AI strategies for newsrooms... and the content apocalypse that is coming
PLUS Chatbot girlfriends and the importance of databases.
Here’s a prediction: some newsrooms are going to take the race to the bottom with the adoption of AI and they are going to suffer for that. Because you can imagine certain newsrooms using AI as an opportunity to reduce costs, fire staff and produce the same product (just cheaper and faster).
However, the next wave of content creation is going to be brutal. People are obsessed with everything getting worse because of AI, but really it is going to mean that the floor for content quality is going to get considerably higher. Which means that newsrooms, instead of cutting costs, are going to have to use AI, retain their staff and aim to make their content way, way better to be able to charge or attract anyone. That’s the only way to avoid the coming content apocalypse. To be frank, for any of us to survive we are going to need to embrace AI and retain smart, flexible honest people in our newsrooms to make our content infinitely better… the competition is about to get insane.
So, if they are staying on the payroll, what should we teach our journalists? When the world keeps moving, what can we ask them to latch on to? We don’t want to teach journalists how to use specific tools, but how to find good ones. They should be able to vet the tools and incorporate them into their workflows. It is not possible to do the tool integration “seamlessly” even though that is what we all want. In a few months (or weeks) the tools will change, everything is going to keep changing and our workflow platforms are going to go out of date. We need to adopt a life of flexibility. And there is going to be a great deal of tool duplication if we are not careful. In our WhatsApp Community we have asked the journalists to appeal to each other to test their apps and eventually become each other’s first customers.
I think journalists are going to be forced to break down their processes into tiny details in order to investigate which parts can be automated. However, there needs to be a trust between the journalists and the bosses to say that no one is going to be retrenched once the journalist reveals their tasks and brainstorms how to automate them. We don’t want journalists to feel like they are being asked to design their own electric chair.
What would be fantastic is an ongoing dialogue between what is being built by an organisation’s tech department and what is needed by the newsroom. Also, there needs to be a way to track how AI is improving the final product. This is key and not something people are talking about too much: you want a company’s internal metrics to judge how the product is getting better through the AI adoption process.
The idea of a policy for your media business is a good one, but you need to ask why you need one and how that folds into the training of your staff and products you might want to adopt and build. It also needs to be produced with input from your journalists, because they want the product to be ethical more than anyone. And your policy will change (probably soon) and that’s okay.
When it comes to disinformation my prediction is that for the entirely fabricated stuff - fake stories, images and videos - people are going to choose to subscribe to something that will block all that out (like how we dealt with spam decades ago). But the content that is incorrect that people are actively seeking out (and won’t believe is wrong no matter what you tell them) is about to become more rampant.
Lastly, what will this do to investigative journalism? The outlook is largely positive. People are seeing an opportunity to find insights and stories in huge data sets. I feel for this type of journalism there could be a larger problem in terms of checking what the AI spits out. Because in pure narrative writing a journalist can pretty easily spot where the AI has gone off script, but if a tiny newsroom gives it a massive dataset then they might not have the manpower to check what it serves up and if asked to “show their working” they might not be able to do so.
Develop AI helps newsrooms (no matter what their size) with their AI strategies. If you are interested, get in touch.
This week’s AI tool for journalists to use…
In this article for One World Media, Ruona Meyer, a superb journalist and good friend, gives a treasure trove of insights around AI tools for journalists. She is joined by Prof. Charlie Beckett, founding director of Polis. One recommendation from Meyer I found very helpful was GPTZero. It’s an AI-detection tool. It is trained to detect ChatGPT, GPT4, Bard, LLaMa, and other AI models. Let the arms race begin!
Coding Corner (the gradual process of a journalist learning how to code)
A huge thanks to The Podcast Sessions for featuring our AI podcast creation app. I’ve been thinking about what is next for the tool. I want to feed the transcripts for each episode into a database and then when the prompt creates a new script I want it to access this database and let that inform what the hosts say. At the moment there is no history or chemistry between the hosts and that is because each new script is created in a vacuum. The changes between episodes come from the different content of the new URL and the whims of the updated ChatGPT model. But if they started to have a HISTORY and comment on what they had said in the past, well that would definitely give the impression of them being more human.
You can listen to our entirely synthetic podcast (based on this newsletter) on Spotify.
More from our coding blog.
What AI was used in creating this newsletter?
Very little besides the image from Bing as usual. I have to admit I tried putting all my conference notes into ChatGPT to write the main section above and the TONE was incredibly grating. I didn’t use anything that it gave me. So, I just wrote it the slow way like a loser.
In the news…
I don’t think we are close to experiencing peak AI media paranoia, but last week saw a flurry of articles and videos about AI generated girlfriends (and how they are destroying young men). The most innovative version is one influencer created an AI bot of herself named Caryn. She gained 1 000 boyfriends in a week with a waiting list of 15 000 more. It feels like every business is trying to create a personalised “assistant” for their customers but there is a dimension of hysteria from general news outlets if these bots are used to talk about sex and emotions.
What’s new at Develop AI? Jamfest in Johannesburg, Innovation Lab in Windhoek & The Egypt Media Forum in Cairo
I am having a brilliant time at Jamfest in Johannesburg. I’ll be doing a workshop today (Tuesday) on AI in Journalism. And I will also be part of a session in the afternoon talking about the fabulous MethodKit For Podcasts.
I would like to congratulate the incredible organisations from Zimbabwe, Botswana, Namibia, Malawi and Lesotho that will be in Windhoek for the Innovation Lab next week. They will be working on AI solutions for African media that are destined to have a big impact. Organised by DW Akademie, I’ll be joining a team of amazing people to help guide these projects to fruition. A bunch of the people from the various teams are part of our WhatsApp Community, you can join now and ask them about their work.
The Egypt Media Forum is a massive conference celebrating media in Africa. I’ll be there representing Develop AI late in November talking about the future of media.
See you next week. All the best,
Join our WhatsApp Community here, visit our website, or contact us on X, Threads, LinkedIn or Instagram. Physically we are based in Cape Town, South Africa.