In yet another unpleasant move in our AI hellscape, Amazon was caught training its generative AI using Twitch stream content earlier this month—a particularly brazen strategy, especially given Twitch's chief product officer admitted that the data collection was opt-out because "nobody would opt in." Hmm, I wonder why!

Anyway, despite attempts to insist that Twitch wasn't doing this, Amazon was—a distinction which pretty much doesn't matter, as Amazon owns Twitch, and even if it didn't, Twitch still would have been permitting this—the streaming platform's caught a ton of flak for trying to slide this one under the table.

Including, per Courthouse News, a class-action lawsuit. Plaintiff Warren Pandiscia, who has around 1,000 followers on the site, claims that "because Amazon AI products are commercialized, Amazon had an overwhelming incentive to acquire training data on an unprecedented scale.

"Rather than negotiate for lawful licenses or seek permission, defendants accessed the Twitch streams and videos to utilize them as a massive dataset necessary to fuel Amazon's AI products."

The full complaint, filed in California, goes on to state: "Defendants' actions were not only unlawful, but an unconscionable attack on the community of content creators whose content is used to fuel the multitrillion-dollar generative AI industry without any compensation."

The complaint also goes on to note that, per Twitch's documentation, a streamer chatting on someone else's stream might have their voice used to train generative AI despite opting out themselves, as whether or not Amazon can scrape your voice is entirely dependent on the host streamer's settings.

Source link