This past summer, I presented actionable AI plans to publishers at several press and revenue events. When I was first invited to present, I wanted to discuss amazing tools and the possibilities. To prepare, I reached out to several of my publisher friends who represented the groups I was presenting to. No judgment, but they all said, “We really haven’t started yet. We’d like to.”
In fairness to everyone, AI is a moving target. It’s overwhelming. It’s moving faster than any technology we’ve ever had in the modern era. Many people don’t even know what they don’t know, but more importantly, they don’t know where to start.
I won’t get into the ethics of AI; there are people who are way more qualified to discuss that.
Let’s start with the number one complaint I’ve heard through all my publisher discussions: “AI just makes things up. It cites things that don't exist.” You've probably all seen this. I have.
Why does AI hallucinate?
AI today is synonymous with LLMs. LLM is short for Large Language Model. Large Language Models are statistical averages of question-and-answer pairs. Billions of question-and-answer pairs. When an LLM receives a question it has never seen before, it, like a human, comes up with the best answer based on the average of all the past question-and-answer pairs it can. This inherent averaging can cause details to be left out or lead to a made-up answer.
We know AI for its large language models or LLMs. Brand names like ChatGPT, Claude, Gemini and Llama are all LLMs. Different versions have different capabilities. Some are better at logic, some are better at creativity, and others are better at programming. What sends an LLM down one discipline or another? Training data. To make it better at coding, feed it code. To make it better at poetry, feed it poems. History? Feed it history books.
To bring it down to human terms, think of an LLM as a brain, and everyone’s brain is wired differently based on their experiences. Your brain knows every class you’ve ever taken, every book you’ve ever read, every program or movie you’ve ever seen, and every interaction you’ve ever had. It all went into training your brain. Suppose I ask you to “write a story about coffee.” Maybe you’ve never written a story about coffee before or have never drunk coffee. In that case, you can still make the word association and recall other shared experiences about coffee to compose a reasonable story.
The more specific you are about what you need, the better the results (more below). If you ask AI to write a story about something generic like coffee, with a good prompt, it will return a decent story. If you ask it to write about a school board meeting or an event that has never happened, be prepared for anything because the AI will make things up.
Here is an anatomy of a news prompt. Note that the reference material related to a story should be made available in a knowledge base or what is known as retrieval augmented generation or RAG. You’ll hear that expression a lot related to news writing or co-pilots.
Anatomy of a news prompt
(Retrieval Augmented Generation - RAG.)
Who will win using AI in local media?
AI is the greatest tool ever developed and also the great equalizer for local media or any business. It gives you access to expertise and talent you could never attract to your market, and if you could, you probably couldn’t afford them. Given the appropriate guidance, it works tirelessly and will do anything you need or want.
That being said, it is available and open to anyone who wants to master it. The cost of using AI tools and techniques is negligible.
That’s the good news. The bad news for publishers is it is available to anyone who wants to use it. The classic “two guys/gals in a garage,” especially if they are technically astute and aggressive in using AI as tools, will dominate the space.
If you are slow to adopt and adapt, you will lose. We are in a major period of disruption that will make the invasion of digital platforms look like a minor inconvenience because, with AI, anyone can become your new competition.
Small is the new big
I personally believe that the greatest success stories and case studies will come from smaller media companies. In the world of AI tools and automation, smaller, more nimble companies can dominate. Committees, approval processes and institutional drag will bog down larger companies. In the new world of AI-driven entrepreneurship, speed is a competitive advantage. The winners will be those who can quickly adopt, integrate and iterate.
I predict you will see the best, most innovative publishers and operators become frustrated with corporate oversight and build next-generation media companies unconstrained by legacy staff, infrastructure and business models.
What is AI’s real superpower?
While most of the use cases for AI are headline writing, summarizations, copy editing and data mining, AI’s superpower is in autonomous agent systems or AI agents. AI agents can take a task with guidelines/guard rails and break them down into several subtasks, then accomplish those subtasks using the set of tools the system has available.
So, how is this different from an automation or an automated task that uses AI in its workflow? Automations typically fail if the appropriate tools or input are not available. Agents will find workarounds to accomplish tasks even if it doesn’t have the proper tools. That sounds scary, but good AI agent systems will surely put a “human in the middle” for checks and balances.
Biggest challenges of AI for local media publishers
I've spoken to dozens of publishers about AI, and the three biggest challenges are:
Invest in expertise, not tools
For the first time in the history of technology, the cost and access to the technology are not barriers to entry. Most of you know that a subscription to paid ChatGPT, Claude and Perplexity is about $20/month. If you require API access to power your tools, they are very affordable.
A tool like Langflow, which will allow you to build virtually any AI app, is free and open source. But what do you do with the pieces?
The reality is everyone needs a “guy.” While I can address 95% of all publishers’ needs and questions, I have a “guy” who I go to to solve my problems. Scott Pierce, a former colleague and chief architect at a media tech startup, is my guy. His Left Join Studio is an innovation lab where he takes my ideas from concept to product at record speed. On that note, finding the right resource is the most important thing in the entire journey. Find a partner whose skills and philosophy are in alignment with yours.
Five things publishers need to do TODAY!
Guy Tasaka is a seasoned media professional with a 35-year track record of leading change in the industry. He has collaborated with renowned organizations such as Macworld Magazine, Ziff-Davis and The New York Times, where he honed his expertise in research, strategy, marketing and product management. As the former chief digital officer at Calkins Media, Guy was acknowledged as the Local Media Association's Innovator of the Year for his work in advancing OTT and digital video platforms for local news organizations. He is also the founder and managing partner of Tasaka Digital, specializing in helping media and technology companies navigate business transformations using his extensive experience and forward-thinking approach. Guy can be reached at guy@tasakadigital.com.
Comments
No comments on this item Please log in to comment by clicking here