Today I have seen the idea of poisoning image alt text with lies or other nonsense to make data scraping harder. I have little doubt about it working, sounds like something that would prove effective.
However, I will refuse to engage in this behaviour. To me, the function of alt text is not to train AI models. True, it may be a side effect of it. But were I to begin fighting that, I'd lose the main function; I would no longer be providing assistance to the genuine people behind the screen that rely on alt text, for any reason from a list too long to enumerate in this post. Or new ones I hadn't even considered, but should have seen coming.
By all means, fight the good fight. But please remember the people you are doing it for while you are at it. Alt text may not be the medium to take.
Writing inaccurate alt text to spoil its usefulness to LLM scrapers is such a misguided idea. I can't help but think it'd only come from people who are not and don't know anyone who benefits from images being described.
Because if they did, they'd know that there are already PLENTY poorly- or non-decribed images on the web. You don't need to add to that number!
And even if it became common, it'd still be a small fraction of an LLM's training data. So it wouldn't work. It'd just hurt humans.
Lass nicht zu, dass jemand
anderes deinen Tag ruiniert.
Es ist dein Tag.
Ruinier ihn selbst.
Diese Frage ist eine der Ersten, die sich viele Männer stellen oder gestellt bekommen, wenn sie ihre Liebe zu Kleidung entdecken, die man normalerweise in der Frauenabteilung findet. Lass uns also über dieses Thema sprechen.
Andere Leute haben einen Mixer in der Küche.
Ich habe einen Mixer in der Küche.
#Umzug