AI and Science of Reasoning(1)

Sherlockian Way of Thinking

1. Birth and Evolution of Artificial Intelligence

We are now stepping into an age in which Artificial Intelligence (AI) has fully come of age. Generative AI, in particular, has advanced to the point where it can understand human language, create images, and even compose music with remarkable fluency. Yet the true key to harnessing this power does not lie in the machinery itself, but in how we think and, above all, in how we ask questions. This practice is known as prompt engineering. In this chapter, I will explore the deep resonance between Sherlock Holmes’s methods of logical reasoning and the art of prompt design.

The true key to harnessing the power of AI lies in above all how we ask questions.

Before doing so, however, it is worth retracing the birth and evolution of AI itself. The field, born from Alan Turing’s provocative question, passed through two long winters and then a quiet resurrection, until it blossomed into today’s age of generative AI. This journey was never merely a technological march forward; it was, rather, a continual inquiry into the very nature of human thought.

Once this foundation is laid, we shall turn to Holmes’s six celebrated methods of reasoning—deduction, induction, the hypothetico-deductive method, abduction, retroduction, and elimination—and see how each can inform the craft of designing prompts. The Sherlockian Way of Thinking is not a matter of theory alone; it offers concrete guidance on how to structure a prompt, how to break down a complex problem step by step, and how to sharpen the accuracy of an answer.

Finally, the reader will be invited to practice Holmes’s art of thought directly: framing precise questions, stipulating the necessary conditions, and even sketching the structure of the desired answer. For crafting a good question is not a talent bestowed by nature—it is a discipline to be cultivated through training.

“Can machines think?” — Turing’s Question

There once lived a brilliant mathematician named Alan Turing (1912–1954) in England.

In England there once lived a brilliant mathematician named Alan Turing (1912–1954). During the Second World War, he achieved lasting fame by breaking the German cipher system known as Enigma, an achievement credited with shortening the war by years and saving countless lives. His life and work were later dramatized in the film The Imitation Game (2014).

Yet Turing’s true greatness lay not in cryptography alone. He was the first to conceive of what we now call the computer. In the early 1940s, he had already outlined the essential architecture of modern computing in the abstract model he named the Turing Machine. He was, in short, the man who asked not only how machines should function, but whether they might one day think as humans do.

In 1950, in his seminal paper Computing Machinery and Intelligence, Turing posed the disarmingly simple question: “Can machines think?” He immediately noted, however, that the word think was hopelessly ambiguous, subject to countless interpretations. Rather than wrestle with semantics, Turing reframed the challenge in a far more ingenious way: “Instead of asking whether machines can truly think, let us ask whether they can behave as though they were thinking.” Thus was born the Turing Test.

What Is the Turing Test?

Even now, seventy years on, the question lingers: “Has AI passed the Turing Test?”

The principle of the Turing Test is strikingly straightforward:

A human interrogator engages in text-based conversation with two unseen interlocutors, one a machine, the other a human.

On the basis of the conversation alone, the interrogator must decide which is which.

If the interrogator fails to reliably distinguish the machine from the human, then the machine must be credited with “thinking.”

Turing thereby suggested that intelligence should not be judged by unseen inner workings but by outward behavior—by the dialogue itself. A computer that could converse naturally, reason coherently, and answer questions as a human would, deserved to be called intelligent, regardless of metaphysical debates about “real thought.”

This was nothing short of revolutionary. In 1950, most people still believed thought to be a mysterious faculty unique to human beings. And even now, seventy years on, the question lingers: “Has AI passed the Turing Test?” With the rise of systems such as GPT-4, the debate has only sharpened. Turing’s test remains a touchstone, a benchmark for evaluating not only whether AI thinks like us, but how we should measure such thinking at all.