@nocci The value is in the pictures themselves, describing them doesn't really make sense. If you have poor sight, the post really doesn't have much of value to offer.
A square, black-and-gold printed circuit board is lying on a light gray surface.
The boards surface is covered by a dense, maze-like gold trace pattern that appears to be made from a small number of extremely long continuous paths rather than many separate short traces. The lines run in tight, evenly spaced meanders across the entire surface, bending around holes and circular pads like contour lines on a topographic map. The result is an intricate interwoven field of conductive tracks, possibly two large nested traces or electrode areas
Around each circular hole, the pattern opens up into ring-shaped contact areas. From there, the traces spread outward and weave between the other holes. Some sections look like tiny rectangular spirals or switchbacks, while other areas form long parallel lanes that snake around obstacles. The spacing between the lines is very consistent, which makes the whole board feel deliberately engineered rather than decorative, even though it also has a very ornamental, almost art-deco technical look.
@nocci Ok, I've gotta level with you, the really big thing that makes me anxious about this is primarily the generation of easily scrapable AI training data. I am so viscerally against providing this data, that I understand there are other improvements to other people's lives that are being lost in order for me to feel less icky, I am doing the best I can to find a middle-ground.
Your comment is a really good explanation - I just can't bring myself to do that knowing what it will be used for.
Any LLM could take the image, analyze it, and then generate a result similar to the one I just did. You don’t have to write what the thing is for. (I don’t know either.)
But an alt-text that says “A circuit board” feels like it’s actively excluding a group of people. I’d feel really offended. You might as well just leave out the alt-text altogether, since it provides just as much information
@nocci They can generate a very similar thing, but providing fresh, human data is what is of the most value. There has to be a way to look out for those who have difficulty seeing without seeding new low-hanging training data. I just don't know what it is.
Please respect my disdain for anything that could remotely help LLMs, but I really have such a serious moral qualm with them, I don't even use them for any form of workflow and would be willing to lose my job over it.
