Speed of information input is a bottleneck for rationality.
Rationality (both in its epistemic and instrumental forms) demand having true beliefs- not just true beliefs about abstract logical facts, but also true beliefs about the world.
The only way to have true beliefs about the world is to observe the world, which we do all the time by various means: vision, hearing, smell, taste, touch (‘senses’ and ‘input’ are basically the same thing).
You might think that because we are constantly receiving a large stream of information, that information would be so abundant that it couldn’t be a bottleneck, but yet empirically this bottleneck does matter.
Some reasons include: (1) a lot of the information we input is highly correlated with information we already have (2) not all information is highly relevant to the things we care about (this is particularly important for instrumental rationality, less so for epistemic r.) (3) a lot of the information we receive gets thrown away if we don’t know what to do with it, or we don’t have the processing power to process it.
So while pure intake of information is quite high, the intake of information related to each of these criteria is oftentimes much lower. Addressing these criteria will allow you to be less bottlenecked on information input, and focus your pursuit of rationality on other bottlenecks.
A note on input speed, as relating to abstract and logical facts
Before I delve deeper into the criteria, both in terms of what I mean by them, and how they might be addressed, I want to note that it is not only our •epistemic rationality relating to the facts about the world• that is affected by information input, but also our •epistemic rationality relating to abstract and logical facts unrelated to the world• that is affected by input.
The human mind is an embedded physical computer, and has a limited processing speed, so even facts that a •theoretical perfect agent of rational cognition• can •calculate instantaneously, without reference to the world•, are unlikely to be known to your mind unless you input that abstract fact from the world. Even if you are Einstein, the total computational power (even among humans) existing right now is ~millions of times greater than what you have alone; much less the total computational power over all history.
The information you receive is highly correlated with itself
You probably witnessed the sun rising this morning. If you had never witnessed a sunrises before, the data you received would tell you a lot about the world. But since you have witnessed many thousands of sunrises, you gained very little information from that.
Same thing when you taste a strawberry, or see a thumbnail of the Mona Lisa in the corner of a powerpoint. You might notice some detail that you never noticed before, maybe many such details if you’re really paying attention, but most of the information you receive will be old news.
This criteria alone moves information from something that’s abundant, to something that’s rare. Information that you didn’t already know is rare.
Suppose you’re reading a book (printed in black and white, without illustrations, as many books are). Maybe you’re in a quiet library, maybe in a noisy café. The words on the book likely contain a lot of new information that you never heard before. But the texture of the book in your hands, the smell of the paper, the sound of the chatter in the café (or the quiet stillness of the café), the shape of the 26 letters of the latin alphabet, the colour of the blank paper and the black ink:
All of these factors contain no new information, you’ve experienced all those things many times before.
But do the words need to be black and white? No. You can colour words and use highlights. In my Anki notes (which I have complete control over the html and css of, and are both written and read only by me), I like to use grey text for information that is worth mentioning, but not the main focus of what I’m trying to communicate.
I sometimes find myself wishing that I could use grey text on LW. If I’m writing something, and want to mention an aside, (but don’t want to put it in a footnote), I can use parentheses to mark the point as secondary, but it’s a bit of a pain on the eyes and still looks complicated, wheras with greyed parentheticals, the eyes and brain know instinctively to skip over the greyed text while parsing the main structure of the text, then return to the parenthetical once I’ve parsed the main structure.
(Aside, I much prefer greying text over using smaller fonts, because the greyed text is still easily legible, at least if you don’t use too light of a shade, and visually flows well with the rest of the text)
And grey isn’t the only color that can be used. Colour can be used to highlight repeating patterns, or to show connections that aren’t otherwise obvious.
[maybe I’ll include an image illustrating what I mean, later]
You can also use color to add emphasis to things, or to categorize certain words or sayings.
I also think using various fonts is an under-explored way to increase information density. Of course we have italics and bold, which are essentially using varied fonts to convey information, but there’s a lot more variety in fonts that can be used (potentially adventageously, if done skillfully) to convey information.
Since we’re not talking about criteria (3), •ability to process info•, right now, I’ll also mention this:
You should be able to put on headphones as you’re reading and listen to a podcast completely unrelated to what you’re reading. Ideally you’d have two different podcasts, one in each ear. You at the same time can rig your lights to slightly change colour (just enough to be noticed, not enough to give you a headache), using morse code to convey the information of yet another book. You’d have yet another book being input by a tactile machine, which taps your skin. In fact you’d have many of these tappers lightly tapping many places over your body, allowing you to read many books at once.
I expect •your gut reaction to what I just said• already says a lot about bottleneck #3 (almost certainly about what I mean, hopefully also about how it might be overcome, if you are a person with a good imagination), so I’ll save both my own time, and your capacity to input information, and refrain from saying more on that.
And since I don’t want to spend all day writing, I’ll also refrain from discussing criteria #2 right now. If there’s (enough) interest in what else I have to say on this subject, I may write more later. (Otherwise, if I don’t desire to say more, I’ll write about other matters)
Speed of information input is a bottleneck for rationality
Speed of information input is a bottleneck for rationality.
Rationality (both in its epistemic and instrumental forms) demand having true beliefs- not just true beliefs about abstract logical facts, but also true beliefs about the world.
The only way to have true beliefs about the world is to observe the world, which we do all the time by various means: vision, hearing, smell, taste, touch (‘senses’ and ‘input’ are basically the same thing).
You might think that because we are constantly receiving a large stream of information, that information would be so abundant that it couldn’t be a bottleneck, but yet empirically this bottleneck does matter.
Some reasons include: (1) a lot of the information we input is highly correlated with information we already have (2) not all information is highly relevant to the things we care about (this is particularly important for instrumental rationality, less so for epistemic r.) (3) a lot of the information we receive gets thrown away if we don’t know what to do with it, or we don’t have the processing power to process it.
So while pure intake of information is quite high, the intake of information related to each of these criteria is oftentimes much lower. Addressing these criteria will allow you to be less bottlenecked on information input, and focus your pursuit of rationality on other bottlenecks.
A note on input speed, as relating to abstract and logical facts
Before I delve deeper into the criteria, both in terms of what I mean by them, and how they might be addressed, I want to note that it is not only our •epistemic rationality relating to the facts about the world• that is affected by information input, but also our •epistemic rationality relating to abstract and logical facts unrelated to the world• that is affected by input.
The human mind is an embedded physical computer, and has a limited processing speed, so even facts that a •theoretical perfect agent of rational cognition• can •calculate instantaneously, without reference to the world•, are unlikely to be known to your mind unless you input that abstract fact from the world. Even if you are Einstein, the total computational power (even among humans) existing right now is ~millions of times greater than what you have alone; much less the total computational power over all history.
The information you receive is highly correlated with itself
You probably witnessed the sun rising this morning. If you had never witnessed a sunrises before, the data you received would tell you a lot about the world. But since you have witnessed many thousands of sunrises, you gained very little information from that.
Same thing when you taste a strawberry, or see a thumbnail of the Mona Lisa in the corner of a powerpoint. You might notice some detail that you never noticed before, maybe many such details if you’re really paying attention, but most of the information you receive will be old news.
This criteria alone moves information from something that’s abundant, to something that’s rare. Information that you didn’t already know is rare.
Suppose you’re reading a book (printed in black and white, without illustrations, as many books are). Maybe you’re in a quiet library, maybe in a noisy café. The words on the book likely contain a lot of new information that you never heard before. But the texture of the book in your hands, the smell of the paper, the sound of the chatter in the café (or the quiet stillness of the café), the shape of the 26 letters of the latin alphabet, the colour of the blank paper and the black ink:
All of these factors contain no new information, you’ve experienced all those things many times before.
But do the words need to be black and white? No. You can colour words and use highlights. In my Anki notes (which I have complete control over the html and css of, and are both written and read only by me), I like to use grey text for information that is worth mentioning, but not the main focus of what I’m trying to communicate.
I sometimes find myself wishing that I could use grey text on LW. If I’m writing something, and want to mention an aside, (but don’t want to put it in a footnote), I can use parentheses to mark the point as secondary, but it’s a bit of a pain on the eyes and still looks complicated, wheras with greyed parentheticals, the eyes and brain know instinctively to skip over the greyed text while parsing the main structure of the text, then return to the parenthetical once I’ve parsed the main structure.
(Aside, I much prefer greying text over using smaller fonts, because the greyed text is still easily legible, at least if you don’t use too light of a shade, and visually flows well with the rest of the text)
And grey isn’t the only color that can be used. Colour can be used to highlight repeating patterns, or to show connections that aren’t otherwise obvious.
[maybe I’ll include an image illustrating what I mean, later]
You can also use color to add emphasis to things, or to categorize certain words or sayings.
I also think using various fonts is an under-explored way to increase information density. Of course we have italics and bold, which are essentially using varied fonts to convey information, but there’s a lot more variety in fonts that can be used (potentially adventageously, if done skillfully) to convey information.
Since we’re not talking about criteria (3), •ability to process info•, right now, I’ll also mention this:
You should be able to put on headphones as you’re reading and listen to a podcast completely unrelated to what you’re reading. Ideally you’d have two different podcasts, one in each ear. You at the same time can rig your lights to slightly change colour (just enough to be noticed, not enough to give you a headache), using morse code to convey the information of yet another book. You’d have yet another book being input by a tactile machine, which taps your skin. In fact you’d have many of these tappers lightly tapping many places over your body, allowing you to read many books at once.
I expect •your gut reaction to what I just said• already says a lot about bottleneck #3 (almost certainly about what I mean, hopefully also about how it might be overcome, if you are a person with a good imagination), so I’ll save both my own time, and your capacity to input information, and refrain from saying more on that.
And since I don’t want to spend all day writing, I’ll also refrain from discussing criteria #2 right now. If there’s (enough) interest in what else I have to say on this subject, I may write more later. (Otherwise, if I don’t desire to say more, I’ll write about other matters)