I’ve been enjoying the new set of Sequences. I wasn’t around when the earlier Sequences were being written; It’s like the difference between reading a series of books all in one go, versus being part of the culture, reading them one at a time, and engaging in discussion in between. So thanks to Eliezer for posting them!
I really liked how there was an ending koan in the last post. It prompted discussion. I tried to think of a good prompt to post for this one, but couldn’t. Anyone have some good ideas?
I was planning to paint my boat today. There’s already a coat of paint on it, drying. If I overpaint today, that’s optimal. If I wait till tomorrow, then I’ll have to sand it down first.
It looks like it might rain, but the forecast is good. I don’t know what effect rain will have on newly applied paint, or indeed on the current partly dried surface.
Do I spend the afternoon painting the boat or carry on sitting in a coffee shop reading Less Wrong?
Impractical, as it happens. I eventually solved the problem by going home, changing into painting clothes, cleaning brushes, arranging tools and stirring paint. At that point it started raining heavily. So I undid all that in the rain, changed back into dry clothes, went back to the coffee shop and am now reading Less Wrong again. I think I just failed rationality for ever.
I don’t think it’s possible to fail rationality “for ever”, as long as you are in a state where you can make observations, record memories, formulate goals, plan and take actions. Though you do seem to have been a bit unfortunate in the timing of the precipitation.
Hmmm. It seems that I should add “as long as you are able to reassign all priors of 1 to priors of 0.999999999, and all priors of 0 to priors of 0.000000001” to my list of exceptions. (It won’t fix the agent immediately, but it will place the agent in a situation of being able to fix itself, given sufficient observations and updates).
Perhaps a perfect agent should occasionally—very occasionally—perturb a random selection of its own priors by some very small factor (10^-10 or smaller) in order to avoid such a potential mathematical dead end?
I couldn’t think of a koan-y question, but here is a discussion prompt.
Let’s make a Worksheet!
Let’s come up with some practice examples of the 2x2 matrix (such as the “Being Lost or Not” example in the OP), that people can fill out. The examples should be short (single paragraph) everyday type problems that people can relate to. Submit examples in the comments. I’ll take the best and put them in a worksheet in Google docs, and link to it here.
That way, when people in the future come and read this post, they have an activity to help them practice it. Also, people can use them at meetups if they want. Worksheets, of course, aren’t the BEST way to learn, but they’re better than nothing.
You’re at work, and you find yourself wanting very badly to make a certain, particularly funny-but-possibly-taken-as-offensive remark to your boss. The comment feels particularly witty, quick-minded and insightful.
(trying to think of stuff that’s fairly common and happens relatively often in everyday life)
You are leaving your home in the morning, to return in the evening; your day will involve quite a bit of walking and public transport. It is now warm and sunny, but you know that a temperature drop with heavy rains is forecasted for the afternoon. Looking out at the window and thinking of the walk in the sun and the crowded bus, you don’t feel like carrying around a coat and umbrella. You start thinking maybe the forecast is wrong...
Yes, that is clearly the optimal solution. I was assuming you don’t own those two items, or that you don’t have a handbag the right size or don’t want to use it—more plausible for a man that for a woman, I guess.
Carrying around a handbag in the first place happens to be something that I find annoying and risky. I’m prone to leaving it in easy-to-notice, easy-to-steal places or outright forgetting it in some public location.
Now that I think about that, that happened to me exactly once (as far as I can remember) with a handbag, though it happens to me very often¹ with other items such as keys, jackets, sweatshirts and sometimes my iPod. (I usually² eventually manage to recover them, but not always.) I guess that’s because I’m more likely to immediately notice that I’m missing my bag than that I’m missing my keys.
You are knitting a fitted garment. Let’s say it’s a sweater. You’ve been knitting for awhile, and you″re starting to get concerned it won’t fit the intended recipient. You can’t tell for sure, because your needle is too short to fully stretch it out, but you just have this feeling. This feeling you hope is wrong, because you don’t want to rip out and re-do all the ribbing you’ve just knit...
You are an ex-smoker overcome with a sudden craving after a particularly bad day, and your helpful friend offers you a cigarette “have just this one smoke!” to relieve tension. You know that anything less than a complete abstinence has a chance of kickstarting the habit.
If a stressful day is enough to give you a craving difficult to resist, I think that saying “anything less than complete abstinence has a chance of kickstarting the habit” is a misleading statement of how it works. It might be more accurate to say that every cigarette you have is one cigarette closer to having a habit you need to kick. It seems, in fact, that there’s sort of a gradient of average craving from abstinence all the way up to two packs a day, with variances around those averages. It seems a bit obfuscatory to suggest that “complete abstinence” is the deciding factor, especially when considering the question “When does complete abstinence start? Why doesn’t it start after the next cigarette?” After all, the “real” complete abstinence has already failed, if you had to quit smoking in the first place.
. . . but that’s kind of off the topic of the worksheet example.
Sharing this sentiment. I’m particularly impressed with the cartoon diagrams. They’re visually very appealing, and they encapsulate an idea in a way that takes just enough thought to untangle that I feel like it makes me engage with the conceptual message.
Same here, I’m certainly happy that this new sequence is starting. I devoured the old sequences, but being forced to stop and digest these makes them feel more impactful.
I’d be curious to see how much more powerful the sequences could be if they all had Koans, too, especially if they were wrapped up in an interactive shell and you had to answer them before the rest of the article (and/or the next one(s)) would show up. Not as good as a Bayesian Dojo, but there doesn’t seem to be enough Beisusenseitachi around to really be effective on that front.
I’ve been enjoying the new set of Sequences. I wasn’t around when the earlier Sequences were being written; It’s like the difference between reading a series of books all in one go, versus being part of the culture, reading them one at a time, and engaging in discussion in between. So thanks to Eliezer for posting them!
I really liked how there was an ending koan in the last post. It prompted discussion. I tried to think of a good prompt to post for this one, but couldn’t. Anyone have some good ideas?
Also, Skill #2 made me think of this optical illusion
I was planning to paint my boat today. There’s already a coat of paint on it, drying. If I overpaint today, that’s optimal. If I wait till tomorrow, then I’ll have to sand it down first.
It looks like it might rain, but the forecast is good. I don’t know what effect rain will have on newly applied paint, or indeed on the current partly dried surface.
Do I spend the afternoon painting the boat or carry on sitting in a coffee shop reading Less Wrong?
LessWrong will still be there tomorrow. The optimal opportunity to paint the boat won’t be.
Is it possible to protect the boat from rain in some manner, such as leaving it under a roof?
Impractical, as it happens. I eventually solved the problem by going home, changing into painting clothes, cleaning brushes, arranging tools and stirring paint. At that point it started raining heavily. So I undid all that in the rain, changed back into dry clothes, went back to the coffee shop and am now reading Less Wrong again. I think I just failed rationality for ever.
I don’t think it’s possible to fail rationality “for ever”, as long as you are in a state where you can make observations, record memories, formulate goals, plan and take actions. Though you do seem to have been a bit unfortunate in the timing of the precipitation.
You may already know this, but the phrase “fail x forever” is a thing.
Merely humanly impossible. If you are a more pure agent just assign probability “1” to enough things and you’ll be set.
Hmmm. It seems that I should add “as long as you are able to reassign all priors of 1 to priors of 0.999999999, and all priors of 0 to priors of 0.000000001” to my list of exceptions. (It won’t fix the agent immediately, but it will place the agent in a situation of being able to fix itself, given sufficient observations and updates).
That’s not the only problem. An agent that assigns equal probability to all possible experiences will never update.
Oh, that’s sneaky.
Perhaps a perfect agent should occasionally—very occasionally—perturb a random selection of its own priors by some very small factor (10^-10 or smaller) in order to avoid such a potential mathematical dead end?
Nice try, but random perturbations won’t help here.
I think that this re-emphasises the importance of good priors.
I couldn’t think of a koan-y question, but here is a discussion prompt.
Let’s make a Worksheet!
Let’s come up with some practice examples of the 2x2 matrix (such as the “Being Lost or Not” example in the OP), that people can fill out. The examples should be short (single paragraph) everyday type problems that people can relate to. Submit examples in the comments. I’ll take the best and put them in a worksheet in Google docs, and link to it here.
That way, when people in the future come and read this post, they have an activity to help them practice it. Also, people can use them at meetups if they want. Worksheets, of course, aren’t the BEST way to learn, but they’re better than nothing.
You’re at work, and you find yourself wanting very badly to make a certain, particularly funny-but-possibly-taken-as-offensive remark to your boss. The comment feels particularly witty, quick-minded and insightful.
(trying to think of stuff that’s fairly common and happens relatively often in everyday life)
You are leaving your home in the morning, to return in the evening; your day will involve quite a bit of walking and public transport. It is now warm and sunny, but you know that a temperature drop with heavy rains is forecasted for the afternoon. Looking out at the window and thinking of the walk in the sun and the crowded bus, you don’t feel like carrying around a coat and umbrella. You start thinking maybe the forecast is wrong...
I put a pocket umbrella and/or a foldable raincoat into my handbag. Duh.
Yes, that is clearly the optimal solution. I was assuming you don’t own those two items, or that you don’t have a handbag the right size or don’t want to use it—more plausible for a man that for a woman, I guess.
Carrying around a handbag in the first place happens to be something that I find annoying and risky. I’m prone to leaving it in easy-to-notice, easy-to-steal places or outright forgetting it in some public location.
Now that I think about that, that happened to me exactly once (as far as I can remember) with a handbag, though it happens to me very often¹ with other items such as keys, jackets, sweatshirts and sometimes my iPod. (I usually² eventually manage to recover them, but not always.) I guess that’s because I’m more likely to immediately notice that I’m missing my bag than that I’m missing my keys.
Around once per month in average.
Around 90% of the times.
What immediately comes to mind for me:
You are knitting a fitted garment. Let’s say it’s a sweater. You’ve been knitting for awhile, and you″re starting to get concerned it won’t fit the intended recipient. You can’t tell for sure, because your needle is too short to fully stretch it out, but you just have this feeling. This feeling you hope is wrong, because you don’t want to rip out and re-do all the ribbing you’ve just knit...
That’s time for a new set of knitting needles, and empiricisim. I have 60in cables.
You are an ex-smoker overcome with a sudden craving after a particularly bad day, and your helpful friend offers you a cigarette “have just this one smoke!” to relieve tension. You know that anything less than a complete abstinence has a chance of kickstarting the habit.
If a stressful day is enough to give you a craving difficult to resist, I think that saying “anything less than complete abstinence has a chance of kickstarting the habit” is a misleading statement of how it works. It might be more accurate to say that every cigarette you have is one cigarette closer to having a habit you need to kick. It seems, in fact, that there’s sort of a gradient of average craving from abstinence all the way up to two packs a day, with variances around those averages. It seems a bit obfuscatory to suggest that “complete abstinence” is the deciding factor, especially when considering the question “When does complete abstinence start? Why doesn’t it start after the next cigarette?” After all, the “real” complete abstinence has already failed, if you had to quit smoking in the first place.
. . . but that’s kind of off the topic of the worksheet example.
Sharing this sentiment. I’m particularly impressed with the cartoon diagrams. They’re visually very appealing, and they encapsulate an idea in a way that takes just enough thought to untangle that I feel like it makes me engage with the conceptual message.
Same here, I’m certainly happy that this new sequence is starting. I devoured the old sequences, but being forced to stop and digest these makes them feel more impactful.
I’d be curious to see how much more powerful the sequences could be if they all had Koans, too, especially if they were wrapped up in an interactive shell and you had to answer them before the rest of the article (and/or the next one(s)) would show up. Not as good as a Bayesian Dojo, but there doesn’t seem to be enough Beisusenseitachi around to really be effective on that front.