The impact of artificial intelligence on journalism and other areas of communication Part II

Notes from the course “How to use ChatGPT and other generative AI tools in your newsroom”, taught by the Knight Center for Journalism in the Americas.

PART II

In the previous blog post, I shared PART I of the course notes: “How to Use ChatGPT and Other Generative AI Tools in Your Newsroom,” which I conducted with the Knight Center for Journalism in the Americas. For the EscribaNos (Escrib & OS) audience, I emphasized the aspects that have a direct impact on journalism and other areas of communication. In this post, I will cover the content of modules 3 and 4 of the course.

In module 3 we once again had Sil Hamilton as our expert instructor, who, as I mentioned in the previous post, is a researcher on Artificial Intelligence (AI) at Hacks/Hackers, “a network of journalists who reflect on and analyze the future of news through talks, hackathons and conferences.”

I concluded the previous post by stating that there are a number of “open discussions about generative artificial intelligence,” such as: given that AI systems have been trained on large amounts of pre-existing information, whose information was used for training and for which no credit has been given? And another question: once an AI system is fed new information and asked to perform some action, such as transcribing, translating, or improving the text, how does it reuse that information to continue training and generating more content for other users?

In this installment, I'll share some of the reflections we made on these and other issues in Module 3 of the course, during which we also introduced a glossary of terms related to this field of AI. We published the list of concepts and their definitions in the third edition of ACADEMO; you can consult it at academo.substack.com or explore the topic further in other sources.

Una de las lecturas realizadas durante este módulo del curso fue el artículo académico de Emily Bender, que analiza varios aspectos de un artículo publicado en The New York Times bajo el titulo “La IA está dominando el lenguaje ¿Deberíamos confiar en lo que dice?”, el cual, desde la perspectiva de Bender, parecía exagerar algunos aspectos negativos de la IA, e incluso presentar otros que ella consideró sin suficiente fundamento científico.

Throughout Module 3, literature on the various AI implementations underway around the world, particularly in the media and news production sectors, was a recurring theme. Despite the anxiety this topic has generated among many journalists, most of the information suggests that AI will enable the development of tools that complement, rather than replace, journalists.

Diversas empresas mediáticas han anunciado a sus colaboradores que se encuentran explorando los usos que estas herramientas permiten y hay otros casos en los que ya se están desarrollando soluciones a la medida que han sido solicitadas por algunas corporaciones de medios. En todos estos, sin embargo, la intervención humana sigue siendo un factor clave, así sea en una fase de supervisión.

Asimismo, vale la pena subrayar una cuestión que ha sido recurrente, la posibilidad de entrenar a los sistemas de IA para que desarrollen tareas rutinarias donde la creatividad  no es un factor determinante, para dejar espacio a los periodistas para tareas que requieran una mayor capacidad creativa. Los aspectos legales relativos al derecho de autor y otras cuestiones relacionadas siguen en discusión, cada vez más intensa.

En el último módulo del curso contamos nuevamente con la instrucción experta de Aimee Rinehart, gerente senior de Producto de Inteligencia Artificial de Associated Press (AP). Este módulo se centró en brindar información de calidad para que las personas, y especialmente las corporaciones de noticias, puedan tomar decisiones informadas sobre cuáles herramientas de IA utilizar, cómo utilizarlas, en cuáles tareas de la cadena de valor de las noticias y otros aspectos relacionados que también fueron rigurosamente abordados. Uno de los materiales a los que tuvimos acceso durante la última semana del curso fue la Guía de AP con nuevas directrices para la cobertura periodística sobre IA, la cual fue escrita por su reportero Garance Burke, y cuya lectura es recomendable para los comunicadores que se encuentren realizando algún trabajo sobre IA.

The results of an AI study with 200 newsrooms:

Instructor Aimee Rinehart discussed the results of a study she and her AP colleague, Ernest Kung, conducted with at least 200 newsrooms. They asked the newsrooms which areas of the news value chain they needed help with, focusing on repetitive and tedious tasks that could be freed up by AI to allow journalists to focus on more complex work. This question was posed in light of AP's AI usage framework, which aims to identify repetitive tasks performed in newsrooms and includes the following aspects:

• Defend the values ​​of the institution.
• Identify staff skills.
• Protect the data.
• Establecer límites.

One aspect that was consistently emphasized throughout this module of the course is that any AI tool implemented in a newsroom must be aligned with its mission and values. Likewise, the importance of not automating a flawed process was stressed; that is, before making a financial investment and a collective learning effort by staff on a particular AI tool, we must ensure that the process is appropriate, timely, delivers the expected results, and is ethically designed, because AI doesn't fix what is flawed; in fact, it could likely make it worse.

Transparency regarding the use of AI is another crucial element for its implementation. This means that the audience must be informed when AI has been used, through a public explanation of why and how it was used. Given that AI represents technological disruption, experts suggest that collective experimentation is the best way to establish the limits of its uses. One known issue is that while AI allows for content personalization, this could come at a high cost to audience privacy, because giving each person what they need or want also means providing the AI ​​system with sufficient information about each individual's profile.

How will online bibliographic searching change with AI tools?

According to Rinehart, internet search engines are expected to undergo significant adjustments by 2024, changing how they currently deliver search results. This overhaul may make it more difficult for people to find bibliographic references online because AI systems will have essentially rewritten them in as many different ways as there have been users searching for information on the topics. In fact, major social media platforms like Instagram, Facebook, and X (formerly Twitter) no longer prioritize local news articles.

It is predicted that as early as next year, search engines will respond to information requests with a paragraph summarizing the topic and a few links below that won't necessarily redirect to the original sources. Therefore, content marketing will become increasingly difficult because Search Engine Optimization (SEO) will shift from being a large data dashboard to having to hit the mark every time, so that your content is one of the few featured links.

The new era that seems to be opening up with AI in the field of news production also seems to suggest that media corporations will have to deploy greater efforts in the production of specialized newsletters, podcasts, WhatsApp groups and other messaging tools in order to direct their audience to their own platforms.

From the inverted pyramid to a new model for presenting news?

Journalism has historically been impacted by technological advancements, and for each of these disruptive stages, the field seems to have devised responses. The classic lesson in journalism schools about the inverted pyramid is a prime example of this constant reinvention. It emerged as a solution to the connectivity problems of the telegraph in 1899, which were addressed by prioritizing the most important information at the beginning of the news story. Comparable to this is the new way in which search engines are expected to present information to internet users, providing direct answers to their questions rather than simply offering links for further exploration. Similarly, new AI tools may change how media corporations present news to their audiences.

Currently, AP is developing at least five open-source AI projects, including those for newspapers, radio, television, and, of course, digital platforms. Some of these projects, which focus on generating summaries from video transcripts, have already encountered an initial hurdle when video quality is low. However, in most cases, the projects are progressing by providing tailored solutions to a variety of problems newsrooms face at different stages of their news production chain. It is important to emphasize, however, that for the implementation of these or any AI project, the following steps are recommended: first, to accurately determine the needs, and second, to consider both the news value chain and the aforementioned implementation framework developed by AP.

Finally, the efforts being made by different organizations to promote the responsible use of AI are highlighted, generating educational guides that provide guidance on this matter and other initiatives and research that have the same purpose.