#poisoning data

[willcodehtmlforfood]

Preventing AI Plagiarism With .ASS Subtitling

Around two years ago, the world was inundated with news about how generative AI or large language models would revolutionize the world. At t

At least as far as YouTube is concerned, the worst offenders of AI plagiarism work by downloading the video’s subtitles, passing them through some sort of AI model, and then generating another YouTube video based off of the original creator’s work. Most subtitle files are the fairly straightfoward .srt filetype which only allows for timing and text information. But a more obscure subtitle filetype known as Advanced SubStation Alpha, or .ass, allows for all kinds of subtitle customization like orientation, formatting, font types, colors, shadowing, and many others. YouTuber [f4mi] realized that using this subtitle system, extra garbage text could be placed in the subtitle filetype but set out of view of the video itself, either by placing the text outside the viewable area or increasing its transparency. So now when an AI crawler downloads the subtitle file it can’t distinguish real subtitles from the garbage placed into it.

Lol a video végén van még egy trükk a subtitle crawler llm-k átbaszására, nem lövöm le a poént

[facelessesper]

You know how spammers use similar characters to replace actual letters to get by filters? Doing that to your writing would give a slightly bizarre visual and utterly annihilate any basic scraping tools, because visual interpretation of characters is wildly more time and resource intensive than plaintext ripping.

It might help or severely hinder dyslexic people, though.

Also some browsers/OS might have trouble loading them?

ɬɧıʂ ƙıŋɖą ɬɧıŋɠ. ɬơɬąƖƖყ ʄųƈƙ ῳıɬɧ ɬɧɛ ʂƈཞą℘ıŋɠ ɖɛ۷ıƈɛ.

None of that will at all register correctly to a scraper. It's in Unicode, so it'll work wherever.

Boldtext.io is the one I used, but there's plenty of ways to do it. You could even do multiple at once and make undoing it basically impossible, but still have it be readable to humans (auto readers or whatever for the visually impaired, not so much...)

ᏇᏂᎩ ᏕᏂᎧᏬᏝᎴ ᏇᏋ ᎷᏗᏦᏋ ᏝᎥᎦᏋ ᏋᏗᏕᎥᏋᏒ ᎦᎧᏒ ᏖᏂᏋ ᏰᎧᏖᏕ ᏖᏂᏗᏖ ᏒᎥᎮ ᏋᏁᏖᎥᏒᏋ ᏇᏋᏰᏕᎥᏖᏋᏕ?

Just gotta pick the style you like that's as clear as possible while also being literal gibberish to any quick digital scan method.

₱Ʉ₮ ₮Ⱨł₴ ₴Ɇ₦₮Ɇ₦₵Ɇ ł₦₮Ø ₲ØØ₲ⱠɆ ₳₦Đ ₳₴₭ ł₮ ₮Ø ⱤɆ₳Đ ł₮ ØɄ₮ ⱠØɄĐ.

Or

ᑭᑌT TᕼIᔕ ᔕEᑎTEᑎᑕE IᑎTO GOOGᒪE ᗩᑎᗪ ᗩᔕK IT TO ᖇEᗩᗪ IT OᑌT ᒪOᑌᗪ.

That second one is almost plain, and none of those characters should be used that way. It just produces gibberish where it can figure out what a particular character is technically displaying.

Some ai might be able to handle this. It will still be an extra layer of defense.

I'll take reduced reading speed for total digital obscurity and I'll be happy.