Five key points for using generative AI effectively with Gerald Holubowicz

On November 7 and 8, two intense days of conferences were held in Nice, as part of the European IQ Media project. Dedicated to innovation in the media, these days brought together renowned professionals from all over Europe to debate and exchange views on the current and future challenges facing the media industry. Among them, Gérald Holubowicz led the workshop :  “How to prompt efficiently on Perplexity“.

Read the summary here.

See our videos to learn more about those two amazing days, just here.

He is a journalist specialising in digital products and artificial intelligence, following his talk entitled « How to prompt efficiently on Perplexity? ». As a user and teacher, he gave us his advice on how to make AI work for journalists.

1. Understanding AI 💡

Artificial intelligence exists for a long time, but generative AIs have arrived in our daily lives rather suddenly. Instead of fearing them, Gerald Holubowicz recommends taking the time to understand them: « The problem is not in manipulating the tool in the strict sense of the term, because empirically you can already get fairly good results, but it’s in understanding what it generates, what you can do to get around a certain number of things, what points you need to watch out for ».

2. Beware of AI bias 🧐

AIs respond to algorithms, so the content they generate responds to rules predefined by their creator. On this point, the journalist calls for vigilance: « There are cultural biases that get through, for the moment they are more inclusive biases, but tomorrow it is not impossible that we will see radical right-wing biases or bases that are polluted by these narratives ». Potential excesses that should lead to more supervision, quote Gerald Holubowicz : « We’re not even a quarter of the way through the first cycle in terms of ethics and regulation. ».

3. AI is a subject before being a tool 📰

AIs are not just used to produce prompts, they are also a tool in medicine, health and education. « Beyond being a tool for journalists, generative AI is also a subject for investigation into video surveillance, discrimination and its use in the army. In fact, it’s less a tech issue than a social science issue. »

4. Using generative AI to save time ⏱️

At a time when information is circulating ever faster, journalists need to be able to respond to this demand without their work losing quality. Gerald Holubowicz now uses AI as a writing aid: « I have a prompt that allows me to write a well-calibrated article from a draft in 25 seconds. After that, I correct the syntax to match my writing style, add information and enrich the text with links. At the end, the text has nothing to do with what was generated, but I’ve eliminated the problem of the blank page, which wastes time . »

5. Varying generative AI models🤝

A multitude of models have come onto the market in recent years, and in his round table, Gerald Holubowicz presented several of them, such as Perplexity, which like chat GPT or Gemini is a Search Generative Experience, or Poe, which brings together a number of AI models in the same way as a TV subscription that gives you access to several pay channels. The journalist himself sticks to the original versions, which he considers more complete than those offered on Poe. « For text, I mainly use Chat GPT and Claude, Flux for images and Eleven labs for avatars. »

Gerald Holubowicz is a French journalist specialising in digital products and artificial intelligence. A photojournalist for nearly 12 years and an award-winning web-documentarist, he has worked with a range of media companies including Libération, Condé Nast and Les Echos. Since 2014, he has been teaching at various journalism schools recognised by the profession, and also works with newsrooms to help them design new formats.

HARON LEVEAU (In partnership with IUT de Cannes, school of journalism)