Anonymous
8/3/2025, 4:27:40 AM No.106122588
There's a lot of progress in the area of ai/ml vision, however (modified) Teeline shorthand presents huge difficulty.
The basic idea of Teeline shorthand is each letter is represented by a certain stroke, that has up to a few ways to write it. You drop some letters, and then connect the lines.
It gets somewhat more complicated, but the essence is there.
Obviously, you can just train all the available and possible "outlines" (teeline words). And you could also use ai to generate novel Teeline outlines.
But Teeline itself lends itself to hiding from ai - but remaining pretty obvious to humans. This is akin to a manual form of captcha - the ai bots can't solve the Teeline "captcha".
But it extends quite far beyond that, because Teeline shorthand also can be modified. Basically, you can change the rules at the moment you create a message. At first, it will be possible to provide the rules in regular text, because the ai will be unable to follow the instructions (basically, reasoning models need to reason on images a lot better). But also, you can just basically provide hints, and the other person can catch on. ai really struggles in this area, and will for a long time.
How is Teeline useful? Only between people who know Teeline. If people don't know Teeline, it's a real pain to try and figure out.
Fun: it would be fun to challenge the world's ai developers to try and defeat us:
1. mere detection isn't that easy. teeline outlines are easily hidden, and can be sneakily merged into other lines, or shapes even (or... even animation paths, like a Blender camera path following a naughty word)
2. reading Teeline
3. generating Teeline
The basic idea of Teeline shorthand is each letter is represented by a certain stroke, that has up to a few ways to write it. You drop some letters, and then connect the lines.
It gets somewhat more complicated, but the essence is there.
Obviously, you can just train all the available and possible "outlines" (teeline words). And you could also use ai to generate novel Teeline outlines.
But Teeline itself lends itself to hiding from ai - but remaining pretty obvious to humans. This is akin to a manual form of captcha - the ai bots can't solve the Teeline "captcha".
But it extends quite far beyond that, because Teeline shorthand also can be modified. Basically, you can change the rules at the moment you create a message. At first, it will be possible to provide the rules in regular text, because the ai will be unable to follow the instructions (basically, reasoning models need to reason on images a lot better). But also, you can just basically provide hints, and the other person can catch on. ai really struggles in this area, and will for a long time.
How is Teeline useful? Only between people who know Teeline. If people don't know Teeline, it's a real pain to try and figure out.
Fun: it would be fun to challenge the world's ai developers to try and defeat us:
1. mere detection isn't that easy. teeline outlines are easily hidden, and can be sneakily merged into other lines, or shapes even (or... even animation paths, like a Blender camera path following a naughty word)
2. reading Teeline
3. generating Teeline
Replies: