A new label discovered in the Instagram app’s code would highlight content created from scratch or edited with generative artificial intelligence (AI) tools.
Meta could be preparing to start experiment with a new label with a subset of its users that would highlight AI-based posts, images and videos on Instagram.
Meta, along with other AI players, recently committed to creating a watermarking system for content that was created or enhanced using AI.
Instagram is working on an AI label
App researcher Alessandro Paluzzi discovered a screenshot of a notice about AI content in the Instagram app’s code, sharing it on Twitter. This seems to be an informational panel offering additional explanations about a piece of content.
Engadget says a spokesperson for Meta declined to comment.
“The creator or Meta said that this content was created or edited with AI,” it reads.
The notice includes a short generative AI explainer about using AI tools like Midjourney, Stable Diffusion and others to “create text, images and video from single descriptions.” There’s also a notice about identifying AI-based posts. “Content created with AI is typically labeled so that it can be easily identified,” it reads.
The feature itself uses AI to automate the discovery of AI-based content because some creators may prefer to avoid such disclosures. Meta’s heart is in the right place—the company is all too aware that AI can be leveraged to effectively spread dangerous misinformation across social media services.
— Alessandro Paluzzi (@alex193a) July 30, 2023
Paluzzi discovered earlier signs of other generative tools that could come to Instagram, like artistic generative AI tools for editing images for stories with text prompts and the ability to summarize long messages.
AI labels everywhere, please!
Generative AI isn’t just the latest buzzword in Silicon Valley, and many companies now experiment with ways to leverage AI to improve their offerings.
But as more people use AI to create images and videos, there’s a legitimate concern about properly highlighting content created by machines.
This is critical for social networks. Look, I don’t want to make a fool of myself by liking that gorgeous photo of a beautiful vista you published on Instagram just because you didn’t bother to clarify it was AI work, not yours.
We need a clear, straightforward way to label AI content everywhere.
Even AI-generated summaries of YouTube videos ought to be labeled appropriately to help humans distinguish AI-based work from human creations.
Doing the right thing
There’s some hope that companies will do the right thing. Recently, Google, Microsoft, OpenAI and other AI companies promised the White House they’ll develop a watermarking system to inform users when content is AI-generated.
So, this will become crucially important as more and more content shared online every day gets created or tweaked using AI. Meta will also release new products using LLaMA 2, its large language model, that’ll work across its services so the need for AI labelling will become even more important.