Innovation Bias: The Chinese AI News Anchor

Analysis

This last week The Verge published some news about an AI news anchor launched by the state-run press agency. The tech used by the ‘digital anchors’ is not clear, however, it seems the agency, Xinhua, used a form of machine learning programmed by real footage to create facial movements to correspond with synthesised Chinese speech.

The article purveys a chilling sense that this is perhaps one more step towards an Orwellian dystopia where fake news is indistinguishable from the truth. This response certainly wouldn’t be unfounded given the nature of the origin of the experiment; a state-run Chinese news agency.

Xinhua AI news anchor.

However, if we put our bais to the side for one moment and view this purely as an innovative development, perhaps this is a step towards elegant communication of information between human and computer. Automation gets a bad rep as it displaces workers, but in this instance look some of the advantages.

The article suggests that two anchors, one Chinese and the other English, would be used to generate news reports for TV, web, and mobile output. Instantly this tool will increase the reach and engagement of this news agency’s output by offering content in multiple languages across various formats, instantly and economically. Furthermore, this reformatting will allow Xinhua to reach and engage their audience through their modern media habits. The agency as simple innovated to create a format that better serves their audience.

This is a natural evolution as news companies continue to tailor their products around the changing habits of the news consumer. This trend dates back decades even before newspapers went from broadsheets to tabloids for easier reading ‘on-the-go’. It just so happens that this innovation comes from a state-run Chinese news agency. But should that be so surprising as China continues to become a leading force for innovation?

At SpeechKit we shared the bias that computer-generated news content would lend itself to dystopian visions. However, once we started to test this technology and experienced its potential to reformat content to be delivered through our headphones, we instantly were convinced of it’s potential to help inform a generation addicted to audio.

Furthermore, once we launched an MVP, we found that this increased convenience was favoured by early adopters. Readers converted to audio and were retained whilst dwell time improved dramatically where automated audio was served. For a format in its infancy, early customers have been impressed and audio impressions across SpeechKit have increased 23% MoM over the last 6 months.

Thanks for sharing your time with us. We’d love to hear how we can improve our service to meet your audio publishing needs. Should you have any issues then feel free to contact us at [email protected].