YouTube’s AI algorithm adds explicit language in captions to children's clips

If you think artificial intelligence (AI) systems are fool-proof, it is time to realise the truth.  

An AI algorithm, which is used by YouTube for automatically adding captions to clips, has been mistakenly inserting explicit language in children's videos.  

The system, which is known as ASR (Automatic Speech Transcription), has been showing words like beach as bitch, corn as porn and brave as rape, according to a report by Wired.  

Also Read: For 'censoring' domestic media, Russia partially limits access to Facebook

A team of Rochester Institute of Technology in New York, along with others, sampled around 7,000 videos from 24 top-tier children's channels to track the problem.   

Around 40 per cent of the videos sampled had 'inappropriate' words in the captions. One per cent of the videos had highly inappropriate words.   

"Well-known automatic speech recognition (ASR) systems may produce text content highly inappropriate for kids while transcribing YouTube Kids’ videos,” the team found.  

Also Read: Evidence of first ear surgery found in 5,300-year-old skull, say archaeologists

"We dub this phenomenon as inappropriate content hallucination," they added.   

It happens as the AI sometimes mishears what is being said, if it is a thick accent, or a child is speaking and doesn't pronounce properly.   

“Our analyses suggest that such hallucinations are far from occasional, and the ASR systems often produce them with high confidence," they further added.   

Children under 13 should be using YouTubeKids, where automated captions are turned off, YouTube spokesperson Jessica Gibby told Wired.  

In a statement, Gibby told Wired, “We are continually working to improve automatic captions and reduce errors.”  

(With inputs from agencies)