Despite living in a world of endless content creation, I still like looking for new avenues for creating content with a bit of genuine value. This is the challenge I set for myself when I decided to carry out my first real project on this site: a set of articles exploring the potential of generative AI in creating recipes. The core idea was simple and personal – to create recipes that would inspire an answer to the question of what to have to eat.
I’m not a food enthusiast. But even for those who are, the task of deciding what to eat each night can be more a chore than a pleasure, especially when you want to make a decision with others. My partner and I often end up messaging each other on WhatsApp with an all-too-familiar question: “What shall we have for dinner tonight?”
My partner, who is much more of a foodie than me, is also sometimes left looking for inspiration. Especially when heading back from work and dashing to the supermarket. So, I thought, why not use generative AI to create articles for us, ones that would include a recipe and a few other meal suggestions. I know a few people who were already using things like ChatGPT to think of what to cook next. Why not do the same, but with a couple of extra steps?
So, I’ve created a few recipe article generators, each involving the creation of a customised GPT on ChatGPT. The first focused on Italian cuisine, drawing inspiration from an old recipe book. I gave the second a set of historic texts, with the aim of enabling it to draw inspiration from ancient and medieval times. I’ve also given it some recipe books from modern history. To these two, I’ve added a GPT that focuses on for quick yet tasty meals, and another for comfort food that didn’t compromise too much on health. Since creating them, I’ve been generating articles for this site each day.
Still, this project wasn’t just about using AI to churn out recipes and ideas for a bit of inspiration. I also wanted to see what would happen if I kept on asking for new recipes. Would it keep on repeating the same old thing or really draw something out from the texts I’d provided the tools with. As I started in November, and I’ve asked the AI cooks to focus on seasonal dishes, the four have produced a lot of recipes or suggestions relating to pumpkins. However, I’ve been happier than I expected to be about the results so far.
I must admit to having had some concerns about what the AI-generated recipes might suggest or include. For instance, The Guardian reported a notable case involving a New Zealand supermarket’s AI recipe bot suggesting concoctions like chlorine gas cocktails and bleach rice. Lovely. Perhaps it shouldn’t come as a surprise that I wanted to keep a level of human oversight as part of the process. I have read through each recipe, ensuring that the AI’s suggestions were not adding ingredients that you wouldn’t want to eat. So far, there’s been so suspect ingredients.
This project remains a personal experiment – an exploration of what generative AI can bring to our daily shopping list. It’s not about creating content for the world; it’s about creating something of value for two people navigating the daily routine of planning the next meal.
And so, the experiment goes on, one AI-suggested recipe at a time.
Here’s a selection of some of the recent articles:
