The Oversight Board calls Meta’s uneven AI moderation ‘incoherent and unjustifiable’


As the meta platforms filled more with AI, the company is still doing a lot to apply the policy around the managed media. The Control Council has once again criticizes the social media company with the management of these tasks, the most recent decision is that the written rules are consistently “inconsistent and unfounded”.

If this sounds familiar, this is Since last year, the Control Board used the word “unavailable” to describe the approach of methane managed media. After the Council previously went to the virus on Facebook, Meta called on the rules after a miscarried video. In response to this, Meta said The use of tags to determine the inculiform content of AI and the application of more prominent labels in the position of “high risk”. These labels, as below, note when a writing is used or edited.

Meta sample of labels when Identify some of the EU-controlled content Meta sample of labels when Identify some of the EU-controlled content

When Meta identifies part of the EU managed content, the tag sample is “high risk”. (Screenshot (meta))

This approach is still less because the Board said. “Despite the growing content of manipulated content in the formats, the implementation of the metal managed media policy is not in line with the implementation. “Inability to automatically apply a label to all cases of the same controlled media automatically, and it is unreasonable.”

A statement came to a decision on a writing Iraq to show the sound of two politicians in Kurdistan. The assumed “recorded conversation” included a discussion on the upcoming election and other “bad plans” fixture for the region. In the article, meta was reported for misinformation, but the company closed the work “without a person.” Meta is later labeled some cases of the sound clip but not first notified.

According to the Board, it is not a job claim. Meta will probably not automatically identify and apply tags to the board, automatically voice and video recordings, for “static pictures”. This can not get the same treatment that one of the same sound or video clip can lead to further mixing of the board’s records. The Supervisory Board has also criticized the meter, according to the fact that the AI-manipulated video and audio to determine the AI-manipulated video and audio.

“This meta must be a prioritize in the world’s best technology and AI companies, the widespread use of sources and meta platforms to identify and tag the volume managed and audio.” “The board is not known that this technical expertise and resources are a company to identify high-risk media to media or reliable partners in high-risk situations.”

In the recommendations of Meta, the Board, when the company’s “high risk” label is added, in the circumstances, the situation must adopt a “clear process” to tag the necessary “identical or similar content”. The board recommended that these labels should not be seen in a language that should be seen in a language with Facebook, Instagram and Subjects.

Meta did not respond to a request for comment. The company has 60 days to meet recommendations.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *