News
Leave a comment

AI ‘will not alter fundamental commercial reality’ in news

AI ‘will not alter fundamental commercial reality’ in news


A report from Enders Analysis has argued that generative artificial intelligence “will not alter the fundamental commercial reality for the news” as the shift online did previously.

The research firm cautioned publishers to be “realistic” about the productivity and revenue gains possible from AI, but added that ignoring AI would be “a mistake”.

The report found there have been some valuable uses for AI in the newsroom — but argued that there may not be an “immediate, killer news use case to raise revenues”.

Worthwhile use cases raised by the analysts included creating audio editions of articles and translating content into foreign languages, although they noted that “translation is not the same as localisation, so such formats won’t mean game-changing audience expansions”.

[Read more: How The Economist is using AI to extend its global reach]


AI can also help to create “more sophisticated metadata for archival material”, they wrote, in turn making it easier for journalists and readers to access a publisher’s back catalogue. This could have revenue implications for local publishers in particular, they said, “where some historical material has barely been digitised”.

‘The opportunity is real, even if… AI in itself won’t give a competitive edge’

In general however the Enders report suggested the amount of use publishers could get out of generative AI depended on their position in the market.

News agencies like PA Media or Reuters, for example, already integrate non-generative AI into their systems with products like RADAR, which reformat structured data sets into news stories.

“The marginal benefits of moving to a generative AI-based system are going to be more limited for an organisation that has already automated a large degree of newsgathering and production,” said the report.

Systems that rely on large language machines — i.e. generative AI — “often require much more oversight by journalists” because of their well-recorded tendency to invent information.

“The output of pre-LLM systems might be more rote, but they are also currently far more reliable, and that is what a news agency needs to be,” Enders said.

At local and regional publishers, on the other hand, the flexibility of generative AI can be helpful because it allows small, less well-resourced newsrooms to spend less time on “more kinds of tasks”.

The report cited as an example Newsquest, which has been hiring “AI-assisted” reporters who can rapidly re-format press releases — freeing up other reporters to do more reporting in the community — and has developed a bot “that can generate story leads through automated FOI requests and identification of newsworthy responses”.

The two main AI use cases the Enders analysts highlighted at national newsbrands were data journalism and story ideation. But they cautioned that given the wide availability of generative AI tools “what will give an organisation an original edge here may be acquiring quality datasets to investigate”.

The Financial Times, they said, sets an example here by using LLMs “to classify information in large datasets, but then use traditional rules based AI to perform analysis on that data once it is classified—this makes hallucination less likely to creep in”.

For now, Enders saw limited room for generative AI in video news content creation, reasoning that “any sensible efficiencies are in editing, not generation, which has a long way to go before it is cost-effective or reliable”.

“Publishers must be realistic about the scale of efficiencies and revenue generation opportunities, and size investments accordingly,” the report said in its summary.

“The opportunity is real, even if broad, democratised access to tools means that AI in itself won’t give a competitive edge.”

The investment necessary may be mitigated through “some pooling of resources”, the analysts recommended.

“The costs involved in some of this development could be shared if UK publishers see this as a way of collectively combating challenges to the industry as a whole.”

Are AI chatbots for news publishers worth it?

Chatbots trained on publisher content have been the main consumer-facing generative AI product news publishers have experimented with. Enders described this as a bid “to be the destination, the recognisable source and brand of the content they supply” amid ongoing attempts by big tech to build link-free, generative AI-powered search engines.

The firm said that “concrete experimentation” had been slower with chatbots than back-end tinkering, “in part due to concerns about reader reception of ‘AI’ things. 

“Some UK publishers feel there is little early mover advantage to the reader-facing side, meaning that there has been a bit of stalemate.”

Consumer demand for such chatbots “is so far unclear and the bar for rollout is high”, the report said, but they could lay the foundations for “other more effective interfaces, like improved search/website functionality—areas where some online news providers have been stalling…

“Improvements here are more modest but effective: like the FT’s new feature allowing readers to highlight text and find articles semantically related to it, moving beyond offering only high-level related articles.”

Email [email protected] to point out mistakes, provide story tips or send in a letter for publication on our “Letters Page” blog



Source link