How artificial intelligence is set to impact media
Photo: Twitter/Wibbitz
But they aren’t alone in their harnessing of the technology. Traditional publishers have started to use AI to create stories that are largely delivered without human intervention. And it does not have to cost them a fortune to do.
So how is AI being used now, and what impact is it likely to have on publishing in the future? At the recent FIPP World Congress, a panel comprised of Francesco Marconi, manager strategy and development, Associated Press, Martha Stone, CEO World Newsmedia Network and Alice Zimmermann, global product partnerships, Google shared some thoughts on developments. They were interviewed by Mike Hewitt, Congress moderator. Here is a condensed transcription of the conversation.
Mike Hewitt – How is AI affecting the media right now, and where is it going?
Francesco Marconi – My perspective on AI is centred around the impact it has on content creation. At AP we realised we had issues around volume and differentiation as we sought to sell our content. We are using AI to address those two opportunities. We look at AI for automation, thereby addressing the volume issue. We also ask how can we use AI to enable the human creativity that occurs in the newsroom to help us with differentiation.
Automation is central to what we do. We have lots of different projects turning data into text and text into video. The idea is not to displace content creators, but rather free resources so they can focus on high level creative work.
With automated content we can open new markets and better serve our client publishers with new forms of content. These are often in areas that we previously found difficult to scale. Automation of earnings reports, for example. We went from doing 300 companies to over 4000 stories, and were able to free up 20% of journalists’ time.
Martha Stone – I have run a big data and AI conference for five years. We focus on leveraging big data and what companies can do with it. Now we can leverage AI to target customers and create new products on the fly. What we do is look at the business side, and ask the question ‘how do we make money and how do we create products?’
A lot of media companies, like Google, are taking bold steps. Though from the traditional media standpoint most are still at the beginning stages. The first thing they need to do is create a data lake, where all the data goes into one technical receptacle where it can be accessed from one place. This takes time. For example, it took The Guardian about a year to put all their data sets into one.
Usage data, about what people are using when reading, where they go and what time they spend etc is the starting point. This can then be used to create profiles of users to enable publishers to deliver information that is relevant to them. For example the BBC have created MyBBC. This is a receptacle of an individual’s data which is both implicitly and explicitly collected. In other words the individual makes choices to personalise the content but they are tracked too. They then serve recomendations on what is of interest to make the service more relevant. They can then use that data to create products that are most relevant for individuals.
Francesco Marconi – I agree the personalisation of content is another important use of AI. If you can track how consumers behave online and communicate with each other then you can personalise their content – choosing tone, locality etc and this will generate a lot of engagement.
Alice Zimmermann – We use AI to help technology to enhance media and make our professional lives and jobs easier. An example is Project Jigsaw which uses machine learning in different ways to protect freedom of speech online. We need to be able to participate in media in discussion boards etc without being subjected trolling and online abuse. We now have AI that can gauge the toxic impact the comments could have on a discussion, and thereby encourage a healthier culture in our media.
AI is also already being used in conversational user interfaces as users engage with online content. This creates opportunities to deliver content in a more natural and useful way. There are tools that publishers can start experimenting with.
Francesca Marconi – There are many good examples of projects with automated insights. We take structured data (eg sports and financial data) and then develop the templates which include specific sentences. This style of thinking requires logic and is closely aligned to how programmers think rather than how journalists think. The AI matches the data with the template to generate a story that is readable by humans. This is automation with little human intervention.
Martha Stone – We need to ask how can we make money from this? This strategy requires large amounts of investment, so to offset this you need to make revenue from it. Media companies need to operate like Facebook and Google. I think their huge success is because in large parts they are driven by data and by AI. In China there’s a huge presence in AI. You need to start watching what they are doing. There’s lots of investment to compete with Google etc in the AI game. They want to win the AI game. If you are not driving revenue through data via AI you need to start now.
Francesco Marconi – Although Google, Microsoft and Amazon are taking the lion’s share of advertising, they are also lowering the cost of entry via APIs and cloud based services. This enables companies to experiment with AI. The revenue question is an important one, one to keep in mind.
Alice Zimmermann – AI should be open to everyone. We have dozens of grants for researchers to build AI outside of Google. Secondly, as a platform we want to enable enterprise tools for anyone to use. You don’t have to be a big company. Anyone can use our tools to build AI. There are already some great stories. There’s a teenager in high school that used our tool in conjunction with a lab to help mammogram detection in breast cancer.
Francesca Marconi – We are working to turn text into video. It works very well for profile of public figures, or the creation of financial data. The approach here with video automation is similar to text automation. Journalists develop the templates and then the technology creates the content by filling in the gaps.
We also using augmentation to discover hidden insights. We used AI to analyse tweeted replies to Trump’s social media posts. By analysing the public discourse we were able to identify some key terms. We could see partisan bias when people are replying to tweets.
Mike Hewitt – What advice would you give to media companies starting to consider using AI?
Alice Zimmermann – Talent resourcing. If you can train yourself to learn about AI and what is available, do it. If not find people that have the expertise which might be enhanced by machine learning today. Invest in training and experimentation.
Martha Stone – Get a grounding in what AI is, and then start formulating strategies and create a team.
Francesco Marconi – Double down on collaboration and experimentation. It is surprising how cheap it is to test new tools. It is a gateway to being exposed to new thinking and lower the barrier of testing the new technologies.
More like this
How reporting the truth can transcend turbulent political times
The Business of Fashion: from blog to international media business
How Hearst UK repackages content across platforms
