Imagine showing someone a box and asking them to guess its contents without any hints. This might seem impossible, yet the box’s nature offers crucial clues. For instance, its size implies the contents are smaller, and the material — metal versus cardboard — hints at what it can hold.

Is there a mathematical way to explain how to make educated guesses based on limited information? Indeed, while outcomes like coin flips or dice rolls are random and unpredictable, many scenarios allow us to optimize our guessing strategies using a few clever tools.

These constrained guesses are essentially estimates, a concept with deep historical roots. A remarkable early example comes from the ancient Greek philosopher Eratosthenes, who resided in Alexandria, Egypt, during the third century BC. Using basic principles, he estimated Earth’s circumference with astonishing accuracy. Though his precise method has been lost, subsequent writings enable us to reconstruct it.

Eratosthenes observed that during noon on the summer solstice, the sun was directly overhead in Syene, causing no shadow in the city’s well. Meanwhile, in Alexandria, a vertical pole cast a shadow of about 7 degrees, or approximately 1/50th of a circle. Knowing the distance between the two cities was 5000 stadia, he estimated Earth’s circumference to be 250,000 stadia.

While Eratosthenes made geometric approximations that can be overlooked, the real challenge lies in determining the length of a stadion — estimated to be around 160 meters. This approximation yields a circumference of approximately 40,000 kilometers, quite close to the modern measurement of 40,075 kilometers. Variations in stadion measurements, ranging from 150 to 210 meters, affect precision, depending on how we interpret Eratosthenes’ work.

The key takeaway is that with simple yet logical calculations, we can deduce significant insights — all without a globe in hand. In the 20th century, physicist Enrico Fermi exemplified this art of estimation, playing a pivotal role in the Manhattan Project which led to the development of the atomic bomb. During the Trinity test, he ingeniously gauged the explosion’s power by dropping small pieces of paper and observing their movements. Though the specifics of his technique remain elusive, his initial estimate of a 10 kiloton bomb was intriguingly close to the accepted yield of 21 kilotons.

Fermi’s knack for educated guesses gave rise to the concept known as the “Fermi problem.” One classic illustration involves estimating the number of piano tuners in Chicago. Starting with a population of around 3 million, estimating the number of households and pianos leads to a rough conclusion of about 150 piano tuners based on several reasonable assumptions.

The crux of this estimation lies in understanding the limits of its imprecision. While we’ve made numerous assumptions during the process, the errors are likely to balance out. An estimate suggesting a million piano tuners would be almost certainly incorrect.

Fermi estimation serves as a valuable tool for generating initial hypotheses, but as we obtain more information, we can refine our guesses. Returning to the box analogy, if a blue ball with the number 32 is drawn from it, our assumption about the contents shifts. Acknowledging that multiple colored balls are likely, we can utilize the statistics pioneered by Thomas Bayes in the 18th century to quantify this uncertainty.

Bayes revolutionized probability by transforming it from a method for understanding randomness into a framework for addressing uncertainty. His Bayes’ theorem offers a way to quantify observations into evidence, comprised of four components: ex ante, evidence, likelihood, and ex post.

Prior values denote fundamental assumptions. Imagine serving three ice cream flavors (chocolate, strawberry, and vanilla) at a gathering. Initially, you might assume each flavor will be equally popular. However, if the first ten guests all choose chocolate, your initial assumption may need reevaluation.

Evaluating the likelihood of ten consecutive chocolate selections under equal preference assumptions reveals a probability of approximately 1 in 60,000—a strong indicator to revise your original beliefs. Such updates provide a more accurate understanding moving forward.

This theorem proves powerful. Referring back to the box example, drawing a colored ball like red ’50’ sharpens the possibilities of what remains inside. Each draw further narrows down our options based on new evidence.

One practical use of Bayes’ theorem appears in spam filters. Early versions used Bayesian inference to categorize a certain percentage of emails as spam (ex ante) and learned to recognize spam emails by examining user-marked emails (evidence) and the likelihood of certain words’ presence in those emails (likely).

This application illustrates how estimation matters in real-world scenarios, far beyond mere mathematics. Especially with modern AI technologies like ChatGPT, understanding and applying Fermi estimation and Bayesian inference techniques is increasingly vital. As observed, AI often seeks to confirm pre-existing information, thus neglecting new data for accurate assessments. Equip yourself with the skills to make informed guesses.