Facebook’s content classifier has been malfunctioning for months

Fake news and other content that could be harmful in many ways could be added to Facebook users’ feeds for many months due to a malfunction of the algorithm responsible for the system’s operation, writes The Verge referring to an internal record obtained from the company. The complexity of the system is well illustrated by the fact that the engineers did not find the final cure for the anomaly discovered last October until mid-March this year.

According to the note, it became apparent to system operators last October that the “viewership” of posts that were clearly classified as sham had suddenly increased significantly by 30% in the system, and that other types of content that the algorithm should have automatically downgraded were not included. transfer to this category.

Thus, during the period, the system misidentified and classified a number of content depicting nudity or violence, and even content from Russian state media that was clearly considered undesirable, which could be particularly worrying in the light of the Ukrainian-Russian conflict.

Perhaps even more worrying is that although Facebook rated the bug as the most critical in its internal systems, the engineers operating the system were only able to fix it permanently on March 11, as the record simply didn’t find the cause of the bug for months.

The Meta-group has received a lot of attacks and criticisms in recent times about the operation of the algorithms of the social services it operates. The company brought back the chronological display of posts to Instagram just a few weeks ago under pressure from external pressure, and it could easily have to limit the operation of algorithms that fundamentally determine the effects and access to content on social networks in the future. States.


Source: HWSW Informatikai Hírmagazin by www.hwsw.hu.

*The article has been translated based on the content of HWSW Informatikai Hírmagazin by www.hwsw.hu. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!

*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.

*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!