Runway’s AI video generator trained on thousands of scraped YouTube videos


Runway trained its AI text-to-video generator on thousands of YouTube videos and pirated films, according to a report from 404 Media. A spreadsheet of training data obtained by the outlet includes links to YouTube channels belonging to major entertainment companies, such as Netflix, Disney, Nintendo, and Rockstar Games, along with creators like MKBHD, Linus Tech Tips, and Sam Kolder.

There are also links to channels owned by news outlets like The Verge, The New Yorker, Reuters, and Wired. “The channels in that spreadsheet were a company-wide effort to find good quality videos to build the model with,” a former Runway employee tells 404 Media. “This was then used as input to a massive web crawler which downloaded all the videos from all those channels, using proxies to avoid getting blocked by Google.” 

Runway is an AI startup that has received millions in funding from Google parent company Alphabet and Nvidia. It has created impressive tools that allow users to make realistic-looking AI videos as well as ones that capture a particular animation type. Runway’s latest tool, Gen-3 Alpha, launched in June and can “create videos in any style you can imagine.” Like other AI models, Gen-3 Alpha needs to ingest a breadth of content when training.

In addition to YouTube channels, 404 Media also found that Runway’s dataset contains links to piracy sites like KissCartoon, which lets you watch anime and other animated content for free. It’s still not clear whether Runway used all of the videos in this spreadsheet to train its Gen-3 Alpha model — and we may never find out. In an interview with TechCrunch in June, Runway cofounder Anastasis Germanidis said the company uses “curated, internal datasets” to train its models, but he didn’t provide further detail.

When reached for comment, Google pointed The Verge to a statement from YouTube CEO Neal Mohan, who told Bloomberg in April that training AI on the platform’s videos is a “clear violation” of its policies. The Verge reached out to Runway with a request for comment but didn’t immediately hear back.



Source link