Critical AI Literacy -- How do AI Detectors Work?

[@Minarctic]

Loved it, very informative. Keep up the good work ❤🎉

[@MoxieLearn]

Thank you so much 😁

[@BenyRubinstein]

Great stuff, keep it up!

[@MoxieLearn]

Thanks, will do!

[@Franchise-infoCa]

Why are image generators so bad with spelling, even when you yell at them? Thanks.

[@MoxieLearn]

LOL. It's such a good question. Here's an explanation: Visual Pattern Focus: Image generators prioritize recognizing and reproducing visual data over understanding text, leading to inaccuracies in spelling within images. Lack of Linguistic Understanding: These models do not inherently grasp language rules (unlike an LLM such as ChatGPT), causing spelling errors when generating text. Because DALL-E is built into the upgraded GPT-4 subscription, it's confusing that they are really two totally separate systems: DALL-E vs. GPT. Complexity of Text Rendering: Correctly generating readable and accurately spelled text within images is a complex task that exceeds the models' primary visual objectives. Goal Misalignment: The main objective of image generators is visual content creation, not textual accuracy, making spelling accuracy a secondary concern. Hope that helps!

[@Franchise-infoCa]

@@MoxieLearn Yes, it does help understand the "why". But, have you got any tips on how to do it better! (Not that I mind yelling in all caps at DALL-E ...)

[@MoxieLearn]

​@@Franchise-infoCa Not yet, but I will investigate and make a video.

[@Franchise-infoCa]

@@MoxieLearnThanks I would love to see that!