How can you use generative AI to personalize news articles?

Always keep humans in the loop. Generative AI has come a long way in the last few years. It should not, however, be blindly trusted to produce news related content. Even the most advance models are prone to hallucination. When using AI for content creation, always have humans review the output for quality and accuracy.

The future lies where Humans and Generative AI can complement each other. AI's role is to assist and/or augment human capabilities across various fields. Whereas, Humans have to be there at every step ensuring the right inputs & quantify outputs, interpret context & make value judgement, training & fine tuning models regularly and finally using AI's capabilities ethically. Remember "With great power comes great responsibility." 😉 The other part, regarding Personalizing news articles could be achieved by collecting Geo information & interest areas as well as related news via various data collection techniques & use AI models to select better articles(via continuous learning & understanding reader's interest). Do not forget-"ethically"

At its core, GenAI tools like ChatGPT, are machine learning-based language models trained on vast amounts of text data. This allows it to understand and generate human-like language. Imagine having countless books, articles, and conversations under your belt—the AI uses this wealth of knowledge to analyze the context and respond in a way that's eerily similar to how a human would. The model doesn't just regurgitate information; it actually analyzes context to generate appropriate and coherent responses. It's all about probabilities—based on the context and the prompt, the model picks words that are statistically most likely to fit.

At its core, Generative AI refers to a subset of AI techniques that can create new content, ranging from images and text to music and even 3D models. It’s like having an intelligent, creative assistant that can generate unique and engaging content with minimal input from the user. This is accomplished through the use of advanced machine learning models, which have the ability to learn patterns and create new content based on what they’ve learned.

Generative AI is a combination of three things: 🧠 the ability to understand the language and questions posed to it, 📚 training on various knowledge bases, and 🎨 the ability to generate natural-sounding content from that knowledge. 👆🏼 super simplification of the concept However, the complexity of Generative AI grows when considering how to train the AI, which dataset to use, and what can be generated beyond text, such as computer code, solutions to math problems, images, and videos.

The key is prompt engineering. It involves crafting precise instructions to guide AI in generating desired content. To customize an article, define the main hypothesis clearly at the outset and the conclusion you aim to reach. Input these into the AI system and any specific angles or data points to be covered. The AI will structure the article around these core elements, ensuring the narrative flows logically from hypothesis to conclusion while incorporating the specified details, leading to a tailored piece reflecting your initial intent.

This is really an interesting question how can we leverage the power of LLMs to generate personalized news articles? Well due to their efficiency and speed of generating text , LLMs can be used to write personalized articles provided that the prompts are designed correctly, again this raises the concerns among journalists that they could lose their jobs to AI , this is not necessarily true at least for the time being because we need journalists to evaluate the quality of the personalized articles AI generates so over the next years we’ll need prompt engineering as a real skill to harness the full potential of AI in every aspect of our lives

The day before yesterday with the news breakout of the Israel-Hamas war, one of our customers immediately analyzed how that is going to impact their 1000+ truck owners, the system at Intelekt AI predicted that it will impact fuel prices. Immediately a personalized news notification popped up to their users with that information. News is going to be hyper-personalized, just-in-time and filtered.

For personalizing news articles, you can use prompt engineering to instruct the model to generate content fitting certain styles, tones, or topics based on a user's preferences. For instance, if a user prefers concise and positive news, your prompt could be: "Summarize this article in a positive and brief manner." By refining your prompts based on user feedback and preferences, you can continually enhance the personalized news experience.

When using LLMs to generate an article, it’s important to establish “voice” and style early on. You can train your LLM on your own voice if you are published and your content can be found. You can also have it write in the voice and style of others whose published works are accessible. There are ethical considerations here of course. For example if it’s your own voice or you are ghost writing for an author, it’s straightforward. If you are emulating another voice that is not your own, it is less clear ethically.

Automated journalism uses natural language generation algorithms powered by AI to automatically convert data into high quality news stories, images, video snippets, and data visualizations. AI generated stories can then also be distributed via automated journalism platforms. Additionally, by using AI-enabled data analysis and natural language generation-based reporting automation tools, we can create quick performance reports that would otherwise take many hours to analyse.

Increased Engagement: Customized content is more likely to capture and hold a reader's attention, leading to higher click-through rates and longer time spent on the platform. User Retention: Tailored content can make users feel that the platform understands their needs and interests, increasing loyalty and reducing churn. Relevancy: Generative AI can filter out irrelevant information, delivering news that is most pertinent to the individual, thus adding value to the user experience. Ad Revenue: More engaged users and higher retention rates can lead to increased advertising revenue for online news platforms. Data Efficiency: AI's ability to analyze and act on data quickly means that the personalization can happen in real-time.

By analyzing and understanding past readers habits, interests, behaviors and even the time spent on different articles or sections of the article can help you understand different variables to create highly personalized articles that engage readers.

Generative AI can filter and prioritize news articles, helping users avoid information overload with personalization. Instead of sifting through a vast array of articles, users can access a curated selection that suits their preferences. Personalized news can provide opportunities for targeted advertising and subscription models. Advertisers can reach a more relevant audience, and news publishers can potentially increase subscription revenues.

Embracing generative AI in curating news stories unlocks a wealth of opportunities to craft compelling, personalized content that resonates with each individual reader. This technological intervention ensures that every piece of information is not just read, but experienced and remembered. - Enhanced Engagement: Personalized stories cater to individual interests and preferences, significantly boosting reader engagement and interaction. - Increased Relevance: AI-driven personalization ensures that readers always encounter news that is most pertinent and meaningful to them, enhancing their reading experience and satisfaction.

You’ve got to make sure the personalisation doesn’t change the fact of the news. News is news, if it becomes overly personalised than we may end up creating false sense of realities. As news creators it’s important to know where the level of personalisation should take place.

I had a frustrating experience with this recently when I asked chatGPT to write an article (as a starting point) for my satyrical blog. Apparently “you are a bougie mom that loves her country club lifestyle” is a voice chatGPT isn’t willing to generate 😂

Navigating ethical concerns like echo chambers, ensuring robust tech infrastructure, and maintaining human editorial oversight are key challenges in using generative AI for news personalisation. These elements are critical to balance personalised content with reliability, diversity, and ethical considerations.

Data Privacy: Collecting user data for personalization raises privacy concerns that must be addressed through secure data handling and compliance with regulations like GDPR. Quality Control: AI-generated content might lack the depth or nuance a human writer could provide, potentially affecting the quality of news. Ethical Concerns: There's a risk of creating echo chambers by only presenting views or news that align with a user's preexisting beliefs. Computational Costs: Real-time personalization can be resource-intensive, requiring robust computational infrastructure. Content Limitations: Generative AI may struggle to fully comprehend complex topics, which might result in overly simplified or inaccurate articles.

We need to be wary about creating “echo chambers” (environment wherein people keep ingesting the same information and become radicalized or non tolerant to other views). This phenomenon has been observed in social media’s content algorithms as well. We need to ensure that periodically, users are also exposed to diametrically opposite viewpoints in their news recommendations.

Step 1- Think of a demographic you want to personalize an article for. Step 2 - Understand their values, beliefs, interests, etc. Step 3 - Pick an LLM (ChatGPT, Bard, Claude) Step 4 - Start playing around. Learn by doing.

A good way to make an AI copy your writing style is to share some of your previous writing with it. This helps the AI learn how you write, your tone, and what kind of stuff you like to write about. Remember, it's a gradual process to AI mimic your style, and the initial outcomes might not be great. But don't give up! Keep giving feedback until it gets it right. A good prompt to start this process would be: I am going to train you to become a GPT article writer ### The following paragraphs are written in a style called 'xyz' Paragraph 1: Sample Paragraph 2: Sample Paragraph 3: Sample ### I'll give you a subject and an audience, and you will write an article for a blog in 'xyz' style.

As part of the implementation process, it's essential to continuously iterate and fine-tune your generative AI model to better align with the evolving preferences and behaviors of your audience. By consistently gathering and analyzing reader feedback and adapting the AI model, you can ensure that the personalized news articles remain engaging, relevant, and in line with the expectations of your readers.

The best way to start is to jump right in without being intimidated. Free generative AI models let you start personalizing content easily. Start with a firm understanding of your objectives and audience, then feed the model relevant data (e.g., local posts/articles/profiles). We lifted engagement by localizing messaging and creative to resonate in each local market vs. one-size-fits-all. Generic advertising that ignores regional nuances, dialects, or culture risks confusing audiences, just like an instruction manual written in the wrong language would baffle users. Start testing models, learn what works, and iterate. No need for perfection; the key is taking the first step and building on small wins. Have confidence; you got this!

To get started with personalizing news articles using generative AI, one should first gather and analyze data on readers' behaviors and preferences. Next, integrate a suitable generative AI model, continuously training it with this data, and test its effectiveness in tailoring content to individual reader profiles while maintaining editorial standards.

Use AI to rewrite headlines, alter tone, adjust length, swap examples and stats to resonate with each reader. Rearrange layout, format, fonts based on user preferences and accessibility needs. Insert location-specific details, businesses, names tailored to the reader's city or town.

This article raises concerns among all of us that AI is gradually insinuating into our lives to perform the tasks we tend to do with unprecedented efficiency making people redundant, again these concerns are understandable and explicable due to the incredible speed with AI technologies are emerging, even AI developers and Gurus like Jeffrey Hinton are absolutely flabbergasted by this development, AI can do just about everything in our lives arts , literature, data analytics, programming, diagnosing diseases , the question is how long will take AI to become autonomous thinker and take decisions surpassing human capabilities? How to prevent AI applications from falling into the hands of bad actors ?

🥇 One of the top players to watch in this space is Artifact. Founded by the co-founders of Instagram, this AI-powered news app curates stories for you based on your interests, preferences and behaviors. It has a summary feature that can present each story in multiple summarized formats including “Explain It Like I’m 5.” Through their partnership with Speechify, you can also listen to articles while choosing from different AI-powered voices and accents (including Snoop Dogg and Gwenyth Paltrow.) Or even mark an article as clickbait based on its title, so the app can rewrite the headline to align it more accurately with the content. Their long-term strategy is the creation of a completely personalized news for an audience of 1.

I'd start by asking myself "why would I want to do this", before asking "how would I do it" so I could decide IF it was appropriate and worthwhile. Context is everything in something like this, so without it I would categorically not "personalize to alter" news or any topic without careful review. Using generativeAI to summarize or recommend is one thing, personalization, especially depending on "intent", is a slippery slope, IMO. That said, an interesting variation would be to apply the concept of six degrees of separation not to people, but to ideas, events, etc. to broaden how we think and correlate what might appear disconnected. Why? The really interesting stuff is where "worlds collide".

A crucial aspect of personalisation involves accommodating individuals' preferred methods of consuming information. While some individuals favour reading text, others may opt for visual content through images, engage with video material, or prefer auditory content. The true potency of personalisation emerges from adapting to people's learning habits and crafting multimodal content.