AI in Journalism: A Paradigm Shift or a Passing Phase?


The integration of Artificial Intelligence (AI) into journalism is not just a technological advancement; it’s a paradigm shift both for the profession and for the business of making news. The Frontline Club event ‘Rise of the Machines: AI in Journalism’ on 24 May shed light on this transformation, raising questions that are both intriguing and essential. As AI’s role in newsrooms continues to grow, the landscape of journalism is undergoing a profound transformation, raising questions that are both intriguing and essential.

Generative AI: the double-edged sword
Generative AI, exemplified by tools like Chat GPT, is rewriting the rules of content creation. Reuters Institute digital journalist Marina Adami‘s observation captures the zeitgeist: “We’re at the cusp of a new era. AI isn’t just a tool; it’s becoming a collaborator.” This collaboration, however, comes with its challenges. The potential of AI to fabricate content that’s indistinguishable from human-generated content is both its strength and its vulnerability.

Dr. Bahareh Heravi added depth to this perspective, noting, “Generative AI is like fire. It can warm your house, but it can also burn it down.” The implications are clear: while AI can enhance journalistic capabilities, unchecked use can erode the very foundation of trust and authenticity.

“Over the last decade we’ve seen increasing use of AI throughout the media value chain; from metadata creation to sentiment analysis to recommendation engines. Generative AI is just the next iteration in this process,” said Patrick O’Connor-Read. Elaborating on the role of Large Language Models (LLMs), he added: “LLMs offer an evolution within this wider AI trend, but they are prone to hallucination. So the workflow will shift, rather than machines supporting humans, humans will support machines, providing input guidance and output verification/sensechecking.”

The trust equation: balancing innovation with integrity
For institutions like the BBC, trust is not just a value; it’s their currency. “In the age of misinformation, our commitment to truth is unwavering. But how do we integrate AI without compromising that trust?” David Caswell, former Executive Product Manager at the BBC, pondered during the Frontline Club event. Parmy Olson, a Bloomberg Opinion columnist covering technology, with a particular focus on AI and chatbots, echoed this sentiment, “The challenge isn’t just about using AI responsibly; it’s about communicating its use transparently to our audience.” This transparency is the linchpin that holds the trust equation together. The traditional model of journalism was linear: journalists create, and consumers consume. AI is disrupting this linearity.

Patrick O’Connor-Read, who has navigated both sides of this equation, producing TV and digital content for over ten years with companies like Zatzu and researching applied AI, offers a unique perspective. He points out, “It’s an open secret that the business models of many media outlets constrain time intensive human research; a lot of this low to mid level work can be automated and achieve near parity outcomes, at least for perfunctory output.”

As David Caswell shared, “It’s no longer just about what we want to say; it’s about how the audience wants to hear it.” This shift towards consumer-centric content is revolutionary. Marina Adami highlighted the transformative potential of this shift: “Imagine a world where news adapts to you, not the other way around. That’s the promise of AI in journalism.” However, with this promise comes the responsibility of ensuring that customization doesn’t lead to echo chambers, a concern raised Heravi, a senior academic at the Institute for People-Centred AI at the University of Surrey.

O’Connor-Read believes that while news outlets are personality-led, the emergence of synthetic personalities is imminent. “News outlets are personality led, people return to their favourite brands and hosts who provide a clear point of view. These sources of authority will continue to command attention, but it is only a matter of time – weeks and months more than years and decades – before synthetic personalities emerge able to process and synthesize information beyond the capabilities of a human, and attract and retain an audience.” He continues to explore the boundaries of what pure/total AI formats look like with his current venture, Zolayola, focusing on blockchain & applied AI.

The automation wave
The article from The Economist, titled ‘The Third Wave of AI in Journalism,’ highlights the initial phase of AI in journalism as automation. Machines have been assisting in delivering news for years. The Associated Press began publishing automated company earnings reports as early as 2014. The New York Times leverages machine learning to decide how many free articles to show readers before they hit a paywall. This automation wave has been primarily data-driven, focusing on generating news stories from structured data like financial reports and sports results.

Augmentation and analysis
The second wave, as described by computational journalist Francesco Marconi in the Reuters Institute article titled ‘ChatGPT: Threat or Opportunity for Journalism?’, shifted the emphasis to augmenting reporting. AI was used to analyze large datasets and uncover trends. This wave saw the use of machine learning and natural language processing to provide deeper insights and context to news stories. The Argentinian newspaper La Nación‘s use of AI to support its data team is a prime example of this phase.

Generative AI: a double-edged sword?
The third wave, which we are currently experiencing, is characterized by generative AI. These are large language models capable of generating narrative text at scale. While they offer applications beyond simple automated reports, they come with their own set of challenges.

While Madhumita Murgia recently appointed AI editor at the FT points out that while generative AI can synthesize information and make edits, it lacks the necessary originality and analytic capability that distinguishes quality journalism.

The future: collaboration or replacement?
The overarching sentiment from all sources is that while AI has a significant role to play in the future of journalism, it cannot and should not replace human journalists. AI can assist, augment, and even automate certain tasks, but the human touch, analysis, and intuition remain irreplaceable.

The future of journalism in the age of AI is not about machines taking over but about journalists and AI working in tandem to deliver accurate, timely, and insightful news to the masses.

Illustration: Lindsey Bailey/Axios

Follow on LinkedIn

Fun in crypto (+ news in 2023)


😎 Worked with Ronaldinho and BigONE Exchange on a NFT‌ collection

👍 Including working with Andre Figer, the son of Juan Figer, whom
Pele said was, “the first businessman in Brazil to respect the football player!”

⚽ See Andre in the panel below..with Matt House of SportQuake
and Daniel Liu CEO of Influxo, in a discussion I hosted on NFTs in football!

💵 My fav article for BigONE was a piece on ‘Why DeFi holds the key to metaverse success’ which appeared in Forkast.News.

📚 Contributed to Dr Mark Van Rijmenham’s book, ‘Step into the Metaverse: How the Immersive Internet Will Unlock a Trillion-Dollar Social Economy’ 

📔 Edited a book on NFTs (that sold out its own promo NFT) for Anndy Lian after the LUNA crash.

🤩 At DappRadar I organised a top NFT influencer event for NFT.London (in a Westminster pub in the heart of Government), which attracted top crypto influencers including Travis Wright of Bad Crypto.

📺 Secured top coverage from cyber reporter Joe Tidy at the BBC in a story about the growth of the metaverse (see on iPlayer, YT and the BBC news website)

🤝Worked on crypto wallet partnership at DappRadar with browser Opera — four years after their first crypto browser launch at Rise in London (me, in the audience 😇)

🏖️🩴DappRadar’s Barcelona ‘workcation’ highlight for me?
The Axie Infinity event.

🏅Where I went to my first eSports contest! 

🎩 And where I met and chatted with top crypto investor
and entrepreneur Yat Siu, Chairman of Animoca Brands. 

Yat Siu has just (17 Jan) facilitated a merger between Forkast.News and
CryptoSlam to form a new web3 info hub, Forkast Labs (Bloomberg report)

😇 However, the recent (4 Feb) FT article ‘Can the pioneer of blockchain gaming survive the crypto winter?’ takes a distinctly sceptical look at Yat Siu and his Animoca crypto empire. (But with such a sceptical story angle, there’s no mention of the smart merger work Siu facilitated between CryptoSlam and Forkast;)

Follow on LinkedIn

From Chain of Density to Finding the Needle in the Haystack: A New Approach to Summarisation

The Chain of Density (CoD paper) has been a revolutionary concept in generating increasingly dense summaries without extending their length. However, the real challenge often lies in identifying the most crucial or insightful information—the “needle in the haystack”—in any given article or text. This article explores how the CoD process can be adapted to focus on this aspect, providing a more nuanced and targeted approach to summarisation.

The proverbial needle in the haystack..

The Limitations of CoD
While the CoD process excels at creating denser summaries, it may not necessarily highlight the most critical points or “needles” in an article. The summaries become denser but may miss the central theme or the most impactful insights, which are often the most valuable takeaways for the reader.

The “Needle in the Haystack” Method
Inspired by the CoD, the “needle in the haystack” method involves an iterative process where the user actively participates in refining the summary. The steps are as follows:

Initial Summary: Generate a broad summary capturing the main themes.
1. User Feedback: The user reviews the summary to identify missing elements or areas needing more focus.
2. Refinement: Based on user feedback, the summary is refined to include these critical points.
2. Review and Decide: The user reviews the refined summary and decides whether it captures the “needle” or nub of the article, or if further iterations are needed.

Case Study: AI in Process Optimization
In a recent article about AI’s role in process optimization, the initial CoD summary focused on AI’s efficiency and safety features. However, it missed a crucial point: the human-created gaps in systems due to the “curse of knowledge.” Using the “needle in the haystack” method, the summary was refined to include this critical insight, providing a more comprehensive understanding of the article’s core message.

The “needle in the haystack” method offers a more targeted approach to summarisation, ensuring that the most critical insights are not lost in the quest for density. By combining the strengths of the CoD process with active user participation, we can create summaries that are not just dense but also insightful and focused on the crux of the matter.

Process using ChatGPT-4, by which the above article was generated through examination of the comparitive strengths of the CoD method, vs the ‘needle in the haystack’ approach.

How do you suggest we try out this CoD prompt for generating a summary?

Great question, Stuart! To try out the Chain of Density (CoD) prompt, we can take the following steps:

Select an Article: Choose an article or a piece of text that you’d like to summarize.

Initial Summary: I’ll generate an initial summary of the article, focusing on 1-3 key entities or main points.

Iterative Densification: We’ll then go through up to 5 rounds where:

I identify 1-3 new informative entities or details that were not in the previous summary.

I create a new summary that includes these new entities or details without increasing the length of the summary.

Review: After each iteration, you can review the summary to see if it meets your expectations for information density and clarity.

Final Summary: At the end of the iterations, we’ll have a summary that is dense in terms of the entities and details it includes, without being longer than the initial summary.

Would you like to proceed with this approach? If so, please provide an article or text that you’d like to summarize.

OK try out the Chain of Density prompt with this article:
AI in Process Optimization: A New Paradigm for Safer and More Intuitive Systems

Fantastic, Stuart! Let’s start by generating an initial summary of the article using the Chain of Density (CoD) approach. I’ll focus on 1-3 key entities or main points for the initial summary.

Initial Summary:

Artificial intelligence (AI) is revolutionizing business process optimization by complementing human decision-making and enhancing system safety. Ethical and security concerns, such as data privacy, are also addressed by AI.

Now, let’s move on to the iterative densification steps. I’ll identify 1-3 new informative entities or details that were not in the previous summary and include them in a new, denser summary without increasing its length.

Iteration 1:

AI in process optimization not only streamlines business but also improves safety and user-friendliness. It addresses ethical issues like GDPR and CCPA.

Would you like to proceed with another iteration to make the summary even denser, or does this meet your expectations?

Yes, proceed

Excellent, Stuart! Let’s proceed with the next iteration. I’ll identify new informative entities or details and include them in a denser summary.

Iteration 2:

AI enhances business efficiency, user safety, and ethical compliance, tackling issues like GDPR and real-world applications in healthcare and automotive sectors.

How does this denser summary look to you? Would you like to continue with another iteration, or does this summary meet your needs?

Yes, try another iteration

Certainly, Stuart! Let’s go for another round of densification. I’ll include more specific details without increasing the summary’s length.

Iteration 3:

AI optimizes business, ensures safety, and adheres to ethics, addressing GDPR and impacting healthcare drug interactions and automotive crash mitigation.

How do you find this even denser summary? Would you like to go for another iteration, or does this version meet your expectations?

The problem is that while technically speaking it is correct these are denser summaries, what they fail to do is find the ‘needle in the haystack’ — so my challenge to you is instead to come up with an iterative process which allows the user to find the needle in the haystack in an article, allowing the user of course through the process of iteration to decide for themself when to stop the process and decide that they have found the so-called needle in the haystack. Does that make sense as a process?

Absolutely, Stuart, your challenge makes perfect sense. The idea of finding the “needle in the haystack” resonates well with the notion of uncovering the most critical or insightful information in a given article. Instead of merely densifying the summary, the goal here would be to iteratively refine it to highlight what you consider to be the most valuable insights or “needles.”

Here’s how we could approach this:

Needle-in-the-Haystack Iterative Process:

Initial Summary: I’ll start by providing a broad summary of the article, capturing its main themes.

User Feedback: You review the summary and identify what you think is missing or what you’d like to know more about.

Refinement: Based on your feedback, I’ll generate a new summary that focuses on the areas you’ve highlighted.

Review and Decide: You review the refined summary. If you feel you’ve found the “needle,” we stop. If not, we go back to step 2 and continue iterating.

Would you like to proceed with this approach? If so, I’ll start by generating an initial broad summary of the article for you to review.

OK start by creating #1 the initial broad summary of the article

Certainly, Stuart! Here’s the initial broad summary of the article:

#1 Initial Broad Summary:

The article discusses the transformative role of Artificial Intelligence (AI) in optimizing business processes. It goes beyond efficiency to explore how AI can make systems safer and more intuitive. The article also touches on the human element in AI, ethical and security concerns, and provides real-world applications in sectors like healthcare and automotive design.

Feel free to review this summary and let me know what aspects you’d like to delve deeper into or what you feel is missing. We’ll use your feedback for the next iterative step in our “needle-in-the-haystack” process.

I feel what is missing from this summary is attention to the problem created by humans that create processes with invisible and sometimes risky gaps (the so-called “curse of knowledge” often leads experts to overlook gaps in systems and processes), and the fact that AI could flag such risky gaps.

Excellent observation, Stuart. Thank you for pointing out the missing elements. Let’s refine the summary to include the “curse of knowledge” issue and how AI can address these risky gaps in systems.

Refined Summary:

The article explores AI’s role in business process optimization, emphasizing its potential to make systems safer and more intuitive. It highlights the “curse of knowledge,” where human experts often overlook risky gaps in systems. AI, being unbiased, can identify these gaps and suggest corrective measures, thereby enhancing system safety and user-friendliness.

How does this refined summary resonate with you? Do you feel like we’re getting closer to finding the “needle in the haystack,” or would you like to iterate further?

I think that is getting to the principle of what the ‘needle in the haystack’ of this article is. Or the “nub of it”, the crux or central point of a matter. Do you agree?

Absolutely, Stuart. I agree that the refined summary now captures the “nub” or the crux of the article, specifically focusing on the overlooked issue of human-created gaps and how AI can step in to identify and correct these risky areas. It seems like we’ve successfully found the “needle in the haystack” in terms of the article’s central point.

Would you like to proceed with anything else, or does this conclude our iterative process for this article?

That’s a good output, I now have the option to write an article showing how the CoD process has helped refine the ‘needle in the haystack’ method.

Would you like to draft that article based on what we have explored in this thread?

Absolutely, Stuart! I’d be delighted to draft an article that showcases how the Chain of Density (CoD) process has evolved into a “needle in the haystack” method for finding the crux of an article. Here’s a draft for your consideration:

From Chain of Density to Finding the Needle in the Haystack: A New Approach to Summarisation

Web3 Tackles the Digital Quagmire: A Deep Dive into the Panel on Hate Speech and Misinformation

“In today’s digital age, misinformation and hate speech are as ubiquitous as smartphones,” Yasir Khan, Editor in Chief of the Thomson Reuters Foundation, declared as he opened the panel discussion hosted by Web3 on September 13, 2023. The panel featured a diverse group of experts, each bringing a unique perspective to the table.

What is Web3?
“Web3 is not just a technological shift; it’s a paradigm shift,” Yasir Khan emphasized. “It’s still in its infancy, but it has the potential to be a game-changer in addressing hate speech and misinformation.”

The Digital Battlefield
“We’re not just talking about hate speech; we’re talking about a life-and-death issue,” Khan asserted. “Misinformation and hate speech online have a direct impact on the access to life-saving information and services by displaced people around the world.”

The Ice Cream Activist
Jessie Macneil-Brown, head of social mission and activism for Ben and Jerry’s in Europe, was candid. “We’re an activist brand, and we’ve felt the heat,” she said. “Someone on the far right picked up one of our messages in France, and it led to a pile-on. It got so bad that someone even turned up at an employee’s house.”

The Humanitarian Perspective
Gisella Lomax of UNHCR was forthright. “Online harms are causing direct harm to refugees, asylum seekers, forcibly displaced communities, and stateless people,” she stated. “We need a multi-pronged approach—understanding the problem, developing tools to respond, and focusing on prevention.”

The Voice of Experience
Nadine Tunasi, a refugee and activist, spoke from the heart. “The impact on the mental well-being that hate speech has on refugees and asylum seekers is significant,” she said. “We’re not just statistics; we’re human beings.”

The UN’s Strategy on Hate Speech
The United Nations has a comprehensive strategy and plan of action on hate speech. Launched on June 18, 2019, by Secretary-General António Guterres, the strategy aims to tackle hate speech at both global and national levels. “Addressing hate speech does not mean limiting or prohibiting freedom of speech,” Guterres noted. “It means keeping hate speech from escalating into something more dangerous.”

The strategy focuses on two main objectives: addressing the root causes and drivers of hate speech, and responding to the impact of hate speech on societies. “As effective action must be supported by better knowledge, the strategy calls for coordinated data collection and research,” Guterres added.

What Can We Do?
“We all have a role to play,” said Khan. “Report hate speech, educate, support organizations, lobby for change, and be mindful.”

“In a world where hate can spread faster than a wildfire, it’s crucial to remember that we all have a role to play in dousing the flames,” Gisella Lomax concluded. “We need to do better. And indeed, we must.”

Call to Action
Contact your elected officials and demand action on this issue. You can also get involved in the fight against hate speech and misinformation by supporting organizations that are working to address this problem.

Particpants in ‘Web3 – Tackling Hate Speech and Misinformation’ with Thomson Reuters Foundation, Ben & Jerrys, Freedom from Torture and UNHCR on Wednesday 13th September:

Yasir Khan: Editor-In-Chief, Thomson Reuters Foundation. An award-winning journalist, editor, podcaster and documentary filmmaker, Yasir leads the Foundation’s digital news platform Context covering the impacts of climate, socio-economic inequality and technology on societies around the world.

Jessie Macneil-Brown: Head of Social Mission & Activism Europe, Ben & Jerrys. Jessie is a specialist with a strong track record of designing and delivering initiatives that inspire the public and drive long term sustainable change. She has experience in both corporate & NGO sectors. Her work is currently focused on advancing and protecting the rights of those with refugee experience. Jessie has previously worked for IKEA, The Body Shop and Amnesty International.

Nadine Tunasi: Survivor Speak Out Coordinator, Freedom from Torture.  A member of One Strong Voice network and Coordinator of the Survivor Speak Out group, Nadine is a refugee from Democratic Republic of Congo.

Gisella Lomax: Senior Advisor on Information Integrity (Misinformation, Disinformation and Hate Speech), UNHCR. Gisella leads UNHCR’s new capacity to address the harmful impact of mis and disinformation, and hate speech on digital platforms. Gisella started her career as a journalist, reporting for national and international media organisations.  She joined the United Nations in 2011, a time of global disruption to the information ecosystem driven primarily by social media. She has served in various digital and comms strategy roles across the UN, including Head of Social Media for UNHCR, where she saw up front the harm misinformation and disinformation can levy on people who are forced to flee.