Japan’s NHK is putting an AI-powered cartoon anime on air to read you the news. 💁🏻‍♀️

This is the national broadcaster's contribution to automated journalism.

By Tim Hornyak
Splice Japan

If you tune into News Check Eleven on Japanese public broadcaster NHK TV, there’s a new feature that might give you pause. Leaping on screen to join hosts Minoru Aoi and Kaori Nagao is an animated, computer-generated character with cute looks and a nose for the latest online trends. “Yomiko”, billed as an artificially intelligent announcer, is Japan’s latest experiment in the automation of journalism.

The broadcaster, formally known as Japan Broadcasting Corporation, rolled out the AI cartoon earlier this month. Wearing a blue blazer, white skirt and earrings in the shape of an “A” and “I,” Yomiko appears to be about 30 or 40 cm tall when “standing” on a desk used by her human counterparts. In a somewhat robotic monotone, she introduces topics such as how school uniforms are being recycled and whether body odor can amount to “smell harassment” before responding to questions from Aoi and Nagao.  

“Yomiko is amazing! She’s really having a conversation!” viewer @toshitaka_szk wrote in a Twitter post that was picked up by NHK and reproduced on screen during Yomiko’s debut.

Yomiko is a virtual news announcer on NHK, Japan's public broadcaster, that uses artificial intelligence. (Image supplied)
Yomiko is a virtual news announcer on NHK, Japan's public broadcaster, that uses artificial intelligence. (Image supplied)

News meets anime

“Yomiko is a new way to deliver the news centered on the concept of ‘viewer participation,’” NHK says in a statement to Splice Newsroom. “We would like to position it as an AI announcer that will be raised along with viewers and improve its news reading and conversation skills. We hope viewers will enjoy this process.”

Yomiko was developed at NHK Science & Technology Research Laboratories, which is known for creating cutting-edge broadcasting technologies such as 4K and 8K. It’s based on software that the lab made to read news from the PyeongChang 2018 Winter Olympics earlier this year. The system uses deep learning techniques from AI and has a vocabulary of some 100,000 words. Yomiko can also read the latest bulletins through Google Home and Amazon Alexa. Viewers can also use an augmented reality app to take smartphone photos featuring Yomiko.

While Yomiko is only a one-year experiment, she has all the ingredients for success, at least with Japanese viewers. The size of a doll, with an earnest, pleasant disposition, Yomiko is far being from a threatening robotic presence.

She’s designed to look like a fresh-faced cub reporter and her name is a tongue-in-cheek play on the Japanese verb yomu, to read.

But her biggest asset is probably the fact that she was designed by manga artist Katsuki Tanaka, creator of the hit capsule toy Fuchiko on the Cup. Marketed by Kitan Club, Fuchiko is a small plastic figurine dressed like an office lady, or Japanese female clerical worker, that can be posed on the edge of cups and other objects; launched in 2012, Fuchiko has notched up sales of over 20 million units in over 1,500 variations. There’s no small resemblance between Fuchiko and Yomiko. Though NHK denies the toy was used as a model, its Yomiko page shows her sitting coquettishly on a virtual rock, holding a cup.

Appearing on News Check Eleven, Yomiko is described as
Appearing on News Check Eleven, Yomiko is described as "evolving every day with AI speech". (Image supplied)

NHK wants to enter the 21st century by moving beyond television.

So it's trying out the internet.

Read

A virtual trend

Virtual news announcers have been around since 2007 and virtual video personalities are nothing new in Japan. Kizuna AI, billed as the world’s first virtual YouTube vlogger, looks like an anime schoolgirl and has a YouTube following of over 1.5 million and was recently selected as a Japan National Tourism Organization ambassador.

But Yomiko is part of an international trend toward news automation, as seen in regional newspaper Shinano Mainichi Shimbun’s use of AI to write news summaries. For the moment, though, it’s unclear whether Japan will see more AI talking heads or if they will  threaten the jobs of human counterparts.

Asked if that could happen, NHK merely says, “We’re still in the development stage regarding Yomiko’s news reading abilities, and we would like viewers to use Yomiko for TV news, news websites, apps, smart speakers and other means.”

Tim Hornyak

Tim Hornyak is a freelance journalist based in Tokyo. He is the author of Loving the Machine: The Art and Science of Japanese Robots. Follow Tim Hornyak on Twitter.

From this week

Platforms

Columbia Journalism Review takes a hard look at the journalism funding done by Facebook and Google.

There are millions of dollars going into this space. While many are happy to take the money on the table, others question the ethics behind it. “The British Empire wanted trains in Kenya and India to run well, too. So their concerns are sincere, but the effect is more often than not a deeper immersion in and dependence on these platforms.” Of course this isn’t an issue unique to the tech giants — grant-giving NGOs have also faced similar critics.
Columbia Journalism Review

Governments & policy

Tools

Trends

New Naratif put together a solid story on how the Muslim Cyber Army works in Indonesia.

If you haven’t heard of the MCA (no, not that MCA in Malaysia!), they have been spreading fake news and driving hate speech along religious and ethnic lines. Worrying trend, especially in a country that’s been fighting fake news factories like Saracen. What makes this one different? “MCA looks more ideological, has thousands of networks in different parts of Indonesia and therefore the destructive power of this group is greater than that of Saracen.”
New Naratif

Transformations

The New York Times is partnering with FX and Hulu on a weekly documentary series called…The Weekly.

It centres around stories from the Times and the journalists that work them. This comes hot on the heels of The Daily, their incredible podcast about one daily story from the Times newsroom. This is part of the Times’ ongoing foray into entertainment: A New York Times Magazine feature is going to be a Netflix documentary series, and Brad Pitt bought the movie rights to the story of how the Times broke the Harvey Weinstein story. Also coming: a four-part series for Showtime about the Times newsroom during the first year of the Trump administration.
New York Times

Media startups

Talent

Design

SilverKris, Singapore Airlines’ in-flight magazine, nailed it with their recent redesign by Ink.

I usually have the same attitude to in-flight magazines as I do to, say, a swift slap across the face: I’d really just rather not, thanks. But this reworked version was good enough for me to forget my Economy Class kneelessness, even though the cover is easily the most forgettable part of the whole redesign: a crowded image with no focal point. But here’s why I love this redesign: 1. The layout and typography have integrity in that they are led by the content. 2. The section fronts have bold, opinionated design. 3. The reading experience is immaculate — even though they crowd little surprise nuggets in the gutter. 4. The illustrations by Stuart Patience are delicious. 5. The writing isn’t all travel-fluff and doesn’t suck. 6. Those are some mad infographics skillz. Here's an interview with the Ink creative director.
The Design Air

The Malay Mail did a website redesign.

Load times were a priority, and the new site scores well on that front. The digital team also prioritised monetizing content and enhancing their “programmatic setup”. For me, this is translating into lots of badly-placed ads for pointless leather accessories in duplicate and Outbrain-forward sewage. They are testing a new section with Mandarin content for Malaysians working in Singapore, which says good things about their user research. Structurally, the website is fine, although better hierarchy on the home and story pages would be a good idea. (Also, those Open Sans headlines need some kerning; they’re w a y t o o l o o s e.) I’m impressed with how their head of digital responded to a question about the cost of the revamp: he said the company saw it as an investment rather than an expense. Respect.
Marketing Interactive

The article page is arguably the most vital page for a news website.

Getting it right across platforms is the Holy Grail. Last week, The New York Times took a giant step towards getting it right. This involved streamlining internal efficiencies on their CMS as well as a better user experience across mobile and desktop on web and native apps. Advertising also got a major overhaul: they killed their cluttered right rail of smaller banner ads for larger, full-width, midstream ads for a much cleaner read—and it’s working: “Ads on the new page are achieving twice the click-through rate of our old design, and initial studies show higher brand recall and four-times the reader attention to ads.” Read about the process here.
New York Times

“Hi, so did you hear that crazy phone call that, umm, the Google Duplex robot assistant made to the hair salon?”

She had the whole uptalk (ending verbal statements with that millennialesque question mark subconsciously designed to maximise responsiveness) thing going on? as well as an “mmm-hmmm?” and even an “er”? It wasn’t just how real it was that blew my mind; it was that the person on the other end of the phone was able to have a complete conversation without suspecting anything. I think the tech is amazing; I think the the whole construct is creepy. Would Turing give the bot full marks? Hear it for yourself.
The Guardian

Google’s Duplex bot will now identify itself as a robot on the phone.

There were some serious concerns that Google was putting the ‘dupe’ in Duplex: “Silicon Valley is ethically lost, rudderless and has not learned a thing”. The company has clarified: “It’s important to us that users and businesses have a good experience with this service, and transparency is a key part of that.” What is it going to say, though: “Hey there, I’m Rishad’s bot assistant, so don’t be freaked out.”?
Twitter

Notables