Will AI be the demise of journalism or a blessing in disguise?

On November 30^th 2022, ChatGPT was released to the world resulting in a frenzy across news outlets and social media as many feared the dusk of humanity had arrived. Within the span of a week, the landscape of journalism seemed suddenly inextricably altered and the discourse between doomsayers and AI enthusiasts began. Across the world many young journalists began to question the future of journalism and whether the profession would survive. Is AI the catalyst for the extinction of human journalists or a beacon of hope for future innovation?

It appears apparent that ‘Artificial Intelligence’ is the buzz word of the modern age as it fills our phone screens on every social media site, but the term itself is little understood. It is important to distinguish the different kinds of artificial intelligence. Although what is often most focused on is generative AI such as ChatGPT, there are various machine-learning models which can aid and improve efficiency in different stages of news production. Instead of focusing on the potentially catastrophic consequences of generative AI writing articles, I argue that artificial intelligence being used as a tool in newsroom can lead to a level of global interconnection and quality that has been unachievable before.

A journalist’s secret weapon

One of the most important uses of artificial intelligence, particularly as it becomes more advanced, is during the process of news gathering. A form of AI that has already become established a helpful tool was text analysis which was used by Quartz with the Mauritius Leaks. The leaks contained several hundred thousand pages full of data creating an impossible task for journalists to sift through and find patterns that expose wrongdoings. Through the use of a machine-learning model which was programmed to identify similar documents such as tax filings to those inputted by the journalists, they could identify discrepancies. Without this technological help, journalists would have most probably not been able to take full advantage of all the resources given to them. However, this machine-learning model was relatively primitive compared to the potential of artificial intelligence in the future.

Artificial intelligence can overcome a great deal of the barriers that journalists currently face when conducting research. The lack of language barrier for a machine learning model allows for greater access to resources on subjects and the speed at which a machine-learning model can scan across all the information inputted is incomparable to anything a human journalist can achieve. The use of such models in news gathering has so much potential. It could lead to greater integration of local news pieces in international reporting and allow more voices of underrepresented people to be heard.

Every day, a countless amount of content is added to the internet and to social media making it incredibly oversaturated and challenging to navigate. Newspaper archives storing previous reports on a subject can be used to check the consistency of statements made by public figures and provide useful historical context for an article. Modern-day journalists are often overwhelmed with information and are time constrained. The sources that are promoted to journalists are dictated by a biased search engine algorithm pushing content aligning closest to their own views and frequently geopolitically biased too. As data gathering models become more advanced, they can drastically reduce the labour of a journalist researching by offering an alternative to traditional search engines by providing a more focused and relevant search. Such news gathering models have the potential to produce more results outside of the echo chamber of their own country’s media by identifying more international sources discussing the same matters. Therefore, the ability to read a greater range of perspectives and find more information would lead to much higher quality journalism and a more informed general public.

News gathering models, I believe, will become the single most important tool in journalism but, as critics have rightly noted, there are many dangers to relying on AI in this manner. I stated earlier the potential benefit of the model gathering data from across the internet without the bias of search engines, but such a model can only exist if it is not programmed with bias to begin with. A machine-learning model builds off the input that is fed into it at the start leaving lots of room for journalists to end up in an echo chamber of their own opinions – a theme that will come up more frequently as this discussion progresses. For newspapers to fully take advantage of such models, they will require constant maintenance, updates and external regulation.

The fall of the language wall

The integration of machine learning models in journalism allows for a huge leap in the interconnectedness of global media. Large language models have the ability to translate articles into a multitude of languages with a degree of speed and fluency that has not been achievable before; this hugely benefits demographics that are underserved by traditional outlets. The translation of articles in a publication means an increase of potential readers. Therefore, journalists can target demographics across linguistics borders and focus on smaller issues, which have not previously been worthwhile or profitable within their publication’s language.

Such models can also be used for the transcription of audio taken from various media forms, enabling real-time translation and making more content more accessible. Just as journalists can benefit from hearing a wider range of perspectives when writing their articles, the readers themselves are given a lot more variety which is most important in more remote areas of the world where a significant proportion of the population cannot understand the most widely published languages. Not only will AI language models make more media accessible to those who cannot read dominant publication languages, it also has the potential to reduce the linguistic monopoly of English, thus protecting local languages.

While the revolutionary impact of such AI models on journalism is undeniable, one cannot overlook the impact on the labour market. With machine-learning models learning how to translate flawlessly into different languages and being able to transcribe audio, it eliminates the need for translators or manual transcription. This brings me to the primary concern of young journalists when discussing AI – if artificial intelligence can do every job and do it more efficiently, what jobs are left for humans?

In the industrial revolution, manual labour was replaced by machinery, but this new technology paved the way for economies to evolve and different careers were created. However, in the face of an AI that can do so much, many struggle to see where the new jobs will be created. Unfortunately, it seems inevitable that most of the jobs within the field of journalism that require repetitive, manual, or assisting work will become extinct. I, perhaps taking a more utilitarian approach, would argue that in this case the ability to share news internationally will allow for growth and the creation of new work to replace what is lost.

Thus far, I have stated the ways that artificial intelligence will improve the quality of articles so, as a reader, one may wonder: if the AI does the work in researching and finding the audience, why should it stop there?

Why not just use generative AI to write entire articles?

The future of journalism may be dominated by artificial intelligence, but it is still dictated by human choices. Recommender algorithms and large language models will be very influential moving forward, but the impact on society will be determined by regulations.

Over the past decade, social media has exposed the public to algorithms and how they tailor content in accordance with preferences. Newspapers have begun implementing such algorithms to push the articles a reader is most likely to read and display them as the front-page headlines. Such algorithms can allow writers to reach a greater audience of interested or targeted readers, however; the trade-off of a more tailored experience is the potential creation of digital echo chambers. Social media has already shown how easily extreme ideology can reach susceptible people, leading to greater political polarisation and social intolerance. If these algorithms are implemented to coordinate our exposure to and consumption of news, it could intensify the ideological bubbles already being manufactured on social media. During the rise of social media, traditional media have often taken on the task of combating misinformation with verified facts, credible sources, and less biased reporting. If these recommender algorithms are not tightly regulated and updated, modern society will be irrevocably harmed as divisions grow and politics become more polarised.

The credibility of journalism

For newspapers to survive in a world of openly accessible large language and visual models, they must rely on their reputation as being credible and trustworthy. A lot of misinformation circulates on the internet and large language models often cannot distinguish between sources that are accurate, and those that are misleading. The advancement of artificial intelligence will lead to more accurate doctoring of voices and images, so-called ‘deepfakes’, which can include faked audio recordings or videos of public figures. Unlike in the past where a video of an event could legitimise a report, such evidence will be weaker in a world of artificially produced recordings. Unfortunately, this has the potential to make reports on governmentally censored issues more challenging, as leaked photos can easily be dismissed as fake by an oppressive regime. Hence, not only can misinformation be dispersed more easily, but trust in journalists is endangered.

As the images, audio and texts AI models can generate become more convincing, the line between real and artificial becomes blurred. It will become increasingly difficult for the readers to determine what to believe. The newspapers that can be trusted to keep strict guidelines about using generative AI and fact checking sources will renew their positions as vital pillars of an informed society. Newspapers with a strong foundation of transparency and credibility should be able to retain the confidence of their audience to report accurately.

What is the value of originality?

Large language models such as ChatGPT and Google Bard have already proven to be capable of generating fluent and coherent language. Newspapers are already beginning to automate repetitive articles on matters such as sports results, real estate or any such articles whose purpose is to convey specific data rather than to report on an event. Whether or not an article could be written by a large language model comes down to what is expected from the reader. Articles written by such models are often predictable, bland, and repetitive, offering little to distinguish one article from another.

Yet, there is one aspect of journalism these large language models cannot recreate: creativity. The danger of recommender algorithms is creating content to gain success on the platform rather than original content that the audience needs to see. Large language models can reproduce a format of content similar to the input given, but they cannot create new, original ideas or articles. As more content by artificial intelligence is published, it will be down to readers to choose quality over quantity and to actively choose the articles written by journalists.

The future of AI

It is clear that the field of journalism will evolve in the face of artificial intelligence models. The journalistic and editorial process will integrate AI to replace the repetitive, arduous aspects and allow them to focus more on the creativity and quality of the piece within their tight deadlines. Where generative AI will become even more advanced and more convincing, the credibility of news institutions will be tested as readers seek out truth in the midst of potential misinformation. As with any technological tool, artificial intelligence must be used with caution, updated, rebooted regularly and used transparently or the trust of the audience will be lost. Ultimately, I believe these forms of artificial intelligence are more likely to be used for editing and complementing the work of a journalist rather than to replace them.

The image used in this article was created using Adobe Firefly – a text to image generating AI.

This article was written as part of the European Youth Press Forum for Member Organisations 2023, which focused on AI in journalism. Special thanks to Jonathan Hendrickxx for organising this wonderful event and to Tshepo Tshabalala and Johannes Skov Andersen for their valuable input to the discourse on AI in journalism.