YouTube plans to evaluate generative AI capabilities, such as a comment summarization feature and a conversational tool.

YouTube has unveiled its plans to explore novel generative AI capabilities, as disclosed by the company today. As a component of the premium offering accessible to paid subscribers, YouTube enthusiasts will soon have the opportunity to test a groundbreaking conversational tool driven by AI. This tool will be designed to respond to inquiries about YouTube content and provide recommendations. Additionally, users will be able to evaluate a fresh feature dedicated to summarizing topics within video comments.

These two functionalities will be accessible through YouTube’s experimentation hub, youtube.com/new, on a voluntary basis. The conversational tool is anticipated to launch in the coming weeks, while the topic summarization feature will undergo testing with a limited group of participants who sign up for the experiment via the website.

According to YouTube, the conversational tool’s responses will be generated by extensive language models that draw insights from both YouTube and the broader web. The aim is to assist viewers in delving deeper into the content they are currently watching. For instance, participants will have the capability to inquire about the video they are presently viewing or request suggestions for related content using phrases like “recommend related content.” These interactions with the AI will occur seamlessly while the video is streaming, ensuring an uninterrupted viewing experience.

Initially, this feature will be accessible in the United States and exclusively on Android devices via a new “Ask” option on the video’s viewing page. Starting today, YouTube Premium members can opt into this trial, although it won’t be available on all YouTube videos immediately. The tool is currently in the testing phase with a select group of users but will be extended to YouTube Premium members on an opt-in basis in the coming weeks.

The comments summarization feature, on the other hand, will employ generative AI to organize and condense the discussions taking place in extensive comment sections beneath videos. This will enable viewers to catch up on the key talking points without sifting through all the comments individually. YouTube believes that this functionality could also be beneficial for content creators, offering insights into the subjects their channel’s subscribers are discussing without the need to manually review every comment. Additionally, it could serve as a source of inspiration for creators when crafting new videos based on comment themes. Furthermore, creators will have the ability to delete comments related to specific topics if they wish.

It’s important to note that the summarization will only be based on publicly published comments, excluding those under review or those containing restricted content. This experimentation, which is currently limited to a small selection of videos, will be available only in English for the time being.

YouTube has shared its commitment to continue testing these features over the upcoming weeks and months, aiming to collect user feedback on their usage and effectiveness.

Google, YouTube’s parent company, has been actively exploring AI applications across its diverse product range, spanning Search, Workspace, consumer applications such as Photos and Assistant, and more. On YouTube specifically, Google has also experimented with AI-driven advertising solutions for targeting cultural moments. Additionally, they have introduced various AI tools designed to assist content creators, including tools for finding music for videos, AI-based dubbing, and “Dream Screen,” a generative AI feature for crafting background visuals in YouTube Shorts. YouTube Music has also recently introduced AI-generated playlists. Previous AI experiments by YouTube include AI-generated video summaries and quizzes for educational content.

YouTube emphasizes the experimental nature of these features and acknowledges that they may not always meet expectations. As such, they are starting with limited availability and actively seeking user feedback, which can be provided directly through the tools themselves.