If you’ve stumbled upon this article — via Google, social media, or any other channel — the first thing you should know is Big Human is a design-driven company. We make beautiful, highly functional, and accessible digital products. The second thing you should know is the website we’re about to introduce to you isn’t beautiful, and it isn’t at the level of function and accessibility where you’ll typically find Big Human-made products.
That’s because it wasn’t built by Big Humans. Meal Missions is a content hub and food blog. The brainchild of our director of engineering David Hudson, it was constructed via AI and our own auto-blogging technology, Unhuman, which utilizes the power of ChatGPT and Google Trends.
ChatGPT came up with the name, provided the branding, acquired the domain, and chose the tech stack. Unhuman uses Google Trends around a topic, and ChatGPT then generates fresh content (in this case, recipes and corresponding articles). As of this article’s publish date, there are over 1,000 recipes on Meal Missions — a number that continuously increases.
How many LinkedIn posts and thought pieces have you seen about the pros, cons, and “hacks” of ChatGPT and AI in general? Some are interesting and informative; others are stale and quickly became archaic within a matter of weeks. At Big Human, we wanted to go further. Meal Missions is one of our first experiments with this important, evolving technology. It’s a concrete step toward understanding how AI can best help our clients. Our end goal is to arm ourselves with information, giving us the ability to teach others how to utilize the technology and draw lines where the limitations are.
We had a big win on the SEO front: Meal Missions's AI-generated website is currently getting indexed by Google. But we’re not celebrating just yet. Sites like this are also Google’s worst nightmare.
As with any AI-generated writing, there’s the plagiarism concern. At first pass, Meal Missions doesn’t appear to be plagiarizing; the content seems to be original.
It doesn’t take an expert in the field to understand the overall branding and design of Meal Missions isn’t great. At some point along the way, we’ll probably give the name as well as the look and feel of the site a revamp. As of now, ChatGPT and AI web design can’t compete with our team of designers and strategists.
If you have any familiarity with the world of recipes and food blogging, you’ll know photos are paramount. Take one look at the photos featured on the AI-generated website and you can tell something isn’t quite right. We’re working on training a ChatGPT prompt to provide better images.
Finally, there are larger, more complex questions of bias and equity. Is the content on Meal Missions biased? If so, is it caused by a bias in Google Trends or in our human-inputted training data? Our team is exploring this further.
One of Big Human’s tentpoles is our belief in the value of testing. The quality of our execution is based on research and understanding; we test, review, and retest to make sure we’ve built the best product or solution. The same principle applies to Meal Missions. That said, we have many questions, including:
Will Google punish us?
Are the recipes actually viable? (We’re going to test a few in the coming weeks.)
How much will we have to train — and retrain — ChatGPT?
What does a food blogger think of this? (We’re going to ask.)
How much traffic will Meal Missions see on a daily, weekly, and monthly basis?
Over the next few months, we’re going to track the AI-generated website’s day-to-day activity, our findings, and the experiment’s results. Through Unhuman, we’re also going to explore other content hubs and ideas; food blogging is just the beginning.
Our learnings — from Meal Missions and beyond — will live on the Big Human blog. If you’re interested in experimental AI and its impact on design, branding, and technology, be sure to follow along on LinkedIn and Instagram.