In planecrash EY shows technique of “never repeating same thoughts”. And it can look useful, like what if you was repeating your thoughts for 10 times? You could get 10x more thoughts by that. Sort of like if you had 10 times more time for thinking, which looks really useful.
But when I actually started to practice it, I noticed, that actually a lot of time I spend on replacing last word by better synonym, a lot of time I rephrase the end of sentence, or whole sentence, or whole paragraph. That I spend a lot of time trying to think up better words to say.
And the same thoughts… I repeated them not just 10 times for whole time. I repeated a lot of them hundreds times per DAY.
A lot of time I wanted to start new thought and I for a long time (~4/5) explained circumstances, next premises, arguments etc. Like if I was doing that in speech. Because my internal dialogue felt like speech, it wasn’t really different that it was in my imagination, not in real world.
So if somebody will say that Yudkowsky thinks 1000 times faster, I can easily believe in that, at least for myself in the span of days. And I can’t even compare in the span of weeks.
But never repeating same thoughts is only one technique of optimising your internal dialogue.
I used a lot phrases which were not actually meaningful in thoughts, only in speech. And these probably were 2⁄3 of my internal dialogues.
I was trying not to use abbreviations, slang etc, and of course not refer to half of ideas like “because of that” if it was something in my visual imagination, or some knowledge I am sure in. I was trying to make my internal speech sound properly in grammatical and rhetorical sense, instead of being maximally compressed.
I was trying to speak only language that I know, Russian or English. Not to mix Russian and English sentences, and even mix words inside of sentences using Russian, English, German, Italian, Esperanto, Japanese, Latin, Greek and some languages that I know a little.
I was not replacing long words and phrases in my speech by imagining of how they will write. And also wasn’t doing the same for math formulas and code. And geometrical blueprints. And schemes. And full visual images or even collages of them.
I was trying to speak legibly, and not in maximal possible speed, but in speed on which people in real dialogues didn’t start to complain that I speak too fast. Which by itself could make my internal dialogue 2-3 times slower.
I was trying to make my internal speech more precise by sending a little tension on my speaking muscles, instead of just doing it fully in imagination. And… Actually my muscles very much slower than my imagination, it by itself was doing my speech 3-4 times slower.
Visual thinking
But even if I will replace all my internal sentences into saying “that” and just mentally referring to idea, that also will not be the best result. As I noticed, just visual images come to my mind faster that I can say “that”, ~2 times faster.
And just the speed is only one parameter in which visual is better than audio. Our world used to refer to video and audio as the same category of information sources. Used to say that you can be visual of audio learner.
But consider the difference in size in video and audio files. 1 min of audio will be ~1mb, 1 min of video ~150mb, 150x (!) difference.
Consider a page of text, it’s something like 400 words, 2000 letters + 400 spaces, how much time do you need to see a page by eyes? Something like 1⁄4 of second maybe. And how much time do you need to listen it as audio? At the maximum speed of human speech it will be 12 syllables, 4 words/s, at x2 8 words, so 1⁄50, 1⁄100 for audio of visual. And on normal speed of speech it’s more something like 1⁄200, 1⁄400.
There are speed reading techniques, and even Wikipedia says that it can give you 5x speed of usual reading in speed (~200wpm, speed of fast speech).
But that is still just 20wps (1000wpm). Human reaction speed is 0.1-0.25s, and in this time yours eyes allegedly can recognise only 2-5 words?? When you are looking on picture you are not restricted by recognising only 2-5 things at the moment, you can see it as a whole.
Though there is a problem of quality of seeing. How how much words you can see in enough quality to recognise? I used a flashlight to create a trace in my eye and next kept it at one word in the center, it looks like I can see in enough quality a square 11x11 or 13x13, 121-169 chars, by one eye, 242-338 chars by both, which could be 50-70 words.
And it’s only for speed reading, in your mind there will be no such restrictions. 5x faster thinking by analogy with 5x faster reading via using visual track instead of audio, it’s only lower limit of possible.
And you don’t need to restrict yourself by usual speech rules than. You can use a lot of long terms and phrases like “utility function”, because length is no longer limits your speed.
Now there is a question, will such speech work? Shouldn’t speech be audio? I don’t think so because of deaf and unable to speak people, who can whole life use gesture language, “speak” by their hands and “hear” by their eyes.
And I suspect that limits on visual thinking should be very high.
Because it works in more natural to brain information processing mode. Not in consequence, but in parallel. Your brain have 200 tacts/s max in consequence, but has 80B neurons.
I was afraid that sentences have dependence of next from previous, so you can’t say them in parallel. But… It’s only in English and German words order is fixed. In my native Russian you can use words almost in any order, and they still will make almost the same sentence. So I think that no.
And imagining words of sentence visually instead of audibly is only the beginning. Next you can do so much. You can use as your speech emojis and even replace words just by pictures, (again, use code and math, schemes, schemes). You can keep threads of sentences in multiple languages at the same time.
You can see the whole sequence of your thoughts, not just current thought. You can group your thoughts into context, and remember these at one moment. You can colour your words like with synesthesia or like in IDE. You can add comments, and reactions, and marks for yourself.
You can you eyes independently to have two tracks of flat image like on computer screen, or you can use both to have 3d! Use distance, texture as additional type of data, like colour.
Practice
I was somewhat incredulous to all of these thoughts. But next I tried that. And all that actually works. Some time ago I noticed that I am 2-3 times when I am using my just beginner’s visual thinking. And I can be now even more faster in learning, because in difference from acting I can learn a few things simultaneously.
Also visual thinking actually helps a lot in navigating in thoughts, not repeating the same a lot of times, because I can just see continuation of this thought sequence.
Problems:
Visual thinking in my case refuses to “just work”, I need to every day start it by deliberate practice before it continues to work automatically.
And it’s prone to gradually fall out from your mind when I am using audio instead, like when thinking by audio, or talking (much less when listening), or even writing, because I have a habit to pronounce things which I write.
Another problems is that some days I can start using visual thinking almost right after getting up. And some other days my imagination just works worse and I struggle to start visual thinking for whole day.
Also I was used to express emotions by mentally saying sentences by intonation and I didn’t find good alternative for that in my visual thinking.
When I started I made a lot of mistakes in the process of training.
Like I thought that if I will make my thoughts less loud and intonated, visual thinking will come by itself. It didn’t. My better idea was just break my internal sentences on half and wait before I will be able to notice their meaning without pronouncing.
Something sort of like when I am trying to speak not very well known language, know what I need, but prohibit to myself to speak my native language, and next without words try to find words in another language.
Another mistake was trying to restrict visual thinking. In saying only things that I would say audibly. Or in imagining visual thought only in subjective space. Or in making them look like usual text, instead of just visualing words in any convenient order.
Intuition of “how it will be more convenient to do that mental action” are generally very useful to listen to, because they usually signal which mental actions will train better and faster.
Another mistake was trying to develop visual thinking as different mode, instead of shifting every my action to it just a little.
(That’s just short list of thoughts that come to my mind immediately, not even close to full list of ideas. And would be interested to know where do exists some similar ideas. Speed reading or Feynman’s modeling method do come to mind, but that’s only which I know.)
Optimising internal dialogue
Never repeating thoughts technique
In planecrash EY shows technique of “never repeating same thoughts”. And it can look useful, like what if you was repeating your thoughts for 10 times? You could get 10x more thoughts by that. Sort of like if you had 10 times more time for thinking, which looks really useful.
But when I actually started to practice it, I noticed, that actually a lot of time I spend on replacing last word by better synonym, a lot of time I rephrase the end of sentence, or whole sentence, or whole paragraph. That I spend a lot of time trying to think up better words to say.
And the same thoughts… I repeated them not just 10 times for whole time. I repeated a lot of them hundreds times per DAY.
A lot of time I wanted to start new thought and I for a long time (~4/5) explained circumstances, next premises, arguments etc. Like if I was doing that in speech. Because my internal dialogue felt like speech, it wasn’t really different that it was in my imagination, not in real world.
So if somebody will say that Yudkowsky thinks 1000 times faster, I can easily believe in that, at least for myself in the span of days. And I can’t even compare in the span of weeks.
But never repeating same thoughts is only one technique of optimising your internal dialogue.
I used a lot phrases which were not actually meaningful in thoughts, only in speech. And these probably were 2⁄3 of my internal dialogues.
I was trying not to use abbreviations, slang etc, and of course not refer to half of ideas like “because of that” if it was something in my visual imagination, or some knowledge I am sure in. I was trying to make my internal speech sound properly in grammatical and rhetorical sense, instead of being maximally compressed.
I was trying to speak only language that I know, Russian or English. Not to mix Russian and English sentences, and even mix words inside of sentences using Russian, English, German, Italian, Esperanto, Japanese, Latin, Greek and some languages that I know a little.
I was not replacing long words and phrases in my speech by imagining of how they will write. And also wasn’t doing the same for math formulas and code. And geometrical blueprints. And schemes. And full visual images or even collages of them.
I was trying to speak legibly, and not in maximal possible speed, but in speed on which people in real dialogues didn’t start to complain that I speak too fast. Which by itself could make my internal dialogue 2-3 times slower.
I was trying to make my internal speech more precise by sending a little tension on my speaking muscles, instead of just doing it fully in imagination. And… Actually my muscles very much slower than my imagination, it by itself was doing my speech 3-4 times slower.
Visual thinking
But even if I will replace all my internal sentences into saying “that” and just mentally referring to idea, that also will not be the best result. As I noticed, just visual images come to my mind faster that I can say “that”, ~2 times faster.
And just the speed is only one parameter in which visual is better than audio. Our world used to refer to video and audio as the same category of information sources. Used to say that you can be visual of audio learner.
But consider the difference in size in video and audio files. 1 min of audio will be ~1mb, 1 min of video ~150mb, 150x (!) difference.
Consider a page of text, it’s something like 400 words, 2000 letters + 400 spaces, how much time do you need to see a page by eyes? Something like 1⁄4 of second maybe. And how much time do you need to listen it as audio? At the maximum speed of human speech it will be 12 syllables, 4 words/s, at x2 8 words, so 1⁄50, 1⁄100 for audio of visual. And on normal speed of speech it’s more something like 1⁄200, 1⁄400.
There are speed reading techniques, and even Wikipedia says that it can give you 5x speed of usual reading in speed (~200wpm, speed of fast speech).
But that is still just 20wps (1000wpm). Human reaction speed is 0.1-0.25s, and in this time yours eyes allegedly can recognise only 2-5 words?? When you are looking on picture you are not restricted by recognising only 2-5 things at the moment, you can see it as a whole.
Though there is a problem of quality of seeing. How how much words you can see in enough quality to recognise? I used a flashlight to create a trace in my eye and next kept it at one word in the center, it looks like I can see in enough quality a square 11x11 or 13x13, 121-169 chars, by one eye, 242-338 chars by both, which could be 50-70 words.
And it’s only for speed reading, in your mind there will be no such restrictions. 5x faster thinking by analogy with 5x faster reading via using visual track instead of audio, it’s only lower limit of possible.
And you don’t need to restrict yourself by usual speech rules than. You can use a lot of long terms and phrases like “utility function”, because length is no longer limits your speed.
Now there is a question, will such speech work? Shouldn’t speech be audio? I don’t think so because of deaf and unable to speak people, who can whole life use gesture language, “speak” by their hands and “hear” by their eyes.
And I suspect that limits on visual thinking should be very high.
Because it works in more natural to brain information processing mode. Not in consequence, but in parallel. Your brain have 200 tacts/s max in consequence, but has 80B neurons.
I was afraid that sentences have dependence of next from previous, so you can’t say them in parallel. But… It’s only in English and German words order is fixed. In my native Russian you can use words almost in any order, and they still will make almost the same sentence. So I think that no.
And imagining words of sentence visually instead of audibly is only the beginning. Next you can do so much. You can use as your speech emojis and even replace words just by pictures, (again, use code and math, schemes, schemes). You can keep threads of sentences in multiple languages at the same time.
You can see the whole sequence of your thoughts, not just current thought. You can group your thoughts into context, and remember these at one moment. You can colour your words like with synesthesia or like in IDE. You can add comments, and reactions, and marks for yourself.
You can you eyes independently to have two tracks of flat image like on computer screen, or you can use both to have 3d! Use distance, texture as additional type of data, like colour.
Practice
I was somewhat incredulous to all of these thoughts. But next I tried that. And all that actually works. Some time ago I noticed that I am 2-3 times when I am using my just beginner’s visual thinking. And I can be now even more faster in learning, because in difference from acting I can learn a few things simultaneously.
Also visual thinking actually helps a lot in navigating in thoughts, not repeating the same a lot of times, because I can just see continuation of this thought sequence.
Problems:
Visual thinking in my case refuses to “just work”, I need to every day start it by deliberate practice before it continues to work automatically.
And it’s prone to gradually fall out from your mind when I am using audio instead, like when thinking by audio, or talking (much less when listening), or even writing, because I have a habit to pronounce things which I write.
Another problems is that some days I can start using visual thinking almost right after getting up. And some other days my imagination just works worse and I struggle to start visual thinking for whole day.
Also I was used to express emotions by mentally saying sentences by intonation and I didn’t find good alternative for that in my visual thinking.
When I started I made a lot of mistakes in the process of training.
Like I thought that if I will make my thoughts less loud and intonated, visual thinking will come by itself. It didn’t. My better idea was just break my internal sentences on half and wait before I will be able to notice their meaning without pronouncing.
Something sort of like when I am trying to speak not very well known language, know what I need, but prohibit to myself to speak my native language, and next without words try to find words in another language.
Another mistake was trying to restrict visual thinking. In saying only things that I would say audibly. Or in imagining visual thought only in subjective space. Or in making them look like usual text, instead of just visualing words in any convenient order.
Intuition of “how it will be more convenient to do that mental action” are generally very useful to listen to, because they usually signal which mental actions will train better and faster.
Another mistake was trying to develop visual thinking as different mode, instead of shifting every my action to it just a little.
(That’s just short list of thoughts that come to my mind immediately, not even close to full list of ideas. And would be interested to know where do exists some similar ideas. Speed reading or Feynman’s modeling method do come to mind, but that’s only which I know.)