Over the past year, I’ve been prototyping generative AI tools in media. A few things I’ve noticed for large language models (LLMs) that I believe will further develop in the upcoming year:

• It’s better to augment skills rather than replace with AI. Using chatGPT to write is okay, sometimes good, but not great. Instead, using it as a place to bounce ideas or get started quickly in a larger workflow is wonderful. I’ve found this for programming too — it’s easier to edit and add rather than start from scratch when coding.
• LLMs make text malleable. You can quickly expand bullet points into paragraphs or contract articles into summaries. Given some text, you can shape it for a particular interface: a summary for a newsletter preview, a quick phrase for a notification. To illustrate this, I made an app where you can pan across a slider and see the article length go from a two-word summary to the full text length.
• User interface for LLMs is an open field of opportunity. LLMs are powerful and mysterious and the way we interact with them is through chat, which is simple, but not entirely welcoming. You have to learn how to prompt and get a feel for what LLMs are good at. A blank input, with no indication really of what it can do, feels analogous to the command line. The command line as a UI gave way to a more welcoming, intuitive, interactive, and expressive interface: graphical user interfaces. Now expanding the analogy, what will the future interfaces look like for LLMs? To put it succinctly  —  command-line : GUI :: llm chat :: ???. This is the space I’m most excited about. We have the opportunity to define a new human-computer interaction for this era of LLM computing.

In past years I’ve written my year-end predictions as articles, zines, or illustrations. This year my prediction comes in the form of a GPT: Nieman Lab Predictions GPT.

I made a GPT and used retrieval augmentation to give the it access to the past 13 years of Nieman Lab year-end predictions. That’s 1,369 articles.

What this means is I now have a custom chatGPT that can draw from these predictions when I ask it questions. Now, what would you ask? You can explore the future of news in an interactive way. Following from my points above, rather than just asking “what 2024 will bring in the world of news,” I’m interested in this as a tool in the process of making predictions, using it in a larger workflow — augment, instead of replace. To use this in your own forecasting, here’s a link to the GPT. If you don’t have access to GPTs yet, you can tweet me a question for it and I’ll share the response.

Below I’ve highlighted a few questions and responses from Nieman Lab Predictions GPT. You can see my full conversation here.

[Editor’s note: This is as good a place as any to note LLMs have a tendency to hallucinate.]

Where have the predictions been most accurate, and where have they been most off? Give specific examples.

The Nieman Lab predictions have been notably accurate in some areas while missing the mark in others.

Accurate predictions:

• Digital transformation: Predictions about the shift toward digital journalism have been largely accurate. The rise of online news consumption and the importance of social media platforms in news dissemination are key examples. Predictions about the growth of mobile news consumption also proved true, as smartphones have become a primary news source for many.
• Monetization struggles: The predictions about the decline of traditional advertising revenues and the need for alternative revenue models were spot on. Many media organizations have shifted toward subscription models and membership programs, as predicted.
• Rise of podcasts and newsletters: Predictions about the burgeoning popularity of podcasts and the resurgence of newsletters have been validated. Podcasts have become a significant part of the media landscape, and newsletters have seen a revival as a direct-to-audience communication tool.

Inaccurate predictions:

• Virtual reality (VR) in journalism: Some predictions overestimated the impact of VR in journalism. While there have been experiments and implementations, VR has not become a mainstream tool for news reporting or consumption to the extent predicted.
• Impact of Artificial Intelligence: While AI has significantly impacted journalism, some predictions may have overestimated its immediate transformative power. The adoption of AI in newsrooms has been gradual and more focused on assisting rather than replacing human journalists.
• Decline of Big Tech’s influence: Earlier predictions that regulatory actions or market forces might significantly diminish the influence of big tech companies in the news ecosystem haven’t fully materialized. Companies like Google and Facebook continue to have a major impact on news distribution and monetization.

These examples showcase the challenge of predicting in a rapidly evolving field like digital media and journalism. While some trends were accurately foreseen, others evolved differently than expected, often due to the complex interplay of technological advancements, market forces, and societal changes.

Which areas were overlooked among the Nieman Lab predictions specifically?

Historically, Nieman Lab predictions have predominantly focused on how journalism will adapt to rapidly changing technology, including digital, social, mobile, and virtual reality advancements. These predictions have been centered on the distribution, reporting, and the overall nature of news in the face of technological shifts​​.

Areas that may have been overlooked in these predictions include:

• Journalistic ethics in the digital age: While technological advancements were heavily discussed, there may have been less focus on the evolving ethical challenges posed by these technologies.
• Mental health and wellbeing of journalists: The impact of the fast-paced, always-on nature of digital journalism on the mental health and wellbeing of journalists might not have been a prominent topic.
• Impact of journalism on democracy and social movements: The role of journalism in influencing democracy, social movements, and civic engagement, beyond just the dissemination of information, could have been underexplored.
• Diversity in newsroom leadership: While diversity in journalism has been a topic of discussion, specific focus on diversity in newsroom leadership and decision-making roles might have been less emphasized.
• Local journalism beyond the financial crisis: Discussions on local journalism may have heavily focused on the financial struggles, possibly overlooking innovation, community engagement, and the role of local journalism in community identity and cohesion.
• The intersection of journalism and education: The relationship between journalism and education, especially in terms of media literacy, and how journalists can contribute to educational initiatives, might not have been a central theme.

These areas represent important aspects of journalism and news technology that could benefit from more attention in future predictions and discussions.

What were the biggest errors in prediction in AI and machine learning in the predictions?

The biggest errors in predictions regarding AI and machine learning in journalism, as gleaned from Nieman Lab’s decade-plus of forecasts, revolve around the overestimation of AI’s capabilities in replicating human journalistic skills and understanding complex socio-political contexts:

• Overestimation of AI’s ability to replace journalists: There was a notion that AI and machine learning would lead to “robot journalists” capable of independently reporting facts without human intervention. This expectation underestimated the complexity of journalism, which involves not just reporting facts but also contextualizing them within a nuanced, multi-polar world. It became evident that machine learning systems were not equipped to handle such complexities in the foreseeable future​​.
• Misjudging AI’s capacity for emotional and affective storytelling: Another error was in expecting AI to competently produce content like news reports or travel reviews with the same depth of emotional and affective narrative as human journalists. Predictive analytics and machine learning operate on mathematical logic and are limited in their ability to emulate the emotional and narrative depth characteristic of human journalism. This realization underscored the unique value of human journalists in creating long-form, affective narratives, an aspect of journalism that AI could not replicate​​.

These errors highlight the limitations of AI and machine learning in journalism, particularly in areas requiring deep understanding, emotional intelligence, and narrative skills. While AI has found a place in assisting journalists, it has not replaced the need for human insight and storytelling prowess.

What did Kawandeep Virdee write about?

Kawandeep Virdee has written on a variety of topics related to media and journalism in the Nieman Lab predictions. Here are some examples:

• “Zines had it right all along” (2017): This article does not provide a detailed summary in the file, but the title suggests a focus on the value and relevance of zines in the modern media landscape​​.
• Growth of micro services for content (2018): Virdee discussed the emergence of useful tools for newsrooms that are embeddable, facilitating diversified content creation. This includes services offering features like data visualization, audience feedback, and 360 images and videos​​.
• “Media wants to take care of you” (2018): In this article, Virdee addressed the issue of media seeking attention from consumers. The piece illustrates the side effects of this attention-seeking behavior and the need for media companies to consider the overall wellbeing of their readers​​.
• “Goodbye, doomscroll” (2020): Virdee wrote about the experience of increased screen time and the phenomenon of doomscrolling during the pandemic. The article highlights the anxiety and stress caused by continuous exposure to extreme content in digital feeds​​.

These articles reflect a diverse range of interests and perspectives, from technological innovations in journalism to the psychological impacts of media consumption.

What do you think he’ll write about for the 2024 prediction?

Based on Kawandeep Virdee’s past articles, which range from technological innovations in journalism to the psychological impacts of media consumption, for the 2024 prediction, he might explore topics that intersect technology, media, and societal well-being. Here are some speculative ideas:

• “Navigating the infodemic: Strategies for media in the era of misinformation”: Virdee could explore how media organizations can effectively navigate and combat the growing challenge of misinformation. The focus might be on innovative tools and ethical approaches for fact-checking and promoting media literacy among the public.
• “The mental health impact of news consumption in a post-pandemic world”: Building on his previous work on doomscrolling, Virdee might delve into the longer-term mental health implications of news consumption patterns that emerged during the pandemic, offering insights into healthier engagement with news media.
• “Embracing digital minimalism: A new paradigm for media consumption”: He could explore the concept of digital minimalism in the context of media consumption, advocating for more intentional and less overwhelming engagement with digital news and social media platforms.
• “The evolution of social media: From attention economy to community building”: This prediction might focus on the evolving role of social media, shifting from purely seeking user attention to fostering genuine community engagement and support, and how media companies can adapt to this shift.
• “The Role of AI in promoting emotional wellbeing through media”: Virdee might investigate how AI and machine learning can be used to create media content that supports emotional well-being, moving beyond information delivery to nurturing a healthier relationship with news.

These speculative topics align with Virdee’s previous focus areas and the current trends in media and technology, offering plausible directions for his 2024 prediction.