AI learned to remove half offensive tweets without moderators

Twitter co-founder Jack Dorsey said that now half of all offensive tweets on the platform are automatically deleted using AI, and users don’t even have to complain about the publication. He associated this event with the company's success in machine learning.

Dorsey added that because of the growing number of users, they could not rely solely on complaints from users and live moderators. Therefore, to automate most of the process, the company resorted to AI.

“We continue to make progress in moderation by improving our ability to actively identify and remove offensive content,” said Dorsey. “At the same time, more than 50% of tweets in the last year were deleted by AI - it was not helped by company employees or our moderators.”

According to the company, the rate of automatically deleted tweets for the year increased by 43%. In their opinion, this is due to the “improvement of machine learning models” to identify potential violations of platform rules.

“We will continue to actively work to reduce the level of offensive publications on Twitter, users will be less and less forced to manually complain about content that they think is inappropriate. We are confident that AI will participate more and more actively in this process, ”the letter says to Dorsey’s shareholders.

hightech.fm