Twitch will start using streamer content to train Amazon AI; see how to disable it

Twitch has implemented a change that allows the use of content generated by its users for training Amazon's generative artificial intelligence models. This feature is active by default, requiring users who do not wish to participate to manually disable it in their account settings.

The news gained prominence on Wednesday, the 12th, after content creator and streaming journalist Zach Bussey published a screenshot demonstrating the new option. It is important to note that this policy covers not only live broadcasts but practically all material from a channel, including VODs, clips, chat messages, images, and text.

The new setting authorizes the use of this content to 'train generative AI content models on Amazon.' Twitch itself cited the use of broadcast audio to improve speech-to-text conversion systems as an example, which could benefit the platform's automatic captions or be applied to other Amazon services.

The scope of the training goes beyond audio; the company clarifies that streams, saved videos, clips, chats, images, and text associated with the channel can be used. There is a specific caveat regarding chat: even if a user disables training on their account, messages sent in another person's channel with the option enabled may still contribute to the training data.

One of the main points of controversy raised by users is the adoption of an opt-out system. This implies that creators are automatically included in the training process and must take the initiative to access settings to express their refusal.

During a dedicated stream on the topic, Twitch Product Director Mike Minton justified the choice of this operational model. According to the executive, if participation depended on prior authorization (opt-in), adoption would be very low. He stated: 'If it were opt-in, nobody would opt-in,' thus explaining the maintenance of the default active option.

For those who wish to prevent their channel content from being used in training Amazon's generative models, the procedure is to disable the option in the account security settings, following the path: Twitch > Settings > Security > Generative AI Training. However, disabling this function does not mean losing all of Twitch's artificial intelligence features; tools such as chat moderation and ad selection will continue to operate normally.

Mary Kish, Twitch Community Head, commented during the stream that certain AI applications are already part of the platform's operation. She expressed skepticism about the possibility of Twitch operating without these technologies and confirmed the current use of AI in various areas of the service. Interestingly, Kish revealed that she personally decided to disable the use of her content for training purposes.

In contrast, Minton stated that if he were a streamer, he would keep the training option enabled. Another issue raised by users was the way the change was introduced, given that the option appeared in the settings without comprehensive prior communication to creators. Executives demonstrated where to find the feature only after receiving user complaints about it being hidden in the settings. Kish acknowledged the criticism regarding the lack of email notification but argued that creators do not always read their mail, suggesting that direct messages and community outreach could help disseminate the information.

The company also guaranteed that, for now, it does not plan to create a public identification indicating which channels have consented to or refused AI training. This decision places Twitch in a broader debate about which entity has the right to use user-generated content to develop AI models. In the context of Twitch, the debate gains complexity due to the vast quantity and diversity of information contained in a single broadcast, which can aggregate hours of video and voice, chat interactions, images, and text produced by both the streamer and the community. For now, users retain the ability to block the use of their own content in Amazon's training, although the default implementation has already generated criticism within the community.

olhardigital.com.brAugust 12, 21:10

Facebook launches Creator Studio with AI to help creators analyze their audience

Facebook has launched its new autonomous application, called Creator Studio, which integrates various artificial intelligence functionalities. This feature aims to support content creators in monitoring the performance of their posts and interacting with their followers.

Currently, the version intended for the iOS operating system is available in the United States and Canada. The core of this application is a sophisticated AI assistant that performs detailed analyses on posting style, audience engagement level, overall performance, and user-defined goals.

Based on this collected information, the assistant provides suggestions and can answer questions related to the content creation routine. Creator Studio seeks to simplify a task that traditionally required producers to navigate multiple charts and metrics dashboards.

Instead of consulting several indicators separately, the creator can simply ask the system, for example, what the ideal time to post is or what the public is commenting on. Furthermore, subsequent questions can be asked based on the answers received.

This launch occurs in a fiercely competitive landscape for creator attention, where platforms such as Facebook, TikTok, and YouTube compete for content generated by this audience. The application also includes a feature that allows selecting the most relevant comments and generating draft responses that mimic the creator's characteristic communication style.

It is important to note that no publication is done automatically; the user maintains full control, being able to review, alter, and approve every suggestion before sending. By accessing Creator Studio daily, the creator also receives a list of priorities.

More information:

Creator Studio is not the only recent launch from Meta. In recent months, the company has introduced other independent applications, such as Forum, which is inspired by the operational model of Reddit and focuses on Facebook groups. Instants have also been made available for sharing ephemeral photos with Instagram friends; Pocket, a game application developed with natural language programming; and StoryKit, which uses artificial intelligence to create children's stories.

After undergoing testing with a select group of creators weeks earlier, Creator Studio was released to all iOS creators in the United States and Canada. This initiative consolidates performance analysis, follower interaction, and AI features in one place, reinforcing Facebook's effort to keep creators active and producing content on the platform.

tecnoblog.netJuly 30, 21:13

LinkedIn implements button to report AI-generated publications

LinkedIn has started testing a new feature that allows users to report publications that appear to have been created by artificial intelligence. This functionality is available in the post reporting menu and uses the description «Seems Like AI Slop» (which can be freely translated as 'seems like AI content').

How the new feature works

According to the portal 404 Media, this mechanism does not result in the automatic removal of publications, nor does it confirm whether the text was generated by an AI tool. Its function is to serve as an indicator to flag repetitive content.

The term «AI slop» gained notoriety to characterize the large amount of mass-produced material by AI, but with little originality, being compared to 'filler content'. On LinkedIn, this type of posting has become a subject of jokes, given that the platform has been receiving a volume of generic motivational texts, exaggerated professional narratives, and posts made solely to increase engagement.

Platform impact and measures

A survey conducted by Panram indicated that 41% of articles published on the social network have already been produced with the help of AI. Previously, LinkedIn had already announced its plans to restrict the reach of content classified as repetitive or low quality, announcing changes to the algorithm in May.

Currently, there is still no clarity on how the platform plans to use the signals provided by this new button or if they will have a direct influence on the recommendation system. LinkedIn has not specified whether the feature will be launched globally or if it is part of a restricted test.

Actions by other platforms

Other social networks are also looking for ways to manage the excess of AI-generated content. YouTube, for example, promoted updates to its monetization guidelines to limit materials considered repetitive or mass-produced, in addition to removing channels that violate its policies.