Last Updated:
Google is going to have labels across YouTube, Lens and more later this year
Google is bringing new AI tools to detect and label images generated or edited using AI tools on the internet. Here’s what people can expect from this.
Google’s AI push in 2024 is getting another big boost as the company looks to thwart the dangers posed by deepfake and AI-generated content on the internet. The AI overviews are already giving you summarised AI-based results in Search but Google realises the perils of the new technology which can be easily used to dupe people and manipulate their thinking.
To make sure it doesn’t become a long-term problem, Google will start showing labels on AI-generated images or even images that are edited using the AI tools. Google will use its expertise in AI and bring the option for YouTube as well, but that’ll happen at a later date. For now, the feature will work in Search, Google Lens and the new Circle to Search option.
Google AI-Generated Label In Search: How It Could Work
Google has a clear plan on how the images will be labelled and how the company will use its back-end tech to identify these images from billions that are hosted on the platform. Having said that, Google’s task of identifying these images will be based on the Coalition for Content Provenance and Authenticity or (C2PA) metadata that gives you granular details about the visual content, its image history and also the software that has been used to edit/create them.
Knowing these details should ideally make it a cakewalk for Google to go about labelling the images, but there is a big challenge in making that happen. After all, Search hosts billions of images, and not all of them have the C2PA metadata. More importantly, this metadata can be removed or made unreadable, which puts Google in a spot with its mission to tackle the deepfake image problem.
The company says if the image contains C2PA metadata, users can go to the About this image section and fetch the details. The good thing is that companies like OpenAI, Amazon, Microsoft and Adobe have joined the C2PA bandwagon. However, the ratio of content with the C2PA metadata and without it is against the company’s strategy to label AI-generated photos.
Like any technology, Google will have to adapt to the evolving nature of AI-generated content, and ensure that people are aware of the images that are original, and the ones that have been edited.
