tech

YouTube brings back human moderators after AI algorithms fail to deliver

YouTube’s over reliance on algorithms led to take down of more videos than usual. The company is now reverting to human moderators.

YouTube brings back more human moderators
YouTube brings back more human moderators (HT Tech)

YouTube had put its human moderators offline earlier this year due to the Covid-19 pandemic. The company relied on machine learning-driven algorithms to manage flagging and removal of the content. The technology did not work out as the filters failed to deliver accuracy. Now, YouTube is bringing back the human moderators to oversee this.

YouTube asserted that the reliance on the AI moderation had led to a surge in video removals and incorrect takedowns, reported Financial Times. The company revealed that the video-sharing platform had taken down about 11 million videos from YouTube between April and June. This is said to be at least two-fold the rate of takedowns. YouTube admitted that about half of 320,000 takedowns were reinstated after appeals.

“One of the decisions we made [at the beginning of the pandemic] when it came to machines who couldn’t be as precise as humans, we were going to err on the side of making sure that our users were protected, even though that might have resulted in [a] slightly higher number of videos coming down,” YouTube’s chief product officer, Neal Mohan told FT [paywalled].

ALSO READ: YouTube Music is getting a major update but not everyone can access it

The admission comes at a time when the social networking companies are relying on machine learning and other latest technologies to weed out unwanted content from their platforms. Though, experts have raised concerns over their accuracy.

For YouTube, the problem is much bigger. The company has time and again come under criticism for not able to pre-screen hate and abusive content on its platform. So far, it has relied on user reports, moderators, and of course, algorithms to take down such content.

Despite not matching the expected accuracy, it is unlikely to internet companies will give up on machine learning algorithms. “Over 50 percent of those 11 million videos were removed without a single view by an actual YouTube user and over 80 percent were removed with less than 10 views. And so that’s the power of machines,” added Mohan.