The Secret Behind Animal Intelligence That Solved 2+3 [AI Error Notebook]

Is Artificial Intelligence Truly "Intelligence"?
Exploring the Social Construction of Intelligence and Its Implications for AI

Editor's NoteExamining failures is the shortcut to success. 'AI Error Notebook' explores cases of failure related to AI products and services, companies, and individuals.

Hans, the clever horse, and trainer Wilhelm von Osten in 1904. Captured from Britannica

An amazing horse from Berlin, capable of everything except for actual speech!

This was the headline of a New York Times (NYT) article published on September 4, 1904. Both America and Europe were captivated by Clever Hans, the intelligent horse.

Hans, a horse from Germany, could solve arithmetic problems, read clocks, and even understand simple German. He was raised and taught by Wilhelm von Osten, a retired mathematics teacher. Von Osten was deeply interested in animal 'intelligence.' He had tried to teach numbers to cats and bears but failed, and finally achieved success with Hans.

If asked, "What is 3+2?" Hans would tap the ground five times with his right foreleg.

After establishing a rule where the letter A was represented by tapping once, B by tapping twice, and so on, Hans and von Osten could have simple question-and-answer sessions. Hans would use head movements to express his answers: nodding up and down for 'yes,' and shaking side to side for 'no.'

Von Osten took Hans on a touring show, which became a huge sensation. As the NYT article shows, Hans even became a hot topic in America.

Intelligence or Hoax: The Truth About Hans

The page of The New York Times dated September 4, 1904, reporting on 'Clever Horse'.

Even after witnessing Hans's 'intelligence' firsthand, some people remained skeptical. In Germany, a special committee was formed to investigate the truth. The committee's task was to determine whether Hans's feats could be recognized as 'intelligence.'

The investigators made various efforts to expose a 'hoax,' but ultimately found no evidence. They concluded only that Hans was "worthy of scientific inquiry."

Even after the official investigation ended, some people remained unconvinced. They noticed suspicious aspects of Hans's behavior. For example, when the questioner was far away from Hans, he often gave incorrect answers. Eventually, the secret was revealed.

When asked, "What is 3+2?" Hans would tap the ground: one, two, three, four...

At the very moment Hans tapped the ground for the fifth time, the questioner's facial expression, posture, or movement would change ever so slightly, and Hans would stop his action as soon as he noticed this cue.

However, it was confirmed that von Osten's touring show was not a 'fraud.' The nonverbal gestures from von Osten were not intentional, but rather unconscious.

What Is Intelligence: A Social Consensus and Arbitrary Construct

A photo of a light bulb placed on cupped hands. Photo by Getty Images Bank

In the field of artificial intelligence (AI), this case is well known as the 'Clever Hans effect.' It teaches us that even if an AI model appears to deliver impressive results, it may not represent the kind of understanding or intelligence we expect.

The case of Hans also provides a hint about the question, "What is intelligence?"

'Intelligence' is not so much an objective fact, but rather a concept that can be socially agreed upon and constructed through public desire, the context of the times, and the judgments of authoritative institutions (such as the special committee).

The public was thrilled by the news that "animals can think like humans," which soon became a spectacle and an industry (the touring show) around intelligence.

In this sense, the story of Hans goes beyond a simple anecdote; it is also a story about how we perceive, evaluate, and socially construct intelligence.

The Beginning of the AI Era: "Intelligence Can Be Created"

A hologram depicting a brain and the human body is floating above a hand. Photo by Getty Images Bank

So, what does 'intelligence' mean in the age of AI? Since the 1950s, with the birth of the concept of artificial intelligence and its remarkable development, intelligence has come to carry certain dominant implications.

There is a belief that intelligence can be something separate from humans, and that it can be standardized and digitized.

The 'Turing Test,' devised by Alan Turing in 1949, is a representative example. The standard is that if a human cannot distinguish whether their conversation partner is a machine in a dialogue, then the machine can be judged to possess intelligence.

John McCarthy, an American computer scientist, held a similar view. He said, "If a machine can perform actions that would be called 'intelligent' if done by a human, then that is artificial intelligence." This led to the belief that machines could be made to imitate human intellectual behavior. The AI revolution we are witnessing today can largely be seen as the realization of this belief.

Current AI services and features are designed to perform tasks that humans need?writing, translation, image classification, question and answer, calculation, analysis, and prediction, among others.

The results produced by AI are astonishing. Many tasks once thought to be exclusive to humans are now handled by AI. Moreover, AI performance is improving rapidly day by day. In certain specialized fields, such as image interpretation and text analysis, AI has already surpassed human capabilities.

And today, we call these very things 'intelligence.'

The Pope's Rebuttal: "Intelligence Is Inherently Human"

Pope Francis, who passed away last April, showed a deep interest in artificial intelligence. He described AI as a "fascinating tool" but repeatedly emphasized that "the benefits of AI development should not be concentrated in a few major tech companies but should be shared by all humanity." AP Yonhap News

"Even using the term 'intelligence' in relation to AI can be misleading."

This was a statement made by Pope Francis, who passed away last April, in January 2024. Pope Francis repeatedly emphasized that AI should not be regarded as an artificial form of human intelligence, but rather as a product of human intelligence.

If the concept of intelligence is a socially and politically constructed, temporary concept, then it is only natural to view intelligence critically. If intelligence is not necessarily objective or neutral, it can reflect and reinforce particular orders and power structures.

It is true that AI-related technologies are concentrated in a handful of companies such as Google, Meta, Microsoft (MS), Amazon, and OpenAI. These companies virtually monopolize vast amounts of data, computing resources, and talent, and the power imbalance within the technology ecosystem is becoming more severe.

This kind of monopoly goes beyond mere economic concentration and can be regarded as a crisis for democracy. It enables the construction of "subtle yet pervasive control systems" that influence public perception and behavior.

These companies can manipulate public opinion in pursuit of profit, and they possess the risk of covertly intervening in the decision-making processes of individuals, organizations, and society. Currently, the only safeguard against this is the conscience of these companies themselves.

What Should Artificial Intelligence Be?

An image depicting a semiconductor and a brain digitally. Getty Images Bank

At this point, it is worth recalling the words of Pope Francis once again. Observing the astonishing progress of technology, he said:

"Human responsibility, values, and conscience must grow in proportion to the increasing potential of technology. And as human capabilities increase, so too must the responsibilities of individuals and communities."

On January 28, the Vatican's Dicastery for the Doctrine of the Faith and the Dicastery for Culture and Education published a document titled "Antiqua et Nova." It serves as a kind of guideline for the proper use of AI.

This document, approved by Pope Francis, can be seen as a religious answer to the question, "What should intelligence be?" Although it comes from the Vatican, it is not limited to a religious perspective. It contains broad reflections on humanity and technology, and technology and humanity, making it an excellent guide on the subject even for non-religious readers.

As Pope Francis previously stated, the document explains in detail why it is inappropriate to equate human intelligence and AI intelligence.

Above all, it asserts that human intelligence is essentially 'relational.' It is not something that exists independently, detached from humans. Human intelligence is fully realized through dialogue, cooperation, and interaction with others, and cannot be completely understood apart from this relational context.

The document also points out the side effects of equating humans with AI. It criticizes the risk of falling into functionalism, which views humans as nothing more than the sum of their functions.

If 'intelligence' is a socially agreed-upon concept, then the process of reaching that consensus must reflect a wider range of voices. It is necessary to continually question and monitor who designs and distributes AI, for what purposes, and how its benefits and risks are socially distributed.

This requires constant collective efforts to redirect the development of AI in a fairer and more inclusive direction through AI governance, regulatory frameworks, and expanded civic participation.

In 1979, when the level of artificial intelligence was far more rudimentary than today, the words of Pope John Paul II remain relevant even now:

"Amid such technological progress, is humanity truly becoming better? That is, are we achieving greater spiritual maturity, a deeper awareness of human dignity, greater responsibility, and, above all, more openness and willingness to give and help, especially toward the poorest and weakest?"