YouTube’s algorithm accidentally blocks CHESS’s “black vs. white” strategy

YouTube’s overly anxious AI may have misinterpreted a conversation about chess as a racist language.

Last summer, a YouTuber that produces popular chess videos saw its channel blocked for including what the site called ‘harmful and dangerous’ content.

YouTube did not explain why it blocked Croatian chess player Antonio Radic, also known as ‘Agadmator’, but the service was restored 24 hours later.

Carnegie Mellon computer scientists suspect that Radic’s discussion of “black against white” with a great master accidentally triggered YouTube’s AI filters.

Running simulations with software trained to detect hate speech, they found that more than 80 percent of chess videos flagged for hate speech did not, but included terms such as ‘black’, ‘white’, ‘attack’ and ‘threat’ .

The researchers suggest that social media platforms incorporate the language of chess into their algorithms to avoid further confusion.

Scroll down to see the video

The popular YouTuber Antonio Radic of chess had its channel blocked last summer for content

The popular YouTuber Antonio Radic of chess had its channel blocked last summer for “harmful and dangerous” content. He believes the platform’s AI mistakenly signaled him for discussing ‘black against white’ in a chess conversation

With over a million subscribers, Agadmator is considered the most popular vertical chess on YouTube.

But on June 28, the Radic channel was blocked after he posted a segment with Grandmaster Hikaru Nakamura, a five-time champion and the youngest American to win the title of Grand Master.

YouTube did not give him a reason to block the channel.

In addition to human moderators, YouTube uses AI algorithms to discover prohibited content, but if they are not fed the right examples to provide context, these algorithms can flag benign videos.

Researchers at Carnegie Mellon have tested two cutting-edge speech classifiers, AI software that can be trained to detect hate speech.  Over 80 percent of the comments flagged by the programs did not contain any racist language, but included chess terms like 'black', 'white', 'attack' and 'threat'

Carnegie Mellon researchers tested two cutting-edge speech classifiers, AI software that can be trained to detect hate speech. Over 80 percent of the comments flagged by the programs did not contain any racist language, but included chess terms like ‘black’, ‘white’, ‘attack’ and ‘threat’

Radic’s channel was reinstated after 24 hours, leading him to speculate that his use of the phrase ‘black against white’ in Nakamura was to blame.

At the time, he was talking about the two opposite sides in a game of chess.

Ashiqur R. KhudaBukhsh, a computer scientist at the Carnegie Melon’s Language Technologies Institute, suspected that Radic was right.

“We don’t know what tools YouTube uses, but if they rely on artificial intelligence to detect racist language, that type of accident can happen,” said KhudaBukhsh.

To test his theory, KhudaBukhsh and fellow researcher Rupak Sarkar tested two cutting-edge speech classifiers, AI software that can be trained to detect hate speech.

The Radic channel was blocked for 24 hours after he posted this video, featuring a conversation with Grandmaster Hikaru Nakamura

The Radic channel was blocked for 24 hours after he posted this video, featuring a conversation with Grandmaster Hikaru Nakamura

Using the software in more than 680,000 comments taken from five popular YouTube chess channels, they found that 82% of the comments flagged in a sample set did not include any racist language or obvious hate speech.

Words like ‘black’, ‘white’, ‘attack’ and ‘threat’ seemed to have triggered the filters, KhudaBukhsh and Sarkar said in a presentation this month at the annual Association for the Advancement of AI conference.

The accuracy of the software depends on the examples provided, said KhudaBukhsh, and the training data sets for YouTube classifiers ‘probably include some examples of chess talk, leading to incorrect classifications’.

Radić, 33, started his YouTube channel in 2017 and has more than one million subscribers.  His most popular video, a review of a 1962 match, had over 5.5 million views

Radić, 33, started his YouTube channel in 2017 and has more than one million subscribers. His most popular video, a review of a 1962 match, had over 5.5 million views

If someone as well-known as Radic is being wrongly blocked, he added, ‘it may very well be happening quietly with many other people who are not so well known’.

YouTube declined to say what caused the Radic video to be flagged, but told Mail Online: ‘When we are told that a video has been removed in error, we act quickly to restore it.’

“We also offer users who carry the ability to appeal removals and we will review the content again,” said a representative. ‘Agadmator called for the removal, and we quickly reinstated the video.’

Radić, 33, started his YouTube channel in 2017 and, within a year, his revenue surpassed his daily job as a wedding videographer.

‘I always loved chess, but I live in a small town and there weren’t many people I could talk to [it]’he told ESPN last year. ‘So, starting a YouTube channel kind of makes sense.’

His most popular video, a review of a 1962 match between Rashid Nezhmetdinov and Oleg Chernikov, has had over 5.5 million views to date.

The COVID blockages have generated renewed interest in chess: since March 2020, the server and social network Chess.com has added about 2 million new members per month since the pandemic began, Annenberg Media reported.

The game of kings has also benefited from the popularity of ‘The Queen’s Gambit’, an acclaimed miniseries about a troubled chess master that fell on Netflix in October.

.Source