To try everything Brilliant has to offer-free-for a full 30 days, visit brilliant.org/PurpleMind/ . You’ll also get 20% off an annual premium subscription.
Since the frequencies of letters and words don't follow the frequencies that all languages fallow (zipfs law) future archeologists will quickly be able to tell its meaningless
@@iamdigory Give the code a couple of updates on the hands of the community and they will probably create a version in which when you press generate the svg image, it instead generates a number of characters or silabes (probably a slider let's you decide) then uses that decided set of characters to generate the image placing them in random cinfigurations, now you can obey zipfs law with total gibberish
I think the springs on the pen add like 90% of the artistic style. All the parameters are nice and all, but the resulting SVG file still looks... well, alien. Every single generated image doesn't look like something that can be made by a human, so adding an actual physical solution to imitate a human hand feels kind of like a cheat? I wonder how complicated it would be to produce the same handwritten-like images, but purely with code, without resorting to physical mechanisms. The project itself is amazing though!
Probably you just add another variable like "direction of the stroke" which randomizes a percentage of lines to have more thickness at one side of the line, so it looks like a pen pressed against the "paper" and then was dragged in one direction to create the stroke. I think that is basically it
@@isayinyang93 The lines on paper also look slightly curved, especially around the connection points. That would probably require some Bezier curves to make it work. True though, I think this algorithm can be improved in more or less the same way
This is something we experimented with actually. That would make sense to do because real languages have a non-uniform distribution of characters in text. However the results when we tried this did not really look much different from an implementation with a uniform distribution of the characters, so we went with that instead to simplify things.
Maybe then it is best to use a uniform character distribution so people that discover the produced texts can tell it's gibberish simply by analyzing the character distribution.
Me: Do I have the permission to let my computer generate characters? My Computer: *NO*. ABSOLUTELY NOT. YOU SHOULD KNOW THAT. I'M NOT MADE FOR MAKING CHARACTERS. Me: Chill It'll only take 10 Megabytes. My Computer: Oh ok then Me: *generates a generator that generates ninety-one charcters*
I like how some anime invent pseudo writings for their world. Somtimes those are just fancy gibberish font for english letters or japaneese, but somtimes they actually gibberish writing.
I am a first year studying linguistics and CS with a focus in computational linguistics, nlp, and AI. I'm also a conlanger. A future project idea of mine is to develop a chatbot that can understand my language. That's the dream...
No. Soviet linguists proved that Book of Veles may not be in a real language. Not just pre-Christianization Slavic, not any Slavic, but not even real. (Because the Book popped up abroad, no one believed them, but fact.) This simple computer model won’t fool them.
Great intersection of technology, art and languages. Fascinating to watch the huge variety of styles and scripts. Calligraphy, cryptography, autography, xenography, robography, tachygraphy.
The simplest approach I can think of is to use a PRNG and map each bit of the output to a stroke. That will lead to some characters that are much busier than others, so maybe come up with a few sets of 8 or 16 strokes that bit ranges can pick.
Is there some way to possibly input this code into some sort of generator that would output a result in real time? Or maybe it exists already and i just havent found it?
You should try coming up with different algorithms to add even more variety, perhaps imitating scripts from different areas of the world like Southeast Asia, Africa, etc. Maybe you could even take inspiration from artificially constructed scripts or even write out sample text of constructed languages.
If I created a language that does not exist… 1. Create script by hand. Sorry, neither cannot trust it to software nor know how to make it programmatically. 2. Take statistics from a real language, port it to script. 3. Invent some small common words like THE. 4. Then write a simple statistical model to invent normal words intertwining with common ones.
@@Salamander876 No, it’s a gibberish language. To make a conlang, you should invent (or steal) phonology, syntax, a few words, and THEN script - spoken word is always the first. Computer is almost useless here (except as a very advanced typewriter).
@@Salamander876 In the xUSSR there were no geeks like Tolkien, but gibberish languages bloom everywhere. Our main example is animal language from Doctor Aibolit. (It has translations for one or two words, but it’s still gibberish.)
The problem i see here is that your words need a letter model. In the example you showed, there are words with tones after characters repeated in the word. In short it's not just about length, but also about entropy.
As English also has similar statistical patterns, proving that the Voynich manuscript is meaningless would have to involve evidence from the Voynich manuscript itself, which is not present in the video.
@purplenanite That's fair. My question wasn't phrased in the most robust way. I was mostly curious if it stands to reason that the author of said manuscript could have simply been nerdy and had an interest in asemic writing, as the creators (and viewers) of this video are. So "prove" was inaccurate to use here, but the possibility is a funny thought. To me at least.
It has been proposed multiple times that it may be meaningless, but given that it was written in the 1500s by hand on vellum it is very unlikely. It is also accompanied by many illustrations that may hint at its meaning, so it is unlikely.
![The Dark Matter of AI [Mechanistic Interpretability]](http://i.ytimg.com/vi/UGO_Ehywuxc/mqdefault.jpg)
To try everything Brilliant has to offer-free-for a full 30 days, visit brilliant.org/PurpleMind/ . You’ll also get 20% off an annual premium subscription.
You guys are ready to confuse the heck out of some archaeologists in the future
Lol
Since the frequencies of letters and words don't follow the frequencies that all languages fallow (zipfs law) future archeologists will quickly be able to tell its meaningless
@@iamdigory Give the code a couple of updates on the hands of the community and they will probably create a version in which when you press generate the svg image, it instead generates a number of characters or silabes (probably a slider let's you decide) then uses that decided set of characters to generate the image placing them in random cinfigurations, now you can obey zipfs law with total gibberish
@@iamdigory except in the future this law won't be a thing, 9 languages die per year according to kurzgesagt
@@Guys-s5v I am not clear on what the rate of language death has to do with zipfs law
Voynich Manuscript Pt. 2: Electric Boogaloo
codex seraphinianus
天书
Me eket? Sem alsa mon abi no basili! 🎉
I think the springs on the pen add like 90% of the artistic style. All the parameters are nice and all, but the resulting SVG file still looks... well, alien. Every single generated image doesn't look like something that can be made by a human, so adding an actual physical solution to imitate a human hand feels kind of like a cheat? I wonder how complicated it would be to produce the same handwritten-like images, but purely with code, without resorting to physical mechanisms. The project itself is amazing though!
Probably you just add another variable like "direction of the stroke" which randomizes a percentage of lines to have more thickness at one side of the line, so it looks like a pen pressed against the "paper" and then was dragged in one direction to create the stroke. I think that is basically it
@@isayinyang93 The lines on paper also look slightly curved, especially around the connection points. That would probably require some Bezier curves to make it work. True though, I think this algorithm can be improved in more or less the same way
i think the different characters shouldn't be the exact same frequency
This is something we experimented with actually. That would make sense to do because real languages have a non-uniform distribution of characters in text. However the results when we tried this did not really look much different from an implementation with a uniform distribution of the characters, so we went with that instead to simplify things.
Maybe then it is best to use a uniform character distribution so people that discover the produced texts can tell it's gibberish simply by analyzing the character distribution.
Love this, not to spoil the ending but the takeaway there really resonated- worth watching fully
Thanks! Glad you enjoyed.
Me: Do I have the permission to let my computer generate characters?
My Computer: *NO*. ABSOLUTELY NOT. YOU SHOULD KNOW THAT. I'M NOT MADE FOR MAKING CHARACTERS.
Me: Chill It'll only take 10 Megabytes.
My Computer: Oh ok then
Me: *generates a generator that generates ninety-one charcters*
I like how some anime invent pseudo writings for their world. Somtimes those are just fancy gibberish font for english letters or japaneese, but somtimes they actually gibberish writing.
I am a first year studying linguistics and CS with a focus in computational linguistics, nlp, and AI. I'm also a conlanger. A future project idea of mine is to develop a chatbot that can understand my language. That's the dream...
Sounds like a neat project!
0:42 almost closed the video since thought it was to simple of a task to require AI. Good thing I watched until the end)
Haha true
such an underrated channel keep uploading im sure ull blow up
Thanks so much!
Hmmm... Now how do I use this machine to sell an ancient manuscrpt
Bro finna start a new Scientology or Mormonism 😭
No. Soviet linguists proved that Book of Veles may not be in a real language. Not just pre-Christianization Slavic, not any Slavic, but not even real. (Because the Book popped up abroad, no one believed them, but fact.) This simple computer model won’t fool them.
Bill Wurtz: You can make a religion out of this!
Great intersection of technology, art and languages. Fascinating to watch the huge variety of styles and scripts. Calligraphy, cryptography, autography, xenography, robography, tachygraphy.
Thank you! Glad you enjoyed.
When is the answer to the voronoi puzzle coming?
(BTW fantastic video and really unique topic)
The simplest approach I can think of is to use a PRNG and map each bit of the output to a stroke. That will lead to some characters that are much busier than others, so maybe come up with a few sets of 8 or 16 strokes that bit ranges can pick.
Is there some way to possibly input this code into some sort of generator that would output a result in real time? Or maybe it exists already and i just havent found it?
You should try coming up with different algorithms to add even more variety, perhaps imitating scripts from different areas of the world like Southeast Asia, Africa, etc. Maybe you could even take inspiration from artificially constructed scripts or even write out sample text of constructed languages.
love this, short sweet, interesting!
Thanks so much!
i love asemic writing
If I created a language that does not exist…
1. Create script by hand. Sorry, neither cannot trust it to software nor know how to make it programmatically.
2. Take statistics from a real language, port it to script.
3. Invent some small common words like THE.
4. Then write a simple statistical model to invent normal words intertwining with common ones.
That’s called a CONLANG. You’re making a CONLANG.
@@Salamander876 No, it’s a gibberish language. To make a conlang, you should invent (or steal) phonology, syntax, a few words, and THEN script - spoken word is always the first. Computer is almost useless here (except as a very advanced typewriter).
@@Salamander876 In the xUSSR there were no geeks like Tolkien, but gibberish languages bloom everywhere. Our main example is animal language from Doctor Aibolit. (It has translations for one or two words, but it’s still gibberish.)
@@Mercury13kiev okay
@@Mercury13kiev thanks. that was kind of embarrassing what I did.
The problem i see here is that your words need a letter model. In the example you showed, there are words with tones after characters repeated in the word. In short it's not just about length, but also about entropy.
Very good video!
Thank you!
1:55 i misheard that as "common jokes" what is wrong with me
Did this video accidentally prove that the Voynich Manuscript is actually meaningless?
As English also has similar statistical patterns, proving that the Voynich manuscript is meaningless would have to involve evidence from the Voynich manuscript itself, which is not present in the video.
@purplenanite That's fair. My question wasn't phrased in the most robust way. I was mostly curious if it stands to reason that the author of said manuscript could have simply been nerdy and had an interest in asemic writing, as the creators (and viewers) of this video are. So "prove" was inaccurate to use here, but the possibility is a funny thought. To me at least.
It has been proposed multiple times that it may be meaningless, but given that it was written in the 1500s by hand on vellum it is very unlikely. It is also accompanied by many illustrations that may hint at its meaning, so it is unlikely.
What about an algorithm for asemic speech? Is it possible to make a TTS produce its own input?
Ooh, neat idea! What's a TTS?
@@PurpleMindCS text to speech
@@brinleyhamer729 Ah
Roman alphabet, not English
I thought this was CScript for a second.
the characters look too complicated to actually work as an alphabet
I did not enjoy the bait and switch
It'll eventually writr.shakespeare.