What artificial butter tells us about Artificial Intelligence

Last year, I finished my PhD at the university of Regensburg. My dissertation, published this month, examines in detail artificial butter in the early 20th century. Some might argue that the topic is remote from Artificial Intelligence. I beg to differ.

Exploitation. A cohort of white, male entrepreneurs established networks across the Global South to appropriate the labor of millions of Black and Brown workers. In Europe and the United States, they developed cutting-edge technologies to turn the fruit of this labor into an ersatz product indistinguishable from the original. Some critics denounced it as the end of culture.

Sounds familiar? The parallels between artificial butter in the 1900s and Artificial Intelligence in the 2020s are striking. Back then, people with no alternatives cultivated oil palms, peanuts and coconut trees. Thanks to advances in organic chemistry, factories almost magically turned these raw materials into a product hardly distinguishable from butter – later to be called margarine. Today, in the very same countries, people with few other options produce text or label pictures. Thanks to advances in computer science, this training data is used to almost magically output text and images hardly distinguishable from human-made artifacts.

Alternative. The history of the so-called “artificial butter” has been largely forgotten today. A century ago, it provided millions with a cheap and reliable source of fat. At the time, almost everyone would have preferred to eat proper butter made from cow milk. Artificial butter wasn’t superior; it was simply cheaper.

Today, millions use chatbots for tasks that LLMs are ill-equipped to handle. Recent surveys have shown that British teens turn to them for psychological support, while young Polish women rely on ChatGPT for gynecological advice. They do so not out of preference, but necessity: psychologists or gynecologists are often too far away or too expensive, or are sometimes seen as untrustworthy. Then, as now, the artificial substitute is rarely the first choice – merely the least-worst alternative.

Artificiality everywhere. A century ago, the situation was strikingly similar. Consumers bought artificial butter by the millions of tons, yet journalists and politicians defended butter as a mark of culture and a national symbol. Entrepreneurs offered machines that supposedly detected artificial butter – though they did not – and losing the fight against it was said to threaten the moral standing of the nation.

Many students today probably understand that using a chatbot to complete an assignment teaches them little. But they also recognize – partly correctly - that universities function more as credentialing institutions than as places of learning. Likewise, many users know that LLMs are bullshit generators, but they also know that bullshit is often precisely what is expected of them.

Difference. The difference between AI and AB lies in the political response they elicited. When governments realized that people would not voluntarily renounce artificial butter, they followed the advice of intellectuals and passed harsh measures on the product. Sales were practically forbidden in some countries – France, for example, from 1897. Everywhere, the name “artificial butter” was banned, and the stuff was to be called “margarine.”

Given that the labels “artificial meat” and “artificial milk” were recently prohibited for their plant-based counterparts, it seems only logical that “artificial intelligence” also warrants a renaming. I propose “margAIrne.”

This is an excerpt from the Automated Society newsletter, a bi-weekly round up of news in automated decision-making in Europe. Subscribe here.