alt Hacker NewsI've thought, in other contexts too, how much easier innovation in script (in writing, glyphs, etc.) is when handwriting instead of printing text. Anyone could create their own Cistercian shorthand - and Medieval writers did use all sorts of shorthand.
Print requires a pre-composed set of glyphs with exceptions that are, I suppose, expensive (i.e., custom made by the printer). Typing right now on your computer, how easily can you create a custom glyph and share it? Look what the OP must do - stretch the bounds of typeface function, something few people are equipped to do.
If HN comments were hand written, each commenter could create custom glyphs on the fly. We could also draw diagrams and pictures, musical notation, draw lines pointing to different taxt from others - gloss each others comments.
Thinking about it (and wandering onto a tangent): If computers could process handwriting the same way as text encodings, would that be preferrable? I can't type as fast as I write but partly because I type far more. I could do so much more with a pen; it would be interesting to try. How well do LLMs handle handwriting recognition?
Replies
How many times in your life have you needed to create a new a glyph? Would the added expressive power make up for the inconvenience of having to explain the meaning of the novel symbols?
> How well do LLMs handle handwriting recognition?
Pretty well for neat modern handwriting, but much worse for cursive or messier writing. They also really struggle if the text is at an angle. I have some recent experience with a project where we tried to use LLMs to digitise handwritten specimen labels from the 19th and early 20th century, and the success rate was far too low to proceed with that approach.
Hallucination was also a common problem, with the output often replaced by a similar (but more common) name or word.
I’d assume you could improve the results by using a model trained specifically on handwriting data sets, grounding the model, or using existing purpose-built OCR tools - but frankly that’s above my pay grade.
You could just provide a canvas to draw on and share images instead of strings of characters. If you restrict it to just black and white, and crop to only the used area, it would probably compress reasonably well, but then you'd also have to deal with the fact that some people have awful handwriting, writing with a mouse is hard (I'm particularly awful at it, being a left-handed person who uses a mouse with my right hand. If I need to draw into a computer, I have to get my drawing tablet out of a drawer), you can't paste into google translate for people writing in foreign languages etc