"AI can help save the press"

Interview
"AI can help save the press"

TX Group, Switzerland's largest private media group, has been deploying artificial intelligence in its newsrooms for the past three years. Titus Plattner, a project manager at the AI Lab of Tamedia (a subsidiary of TX Group), explains how it is being used.

While some in the profession fear the rapid adoption of artificial intelligence (AI) in newsrooms, certain journalists see it as a lifeline for the media industry. Without denying AI's darker side, Titus Plattner – who spent nine years as a journalist in the investigative unit at Tamedia (a subsidiary of TX Group) – falls firmly into the latter category. Since returning from a year at Stanford University (near Silicon Valley) in 2018, he has been devoting his days to developing AI tools for the group's newsrooms. In his view, AI has much to offer the media, provided it is used properly.

With nearly 600 journalists and prestigious titles such as Tages-Anzeiger, Finanz und Wirtschaft and 24 heures, TX Group is Switzerland’s largest private media group. How are you using AI in your newsrooms?

AI is used throughout the entire article-production process, from information gathering to distribution, including document research and writing. Over the past three years, we have been gradually introducing generative AI functions into our systems. We began by automating numerous small tasks that, while necessary, are not particularly engaging for our teams. For example, when a journalist enters an article into the content management system (CMS), an AI tool can automatically suggest a range of headlines or propose tags for search engine optimisation. Image descriptions for visually impaired readers have also been automated for some time. But now we’re taking it much further: for instance, we’ve just rolled out a tool that can flag awkward phrasing and suggest improvements. The journalist remains in control and can choose whether or not to accept these suggestions. Everything is moving very quickly: today, we are capable of doing things that I couldn’t have imagined possible back in January. The idea is not to replace journalistic work, but to streamline or speed up the production process. In this ongoing transformation, Tamedia’s advantage is that it has the scale to support the necessary investments.

 

Isn’t it risky to entrust such tasks to a machine that could make a mistake?

Our guiding principle is ‘human in the loop’: whatever is published is checked by a human and remains the responsibility of an individual. This is the opposite of chatbots, which state in very fine print that their system may make errors, thereby absolving themselves of any responsibility. The danger would be a form of ‘cognitive surrender’, where we delegate to the machine without further verification. These systems often produce good results – sometimes even excellent ones – but they can also go off the rails in very insidious ways. Even though large language models (LLMs) are improving at a phenomenal rate, generative AI remains an emerging technology that is prone to error. Yet, in the media, we cannot afford to publish false information. Our credibility is at stake. That said, I believe that, when used properly, AI improves quality. We are, for example, in the process of rolling out a fact-checking tool. Models remain limited when it comes to very recent or exclusive news, as we often work right at the forefront of current events. But they can spot errors in more widely known information, such as the name of a person quoted in an article. Previously, this sort of error would occasionally slip through the net. Today, AI is helping to tackle this task. And I think we’ll have fewer mistakes in the future.

 

Errors can also arise from people who misuse AI tools…

Yes! It is essential that journalists use these tools in areas they are familiar with. Someone who knows nothing about elections in a particular canton will not be able to assess the reliability of a machine-generated text on that subject. That is why we devote a great deal of energy to training, probably more than to implementing the tools themselves. It is important to understand not only their capabilities, but also their biases, risks and limitations.

 

How is this transformation being received in Tamedia’s newsrooms?

The AI Academy – a nine-day training course we launched in-house – received three or four times as many applications as there were places available. That’s a good sign. In 2023, there was a certain amount of scepticism, which was perfectly understandable. Today, the situation has changed: everyone uses generative AI on a daily basis, sometimes without really realising it, for instance on their mobile phones or via search engines. Everyone now has a chatbot in their pocket.

 

But the automation of tasks could lead to job cuts…

If there have been redundancies in recent years, it isn’t because of AI, but because the press’s business model is broken. Advertising revenue for private media (ed. note: print and online press) in Switzerland has fallen from 3 billion Swiss francs in 2000 to less than a billion in 2024. In French-speaking Switzerland, where the market is smaller, all the daily newspapers are struggling to break even. And in the country’s German-speaking part, the situation is also very difficult. To survive, our industry must do more with less. AI, however, enables us to become more efficient. The responsible implementation of this technology could help the press regain a viable business model. It is not a one-size-fits-all solution, but it will be an essential component.

 

The major LLM platforms use journalistic content to train their models. Is Tamedia considering signing a licensing agreement, as major groups such as Le Monde in France have done?

To my knowledge, no Swiss media group has entered into any significant agreement with these companies. We are keeping our options open, but the agreements seen abroad involve relatively small sums. They may be welcome, but they will not save the media. Switzerland is a small market, even when considering its various linguistic regions. Selling one’s archives may bring in a few million on a one-off basis, but that will not provide a lasting solution to the economic difficulties.

Get the print magazine delivered to your doorstep

Subscribe to Swissquote Magazine and get the scoop on what’s happening in the financial world. 

Enjoy six issues over 12 months for just CHF 40 — cheaper than the newsstand price.

Who will survive AI? cover

Swissquote Bank SA
Chemin de la Crétaux 33
CH - 1196 Gland

Email: magazine@swissquote.ch

Tel: +41 (0)22 999 94 54

Switzerland

Progettato con passione in Svizzera

Considera i rischi

Il trading di prodotti a effetto leva sulla piattaforma Forex, come cambi, prezzi dei metalli preziosi spot e i contratti per differenza (CFD), comporta un rischio significativo di perdita dovuto alla leva finanziaria e potrebbe non essere adatto a tutti gli investitori. Prima di aprire un conto con Swissquote, è necessario considerare il proprio livello di esperienza, gli obiettivi di investimento, gli attivi, il reddito e la propensione al rischio. In teoria, le perdite sono illimitate e potrebbe essere necessario effettuare pagamenti aggiuntivi se il saldo del conto dovesse scendere al di sotto del livello di margine richiesto. Pertanto, non si dovrebbe speculare, investire o fare hedging con un capitale che non ci si può permettere di perdere, che sia preso in prestito o che sia urgentemente necessario o indispensabile per il sostentamento personale o familiare. Negli ultimi 12 mesi, il 68.22% degli investitori retail ha perso denaro nell'ambito del trading su CFD, ha subito una perdita totale del margine alla chiusura della posizione o si è ritrovato con un saldo negativo dopo la chiusura della posizione. In caso di dubbi, è necessario essere a conoscenza di tutti i rischi associati alla negoziazione in valuta estera e rivolgersi a un consulente finanziario indipendente. Per maggiori dettagli, comprese le informazioni sull'effetto leva, sul funzionamento dei margini e sui rischi di controparte e di mercato, si rimanda alla nostra Informativa sui rischi di CFD e Forex. Il contenuto del presente sito web costituisce materiale pubblicitario e non è stato sottoposto all'attenzione di o approvato da alcuna autorità di vigilanza.

Contenuti generati dall’IA

Alcuni dei contenuti visivi presenti sul nostro sito web sono stati generati e/o migliorati utilizzando applicazioni di intelligenza artificiale (IA). Tuttavia, tutti i contenuti sono sottoposti a un'attenta revisione e approvazione da parte di esseri umani per garantirne l'accuratezza, la pertinenza e la conformità alle esigenze dei nostri utenti e clienti.