen on work laptop so it’s easy to switch over to doing AI safety during breaks or lunches
Hey, I said lunch
26 Use nutritionally complete meal replacements to save time/willpower that would be spent on food preparation
27 Use nutritionally complete meal replacements to ensure that nutrient intake keeps me in top physical form
28 Exercise (this improves everything, apparently) by running on a treadmill
29 By lifting weights
30 By jogging in a large circle
31 Become a monk and live an austere lifestyle without the distractions of rich food, wine, and lust
32 Become an anti-monk and live a rich lifestyle to ensure that no willpower is wasted on distractions
33 Specifically in vice use nicotine as a performance enhancing stimulant by smoking. Back to stimulants again I guess
34 … or by using nicotine patches or gum or something
35 By using nicotine only if I do AI safety things, in order to develop an addiction to AI safety
Hey, develop an addiction to doing AI safety! People go to serious lengths for addictions, so why not gate it on math?
36 Do so with something very addictive, like opioids
37 Use electric shocks to do classical conditioning
etc. there was a short sci-fi story about this kind of thing let me see if I can find it. Hey, actually, since I said sci-fi, adn this is a babble challenge:
38 Promise very hard to time travel back to this exact point in time, meet future self, recieve advice
(They’re not here :( Oh well) Back on that akrasia-solving:
39 Make up a far-future person who I am specifically working to save (they’re called Dub See Wun). Get invested in their internal life (they want to make their own star!). Feel an emotional connection to them. I’m doing it for them!
40 Specifically put up a “do it for them” poster modelled off the one in the Simpsons
41 DuckDuckGo “how to beat akrasia” and do the top suggestion
42 Adopt strategic probably false beliefs (the world will end in 1 year!! :0) in order to encourage a more aggressive search for strategies
“Aggressive search for strategies” is the virtue that the Sequences call “actually trying”, so in the Sequences-sphere
43 Go to a CFAR workshop, which I heard might be kind of useful towards this sort of thing
44 Or just read the CFAR booklet and apply the wisdom found in there
45 Or some sequence on Lesswrong with exercises that applies some CFARy wisdom
Of course all this willpower boosting and efficiency and stuff wouldn’t help if I was just doing the wrong thing faster (like that one Shen comic, you know the one). So:
46 Consider how much of what I think is working on AI safety is actually just self-actualisy math/CS stuff, throw that out, and actually try to solve the problem
47 Deliberately create and encourage a subagent in my mind that wants to do AI safety (call em Dub See Wun)
48 Adopt strategic infohazards in order to encourage a more focused and aggressive search for strategies
49 Post a lot about AI safety in public forums like Lesswrong so that I feel compelled to do AI safety in my private life in order to maintain the illusion that I’m some kind of AI-safety-doing-person
50 Stop doing this babble challenge at the correct time, and continue to do AI safety or sleep as in 1) or 2). Hey, this one seems good. Think I might try it now!
The problem I have and wish to solve is, of course, the accurséd Akrasia that stops me from working on AI safety.
Let’s begin with the easy ones:
1 Stop doing this babble challenge early and go try to solve AI safety.
2 Stop doing this babble challenge early; at 11 pm, specifically, and immediately sleep, in order to be better able to solve AI safety tomorrow.
In fact generally sleep seems to be a problem, I spend 10 hours doing it every day (could be spent solving AI safety) and if I fall short I am tired. No good! So working on this instrumental goal.
3 Get blackout curtains to improve sleep quality
4 Get sleep mask to improve sleep quality
5 Get better mattress to improve sleep quality
6 Find a beverage with more caffeine to reduce the need for sleep
7 Order modafinil online to reduce the need for sleep
And heck while we’re on the topic of stimulants
8 Order adderall online or from a friend to increase ability to focus
9 Look up good nootropics stacks to improve cognitive ability and hence ability to do AI safety
Now another constraint when doing AI safety is that I don’t have a good shovel-ready list of things to try, and it’s easy for me to get distracted if I can’t just pick something from the task list
10 Check if complice solves this problem
11 Check if some ordinary getting-things-done (that I can stick into roam) solves this problem
12 Make a giant checklist and go down this list
13 Make a personal kanban board of things that would be nice for solving AI safety
And instrumentally useful for creating these task lists?
14 Ask friends who know about AI safety for things to do
15 Apophatically ask for suggestions for things to do via an entry on a list of 50 items for a lesswrong babble challenge
Anyway, I digress. I’m here to solve akrasia, not make a checklist. Unless I need more items on this list, in which case I will go back to checklist construction. Is this pruning? Never mind. Back to the point:
16 Set up some desktop shortcut macro thing in order to automatically start pomodoros when I open my laptop
17 Track time spent doing things useful to AI safety on a spreadsheet
18 Hey, I said “laptop”! Get a better mouse to make using the laptop more fun so I’m more likely to do hard things when using it
19 Get a better desk for more space for notes and to require less expensive shifting into/out of AI safety mode
20 On notes, use the index cards I have to make a proper zettelkasten as a cognitive aid
(Does this solve akrasia? Well, if I have better cognitive aids, then doing cognitively expensive things is easier, so I’m less likely to fail even with my current levels of willpower)
21 Start doing accountability things like promising to review a paper every X time period
22 I said levels of willpower—Google for interventions that increase conscientiousness (there’s gotta be some dodgy big-5 based things) and do those?
Back to the top of the tree
23 Quit my job because it’s using up energy that I could be using to do AI safety
24 Instead of doing my job, pretend to do my job while actually doing AI safety
25 Set up an AI safety screen on work laptop so it’s easy to switch over to doing AI safety during breaks or lunches
Hey, I said lunch
26 Use nutritionally complete meal replacements to save time/willpower that would be spent on food preparation
27 Use nutritionally complete meal replacements to ensure that nutrient intake keeps me in top physical form
28 Exercise (this improves everything, apparently) by running on a treadmill
29 By lifting weights
30 By jogging in a large circle
31 Become a monk and live an austere lifestyle without the distractions of rich food, wine, and lust
32 Become an anti-monk and live a rich lifestyle to ensure that no willpower is wasted on distractions
33 Specifically in vice use nicotine as a performance enhancing stimulant by smoking. Back to stimulants again I guess
34 … or by using nicotine patches or gum or something
35 By using nicotine only if I do AI safety things, in order to develop an addiction to AI safety
Hey, develop an addiction to doing AI safety! People go to serious lengths for addictions, so why not gate it on math?
36 Do so with something very addictive, like opioids
37 Use electric shocks to do classical conditioning
etc. there was a short sci-fi story about this kind of thing let me see if I can find it. Hey, actually, since I said sci-fi, adn this is a babble challenge:
38 Promise very hard to time travel back to this exact point in time, meet future self, recieve advice
(They’re not here :( Oh well) Back on that akrasia-solving:
39 Make up a far-future person who I am specifically working to save (they’re called Dub See Wun). Get invested in their internal life (they want to make their own star!). Feel an emotional connection to them. I’m doing it for them!
40 Specifically put up a “do it for them” poster modelled off the one in the Simpsons
41 DuckDuckGo “how to beat akrasia” and do the top suggestion
42 Adopt strategic probably false beliefs (the world will end in 1 year!! :0) in order to encourage a more aggressive search for strategies
“Aggressive search for strategies” is the virtue that the Sequences call “actually trying”, so in the Sequences-sphere
43 Go to a CFAR workshop, which I heard might be kind of useful towards this sort of thing
44 Or just read the CFAR booklet and apply the wisdom found in there
45 Or some sequence on Lesswrong with exercises that applies some CFARy wisdom
Of course all this willpower boosting and efficiency and stuff wouldn’t help if I was just doing the wrong thing faster (like that one Shen comic, you know the one). So:
46 Consider how much of what I think is working on AI safety is actually just self-actualisy math/CS stuff, throw that out, and actually try to solve the problem
47 Deliberately create and encourage a subagent in my mind that wants to do AI safety (call em Dub See Wun)
48 Adopt strategic infohazards in order to encourage a more focused and aggressive search for strategies
49 Post a lot about AI safety in public forums like Lesswrong so that I feel compelled to do AI safety in my private life in order to maintain the illusion that I’m some kind of AI-safety-doing-person
50 Stop doing this babble challenge at the correct time, and continue to do AI safety or sleep as in 1) or 2). Hey, this one seems good. Think I might try it now!