Google’s push towards enabling the creation of AI-generated articles is poised to introduce new challenges in content moderation for Google’s systems.
The rapid advancement of Google into the realm of AI signifies a convergence within different divisions of the company. For instance, while Search is leveraging AI to produce summarized results, the proliferation of AI-generated content across the internet will inevitably complicate the search process. This convergence will bring about a mix of noise, complexity, and excitement. Specifically, the teams at Google responsible for platform moderation, such as YouTube and Search, will find themselves intersecting with the units focused on facilitating content creation, like Gmail and Bard. By delineating these entities and connecting them, a plethora of intriguing issues are bound to surface.
In the latest clash of Google’s AI technologies, the Pixel 8 introduces its Best Take feature, utilizing AI to enable users to manipulate facial expressions in group photos to achieve “flawless” images. This development intersects with YouTube’s recent implementation of an AI-generated content policy that mandates authors to disclose any altered or synthetic content that appears realistic. The policy specifically highlights the necessity of labeling content that “realistically depicts an event that always happened,” with potential repercussions for failing to consistently label such artificially created content, including content removal and demonetization.
Moving forward, the Pixel 8, equipped with Best Take and Magic Editor in Google Photos, has the capability to generate lifelike images of events that never occurred. Best Take allows users to fabricate interactions between individuals by selecting different facial expressions for them, while Magic Editor can seamlessly replace elements in the background. This technology can easily manipulate scenarios to depict individuals looking at the same object, engaging with various items, or reacting in ways that are entirely fictional—a Pixel 8 image featuring Taylor Swift and Travis Kelce could narrate a completely fabricated story with just a few taps. Similarly, a Best Take image of Donald Trump during an upcoming election could distort reality. Google’s promotional materials for the Pixel 8 vividly illustrate the simplicity of this process:
So, the question arises: Should Pixel 8 images edited with Best Take or photos altered using Magic Editor in Google Photos be labeled on YouTube? In response to this query, YouTube spokesperson Jack Malon emphasized the significance of context and offered some preliminary guidance.
He asserts that rather than penalizing creators for utilizing AI tools, the aim of this update is to enhance viewers’ awareness of realistically altered content. For instance, if one uses Best Shot to select a photo where someone’s eyes are open, this may not necessitate disclosure as it could have been achieved through conventional means (and is not something that requires disclosure). However, if technology is employed to manipulate the background of a photo from a significant event or to create a false impression of one’s whereabouts, it is imperative to inform the audience of such modifications.
While the standard of “it could have occurred with more photos” may seem somewhat ambiguous, it aligns with the principle of transparency. Nevertheless, YouTube’s decision to mandate labeling of synthetic content created by its proprietary AI tools significantly diverges from this norm. For instance, Best Take images lack metadata indicating AI alterations, unlike images edited using Magic Editor.
In this intricate landscape, there is no definitive right or wrong; instead, there exist blurred boundaries where tool developers, content creators, and distribution platforms set their own parameters. The only certainty lies in the rapid pace at which Google’s innovative technologies will compel these entities to define and adhere to these boundaries.
