Why the Periodic Table creator would have hated AI
Mendeleev taught us the power of thinking deeply to find explanations
Black box approaches are increasingly popular in AI, machine learning, and computational data science. By black box, I mean computational approaches that make predictions without explaining their rationale. For example, the popular chatGPT and the menacing Bing chatbot provide information and amusing conversations, but we don’t know why they choose the responses they choose.
Similarly, deep learning and neural networks are computational tools that make powerful predictions without revealing the mechanism or rationale for their predictions. However, we lose something important when we use these tools — the explanation for why. The story of Mendeleev provides a powerful example of why this matters.
The search for the fundamental elements
For thousands of years, humans sought to understand the substances that make up the world. Philosophers speculated since ancient times about the fundamental composition of matter, but it was difficult to make progress on this problem. Early Greek philosophers proposed that since water is abundant and can change form from solid to liquid to vapor, perhaps water was the fundamental basis of the world.
Other philosophers suggested it could be air that was fundamental, since air forms clouds, hurricanes, and water vapor, and is all around us.
Similar arguments were made for fire and earth.
The philosopher Aristotle integrated these ideas and proposed that these four substances are the four fundamental elements that make up the material world. For thousands of years this idea held sway, and most people believed that the four fundamental elements of matter were indeed water, air, earth, and fire. This mesmerizing hypothesis remains oddly popular in fiction, tv shows, pseudoscientific circles, and in popular culture even today.
The four elements cannot explain all matter
An iconoclastic researcher and physician named Paracelsus began the process that led to the unraveling of this hypothesis and the modern-era search for the fundamental chemical elements, leading to the birth of a new field of science. Paracelsus was the first to use the term chemistry to refer to efforts to make pure versions of natural elements.
Robert Boyle, often viewed as the father of modern chemistry, published his book The Sceptical Chymist in 1661, in which he argued against the prevailing view of the famous four “elements” of air, fire, water, and earth being the basis of all matter. He noted that many experiments showed that there had to be more than four fundamental elements of matter, and that some substances, such as gold, seemed to be pure elements on their own, as they could not be broken down further into these four “substances”.
Boyle argued that the identity of elements had to be determined experimentally rather than by theorizing in the absence of data, with elements being defined as substances that could not be broken down into any simpler substances.
So many elements
As a bewildering number of pure elements were discovered, chemists wondered why there were so many, and how they could be organized. A Russian chemist, Dmitri Mendeleev, was struggling to write a comprehensive chemistry textbook, The Principles of Chemistry. He was trying to summarize what was known about each of the 63 elements that had been discovered, from simple hydrogen to carbon, oxygen, and nitrogen that make up most living matter.
Mendeleev’s mechanistic breakthrough
In thinking about how to present the accumulating data on each of these elements, Mendeleev analyzed the similarities in their properties, so that he could classify them into families of related elements. As he studied the list of the elements in order of increasing atomic weight, he noticed that they had a repeating pattern in their properties, which led him to organize the elements in the form of a table, so that the periodic pattern would be evident in each row and column.
History is a bit confused on what happened next: due either to illness or lacking an appreciation at first for the impact of his organizational scheme, Mendeleev asked his friend Nikolai Menshutkin to present his table for him at a meeting of the Russian Chemical Society. Either way, as Mendeleev elaborated upon his organization of elements in the coming months, he developed a growing appreciation that his periodic presentation allowed him to predict the existence of new elements, including those that became known as aluminum, silicon and boron. When these elements were discovered and had the properties he predicted, the chemistry community embraced Mendeleev’s Periodic Table of the Elements as a profound breakthrough in understanding the building blocks of the material world.
As Richard Morris noted in The Last Sorcerers, “It was Mendeleev who discovered the periodic law, a principle that describes the periodicities that are observed in the properties of the chemical elements. This enabled him to predict the existence of as-yet-undiscovered elements, to predict atomic weights, and to describe their chemical and physical properties as well.” This discovery reverberates today through all of chemistry and biology, and its impact cannot be overstated.
The Periodic Table of the Elements gives us a kind of Netflix rating system for elements, teaching us that helium, neon and argon are unreactive noble gases because of their position in the rightmost column of the table. Or that sodium and lithium make similar salts, because they are found together in the leftmost column of the table.
Could chatGTP or Microsoft Bing have discovered the Periodic Table?
What if Mendeleev had used chatGPT, Microsoft Bing, or a deep learning network to predict the properties of elements? Sure, he would have had fantastic success, but without knowing why. He wouldn’t have unraveled the fundamental reason for the relationships among the elements — their underlying atomic structure. We wouldn’t have the periodic table.
AI is powerful, but we also need mechanisms and explanations to advance our fundamental knowledge of the world.
