Google-owned YouTube is using artificial intelligence (AI) to categorize and derank what it deems to be “right-wing” and “news talk” videos according to a new investigative report from Project Veritas which features a Google insider.
This AI deranking reportedly started in May and the Google insider claims that meme maker Carpe Donktum, talk show host Dave Rubin, educational channel PragerU, independent journalist Tim Pool, and comedian Steven Crowder have all had their content suppressed by this AI.
According to the Google insider, YouTube staff were told during a meeting that there was going to be more content filtering on the platform. After this meeting, lots of YouTube channels were demonetized and deranked, including the five channels listed above.
The insider says that YouTube is targeting what it believes to be “right-wing news commentators,” although some of the people named in the report, including Rubin and Pool, identify as “liberal” and “left-leaning.” They claim that YouTube is targeting these commentators because it doesn’t want the opinions they express to have wide appeal and so it’s deciding which content users are allowed to see.
YouTube reportedly targets these commentators by using AI to transcribe the text of their videos and then assigning certain categories to it based on the transcribed text such as “right-winger” or “news talk.” The Google insider adds that once the text has been categorized, a machine learning (ML) “fairness” algorithm kicks in and suppresses the video across YouTube.
The insider also revealed that Rubin and PragerU were given heightened analysis in an AI program called Viacon which polices YouTube distribution and specifically classified these channels as “right-wing” and “news talk.”
The Google insider also commented on the chilling effect of this AI based suppression of YouTube content:
“What’s scary is that Google’s deciding what’s important and what’s not important. They are going through and they’re effectively deleting conversations from the national narrative. It reminds me of a book called 1984 and that should have been a warning. 1984 should not be a user manual on how to run society and Google’s falling directly into that trap where they’re deciding what gets read, what gets consumed, what people are able to click on, what appears.”
This report from Project Veritas is the latest of many leaks and insider accounts that have exposed Google’s bias against right-wing viewpoints. Recent reports have suggested that the company blacklists multiple right-leaning news sources in its search engine and that it has an internal culture that’s biased against conservatives. However, this is the first time we’ve had a glimpse into how AI is being used and trained to enforce this bias.