World Association of Newspapers and News Publishers


Publishers Should Understand Business Needs in Adopting AI Tech

World News Publishing Focus

World News Publishing Focus
Your Guide to the Changing Media Landscape

Publishers Should Understand Business Needs in Adopting AI Tech

Jensen Boey of Singapore Press Holdings debunks myths associated with the integration of artificial intelligence technologies and newsrooms.

By Chad de Guzman

 

Publishers must understand their business needs before fully utilizing artificial intelligence technology (AI), Jensen Boey said during the Digital Media Asia Conference 2019, held in Hong Kong in October.

“It’s important for the infrastructure, the raw computing power that’s required to support AI models, and equally, if not more important, is the training required for adoption by the newsroom of all these AI technologies. It’s no use if they don’t understand it,” Boey, who works for Singapore Press Holdings, said. 

The Head of Programme Management Office and Value Engineering of SPH Tech said their newsrooms have partnered with Fake News Guard, a platform which links news items up for verification against published articles on The Straits Times website.

The platform, however, does not tag the news as fake and leaves the implications of its recommendations up to the user. 

Boey said publishers must consider three factors when adopting AI technologies: the latency of the response, the number of false positives or false negatives which their publication deems acceptable, and the ability of the AI model to provide an explanation of its output. 

He also debunked several myths revolving around AI, including basing a project’s strengths on its algorithms and models.

“You need to be able to download the data, manage the data, secure your data. Otherwise you are unable to operationalize your AI project, which may not really achieve your business outcomes. As the saying goes, garbage in, garbage out - data is key,” he said.

He also subverted notions of Deep Neural Networks (deep learning) being the best form of AI.

“There are many other AI solutions that provide a much better or more favorable cost-benefit ratio,” Boey said.

Boey also said that contrary to popular belief, AI is biased, since its architecture is still based on human input.

“The reality is that AI products are based on the data, the rules, the inputs from human experts, that is intrinsically biased,” he said.

 

Author Chad de Guzman is a Filipino student journalist from Hong Kong Baptist University, and a scholar of the university’s Master of Arts in International Journalism Studies program.

Author

WAN-IFRA External Contributor

Date

2019-11-21 09:19

Author information


© 2020 WAN-IFRA - World Association of News Publishers

Footer Navigation