YouTube is offering on a small scale the ability to use the AI assistant Gemini as an inspiration and sounding board. This feature should kickstart YouTubers’ creative process even when they momentarily lack inspiration.
According to a recent video on YouTube’s Creator Insider channel, the video platform is experimenting with integrating Google Gemini into its backend environment. The Brainstorm with Gemini tool should help video makers brainstorm ideas, come up with titles and create thumbnails, among other things.
Juggling multiple AI assistents
The new tool bears similarities with another AI-based tool that the tech giant launched for YouTube back in May this year, Techcrunch writes. This AI-based content inspiration tool presents creators with topics for videos that might interest their viewers. The tool then provides an outline of content points that they can use to start the creation process.
Users can ultimately decide which AI-supported tool they want to use for their content inspiration. In YouTube Studio, when they type a video idea into the search field, they are presented with the option to choose between the two tools.
Cutting off ChatGPT
The new Brainstorm with Gemini option for YouTube creators is currently only available to a small group of selected content creators. Based on feedback from these end users, YouTube will determine whether it will make this feature generally available.
The arrival of GenAI to YouTube makes sense, for creators, too. It allows the social media and video platform to capture more market share from its already meagre competition that does not offer this functionality. In addition, this diminishes the chance that YouTube creators will not use other GenAI tools, such as ChatGPT, to prepare and set up their video content.
Inspiring Youtubers with YouTube
Where does Gemini get its inspiration from? Probably everywhere, but also YouTube itself. The New York Times reported earlier that owner Alphabet/Google uses the video platform to train its own models. Google is even said to have recently stretched the terms of use of several services, allowing it access to public material for training its own LLMs.
These include public documents in Google Docs, reviews of restaurants on Google Maps and, of course, YouTube videos.
The scraping of such videos for training LLMs shows that the major AI providers are looking for new training data and are getting increasingly “creative” in that regard. For example, Meta considered acquiring major U.S. publisher Simon & Schuster to obtain data from its portfolio, and the company supposedly collects copyrighted data from all over the Internet, even though this could lead to lawsuits.
OpenAI trained GPT-4 on million of hours of audio from YouTube videos. The AI giant did not ask Google’s permission to do so. The latter did not object however, because it also uses YouTube to train its own LLMs.
Read more: OpenAI trained GPT-4 on millions of hours of YouTube audio