TikTok is introducing a updated setting that allows them to lessen the quantity of AI-made videos appearing in their feeds. This initiative comes as the company disclosed that it now features in excess of 1 bn content made with AI.
Trial and Implementation
This change is undergoing trials over the coming weeks before a full-scale launch. This advancement comes amid a rise in AI-generated material, driven by new content generation software.
AI Proliferation Online
Recent reports that almost a tenth of the quickest rising video streams on a competing service primarily display AI-generated videos. A significant portion of this output is classified as “AI slop”, a label for poor-quality, factory-style content that is typically absurd or surreal.
Based on user feedback that many people enjoy content created using AI tools, ranging from digital art to educational videos, and we aim to give people the option to adjust the amount of that, depending on their own likes,” said a platform representative.
Scale of AI Content
TikTok announced that there are presently 1.3 billion videos on its service tagged as AI-generated. In total, in excess of 100 m pieces of content are submitted to the app daily, making AI-generated material a minor portion of its overall inventory.
User Control Features
Viewers will be able to limit or boost the level of AI content they watch by accessing the topic control feature in the application and picking “AI-generated content”. Additional categories that are already be controlled under the tool include news, style, beauty, and dance.
Platform Guidelines
The platform’s guidelines require users to tag “realistic” AI-generated content, and it prohibits malicious AI impersonations of well-known individuals or critical incidents. Any realistic AI-generated clip that is not labeled can be taken down under its terms of service.
In addition, the service will now attach an AI-generated watermark to content created with its built-in AI features, or identified by an industry-wide initiative for identifying artificial media. This step is aimed to help prevent users from seeking to circumvent the labeling process.
Responsible AI Funding
The company is also launching a $2m awareness program for specialists and institutions to develop educational content about using AI responsibly. This program will support educational bodies focused on training programming and robotics.
Platform Protection
The platform has supported its content review process, following debate over proposals to cut jobs for numerous safety staff in the UK. It is planning to reduce a significant number of roles in its user protection team in its British office, with trade unions and digital protection specialists raising concerns that people are being substituted by automated systems.
A company official explained that human moderation would remain essential, but automated tools can shield workers by eliminating the most distressing content before it gets reviewed by people. The company reported a significant reduction in explicit and harmful content reviewed by safety staff over the previous year due to the use of its machine learning processes.
“We think that it’s essential to combine people and machines to ensure user safety,” the executive stated. “We’ll have staff as involved in the system, but we further consider it’s highly necessary to minimize viewing of harmful content as swiftly, which is where a lot of the machine support is really helping us process content quickly.”
