📹 It’s getting increasingly harder to spot AI generated images and videos. Recent revelations about YouTube suggest it might only get harder from here. AI-infused social media seems to have reached unnerving levels as viewers on the platform notice signs of AI enhancement in the videos they watch, according to the Atlantic. Even more unnerving is the fact that creators are not informed.

AI upscaling, trust downgrading. Multiple creators have spoken up about this apparent violation, with one YouTuber, Mr. Bravo, claiming his VCR videos with an “authentic 80s aesthetic” end up looking completely different after being uploaded. The filter used by YouTube erased the intended visual purpose behind his videos. Other YouTubers also complained about seeing the same effect on their YouTube shorts.

This over-enhancement in an image’s resolution is noticeable to viewers, and could lead to mistrust between a creator and their audience, who could reasonably assume they’re “cutting corners” and creating fake content using AI. Although these enhancements only appeared on select videos as part of the experiment for now, the prospect of a platform-wide change is not a promising one for both creators and viewers.

YouTube is “experimenting.” According to YouTube’s spokesperson Allison Toh, the Google subsidiary is using image enhancement technology — supposedly not generative AI — to sharpen content. More specifically, YouTube is trying to “unblur, denoise, and improve clarity” through traditional machine learning. These terms indicate use of a diffusion model to refine existing images that bears similarities to generative AI’s process to create new videos.

AI is already settled in. Apart from the unexplained image enhancement tests, YouTube has already embraced AI use through a feature that allows users to create and post AI-generated short videos. Other platforms like TikTok, Snapchat, and Meta’s Facebook, Instagram, and Whatsapp have also established AI-generated tools for content creation. And lest we forget, Meta was found lurking in its users’ private and unpublished photos for AI training purposes.

The AI image takeover is only getting started. Just announced, Meta secured access to Midjourney aesthetic technology to integrate into its own models and products, according to the Verge. Using Midjourney’s leading tech in AI image and video generation, Meta plans on upgrading feeds with AI-generated content. As Meta goes full steam ahead with AI imagery across its platforms, we could be close to witnessing social media turn into an AI-powered simulation of what it used to be.

Tags: