I, for one, am looking forward to the rise of generative AI trained on 2014 tumblr, hallucinating Superwholock jokes where they don't belong, cosplayers dying themselves grey in a bathtub, and DashCon references where nobody expects them
To complicate matters even further, advertising content that isn’t even owned by Automattic, including ads from an old Apple Music campaign, has also reportedly made its way into the training data set.
The plans at Automattic have been so controversial internally, that a product manager has even started pulling his own photos off Tumblr to make sure they’re not used to train AI, according to 404.
Generative AI has become a big business ever since OpenAI first launched ChatGPT in late 2022 and text-prompt image creators soon followed from a number of companies.
But major publishers have complained, with some even filing lawsuits, alleging that much of the data used to train these systems was either pirated or doesn’t constitute “fair use” under existing copyright regimes.
In response to emailed questions on Tuesday, Automattic directed Gizmodo to a new post that more or less confirmed 404 Media’s reporting, while trying to sell the move to consumers as an opportunity to “give you more control over the content you’ve created.”
We also plan to take that a step further and regularly update any partners about people who newly opt-out and ask that their content be removed from past sources and future training.”
The original article contains 536 words, the summary contains 201 words. Saved 62%. I'm a bot and I'm open source!
