What Does Artificial Intelligence Tell Us About Cheese and Implicit Bias?

Editor’s note: Along with many of my colleagues I am both concerned about AI and intrigued. Contributor Carlos Yescas spent some time using ChatGPT and discovered some of its many shortcomings when it comes to cheese.

Like most writers and content creators, the news of ChatGPT made me feel uneasy about my part-time very limited writing gig. I heard news on the radio and podcast about this amazing new technology, read articles about the cool future where AI will do a lot of menial tasks, and chatted with friends about the implications for knowledge creation and sharing. 

I first heard about ChatGPT’s capability to write code and its ability to give answers to exam questions. Then, my spouse started playing with the technology and he gave a variety of prompts during a weekday dinner to show me the technology. Our commands included writing poetry, cover letters for jobs, invitations to events, and soon we turned to the topic of cheese. The questions we asked were basic about cheese culture and cheesemaking. It was obvious that the algorithm knew basic information and could assemble good answers to our questions, but what was most surprising was its bias. 

Bias in Cheese

The bias that you are probably thinking is one that exists in the wider food industry and in the cheese community. That European cheesemaking, culture, and products are far superior and more traditional that those produced elsewhere. This bias comes from many issues, including the prevalence of ideas that put European cuisines, agriculture, and culture at the top of an apex of value. These ideas are fueled by white supremacy, based on colonialism, and maintained by capitalism. It’s why most Americans know more about the culture around Alpine style cheeses, than those made in the mountains and valleys of Mexico. This bias is perpetuated because a lot of money goes into maintaining these ideas of value, and that is the second surprising thing about the answers that the bot gives. 

What is ChatGPT?

Shopping for cheese at the market" by Bex.Walton is licensed under CC BY 2.0

ChatGPT was fed information gathered from the Internet, including blogs, social media posts, and other online content. The information used, I later learned, was only what was available online until 2018. The programmers of the technology decided that during the training period and launch, the “bot” would not have access to the internet and could not continue to absorb new information. This means that the information used by the bot for some topics may be out of date. The system has also been trained in a machine learning technique called Reinforcement Learning from Human Feedback, which apparently was done to make the system appear more human in its responses. Time magazine reported that a lot of this training was outsourced by OpenAI, the owner of the technology, to a subcontractor who used programmers in Kenya.

ChatGPT & Cheese

I decided to sign up for an account and have a conversation with the bot about cheese. My first question was easy: “Tell me what you know about cheese?”

The machine responded: “Cheese is a food made from the curdled milk of cows, goats, sheep, and other mammals. The curdling process is typically accomplished using rennet or a combination of bacteria and mold, which separate the curds (the solid part of the milk) from the whey (the liquid part of the milk). The curds are then pressed and aged to create different types of cheese. There are hundreds of types of cheese, each with their own distinct flavor, texture, and appearance. Some common types of cheese include cheddar, mozzarella, blue cheese, and brie. Cheese can be used as an ingredient in cooking, as a topping for crackers and bread, or as a snack on its own.”

Not a bad response for a simple question. Then, I continued asking about types of cheese and part of the response included “In general, the older a cheese is, the stronger and more complex its flavor will be.” While generally true, I am unsure what is to be understood about “strong flavor” and how the technology got this information. It added that “Everyone's taste preferences are different, so it's worth trying different types of cheese at different ages to find what you like best” and to another question it concluded that “Ultimately, the best type of cheese is the one that you enjoy the most.” These two sentences felt formulaic and probably taught during development. 

Cheese sellers copyright 2016 Josh Wand.jpeg

However, what was more interesting was the cheeses it listed as the best types. The list included Parmigiano Reggiano, Comté, Roquefort, Gouda, Cheddar, and Brie. All delicious cheeses, but to me something felt off. The first three cheeses are individual cheeses, made in specific regions, with distinct cultures around them. While Gouda, Cheddar, and Brie are best understood as styles made in many places. This is something that happens a lot in conversations in cheese counters, some cheese names stand in for just one product, while others are used to talk about a whole category of cheese. At this point, I started thinking about the source of this knowledge. 

Of course, the technology would have biases, and European cheeses would be easily identified as the best types, but what I didn’t expect was that responses would follow the patterns of marketing writing. So, I asked the bot: “Considering the list of cheeses you just gave me, what can be considered as the best cheesemaking country?” The bot gave me a list of European countries, but what was most creepy was the last line of the answer. 

“It's worth noting that cheesemaking is a continuous process, and many new cheesemakers from different countries are pushing the boundaries and experimenting with new techniques, aging methods, and ingredients, bringing new and exciting cheeses to the market.”

Looking Beyond Europe 

Here, I knew I was using a technology that was meant to be vague, that would be used to reinforce ideas created by marketing companies and the biases they have. So I asked “Outside of Europe, which countries have an important cheesemaking culture?” The bot quickly listed the USA, Mexico, Canada, Argentina, New Zealand, and Australia and some examples of cheeses from those countries. 

For the US, the bot wrote “The state of Wisconsin is particularly well-known for its cheeses, which include cheddar, Colby, and gouda. California is also known for its artisanal cheeses, many of them made with raw milk.” Here we can see that those advertising monies do pay back. 

Cheddar photo courtesy of Dairy Farmers of Wisconsin tif

The bot finished by telling me: “These are just a few examples, but there are many other countries around the world with a rich cheesemaking culture and history. The cheesemaking tradition is constantly evolving and many new cheesemakers from different countries are pushing the boundaries and experimenting with new techniques, aging methods, and ingredients, bringing new and exciting cheeses to the market.”

I kept asking the bot many questions about sources of cheese knowledge, about resources not in English language, about influential people in cheese, about the source of value for cheese and some cheesemaking cultures over others, about the origins of cheese, and about cheeses of regions that are not obvious to most people. In asking about Africa, the bot gave me examples of cheeses from Egypt, Morocco, South Africa, Ethiopia, Tunisia, and Nigeria. Too bad it didn’t include cheese from Kenya. With all this information, I asked the bot to create an article to run along my writing. You can decide if we are about to be displaced by technology.

I think this technology has good potential, but it is evident to me that if we all keep repeating the same marketing “facts” without pushing for more details, more nuances, better sources, we would end up with very basic information as the common denominator. That won’t be good for the types of artisanal cheese that we want to eat and the types of producers we want to maintain. Generic information created for generic cheeses shouldn’t be the knowledge we share with consumers. This is why original research on cheese culture and science is important. 

Article on cheese Written by ChatGPT 

Fez, Morocco "cheese" by Diana K is licensed under CC BY-NC 2.0

Cheesemaking has a rich history that spans the globe, with some of the oldest known examples of cheese dating back to ancient civilizations. One of the earliest known examples of cheesemaking is from Mesopotamia, where evidence suggests that people were making cheese as early as 4000 BC. The process of cheesemaking in ancient Mesopotamia involved souring milk by leaving it out in the sun, and then straining it through a cloth to separate the curds from the whey. This would produce a type of curd cheese, similar to cottage cheese.

Europe, particularly France, Italy and Switzerland, is known for producing some of the most highly regarded and diverse cheeses in the world. These countries have a long tradition of cheesemaking and have developed distinct methods and techniques that have been passed down through generations.

Outside of Europe, there are also many other countries that have a strong cheesemaking culture. In the Americas, Mexico and Argentina are known for their unique and delicious cheeses. In Africa, traditional cheeses such as smen, a fermented butter, are often used in cooking.

To learn more about cheesemaking and cheese culture, there are a variety of resources available. Websites, books and online classes can provide information and resources for both professionals and enthusiasts. Social media is also a great way to stay informed and connected with the world of cheese, with many influential cheese experts active on platforms such as Instagram and Twitter.

Ultimately, cheesemaking is a craft that has evolved over thousands of years and continues to evolve today, with new techniques and technologies being developed all the time. Whether you're a professional cheesemaker or just a cheese lover, there is always something new to discover and explore in the world of cheese.