The AI Pioneer Who Won a Nobel Prize Was Once a 'Supoja' [AI Error Notes]

Geoffrey Hinton Born into a Family of Scholars
Frustration with Mathematics, Switching to Philosophy, Dropping Out, and Working as a Carpenter
41 Years of Solitary Research... Changing the Course of AI History
How He Overcame the Barrier of Mathematics: It's Okay Not to Be Perfect

Editor's NoteExamining failures is the shortcut to success. 'AI Wrong Answer Notes' explores failure cases related to AI products, services, companies, and individuals.

Artificial Intelligence (AI) may at first glance appear to be the epitome of advanced mathematics. It is often depicted as filled with complex formulas and algorithms. Therefore, it is easy to assume that pioneers in this field must be 'mathematical geniuses.' However, interestingly, did you know that the great figure who sparked today's AI revolution was actually someone who struggled with math? This is the story of Geoffrey Hinton, the 'Godfather of AI.'

Geoffrey Hinton Born into a Family of Scholars

The AI Pioneer Who Won a Nobel Prize Was Once a 'Supoja' [AI Error Notes]

Professor Geoffrey Hinton of the University of Toronto, Canada, who won the Nobel Prize in Physics last year. Photo by AP Yonhap News

Hinton was born in 1947 into a distinguished family of scholars in the United Kingdom.

His great-grandfather, Charles Howard Hinton (1853?1907), was a mathematician and science fiction writer. He was a pioneer in the study of four-dimensional geometry. He left behind several writings aimed at understanding four-dimensional space and also coined the concept of the 'tesseract.' The origin of the tesseract mentioned in today's movies and novels traces back to Hinton's great-grandfather.

The son of his great-grandfather and Hinton's grandfather, George Hinton, was an engineer, and Hinton's father, Howard Hinton, was an entomologist and beetle specialist.

His great-great-grandfather was George Boole (1815?1864). Mathematicians and computer scientists cannot be unfamiliar with this name. He devised the logical system known as 'Boolean algebra.' It allowed complex logic to be simplified using True and False values. This greatly influenced the binary logic system of 0s and 1s, which is the foundation of how computers operate.

This academic tradition in the family likely naturally instilled a scientific interest in young Geoffrey Hinton. In particular, the four-dimensional geometry research of his great-grandfather can be seen as somewhat connected to the high-dimensional mathematical concepts of artificial neural networks that Geoffrey Hinton later developed.

Frustration with Mathematics, Switching to Philosophy, Dropping Out, and Working as a Carpenter

A picture drawn by ChatGPT depicting a college student struggling with math. DALL·E 3

He enrolled in the Department of Physics at the University of Cambridge. However, Geoffrey Hinton's early academic journey was a series of repeated setbacks. Unfortunately, he struggled with mathematics. He thought he lacked mathematical talent. It was so difficult that he even switched his major to philosophy.

While studying philosophy, he contemplated the nature of knowledge and the process of thinking. This laid the foundation for his later research in AI and neuroscience. Nevertheless, he failed to fully settle in philosophy. He changed his major again to psychology. Despite multiple changes, he never found the one thing that suited him. Eventually, he left school altogether. Instead of a pen, he took up tools and worked as a carpenter.

In 1971, with a sense that it was his last chance, Hinton knocked on the door of the University of Edinburgh. He was captivated by Professor Longuet-Higgins' research on perceptrons. Longuet-Higgins studied AI that mimics the human brain and neural networks. Hinton was also interested in understanding how the human brain processes information and learns, and mechanically reproducing this.

Just as he regained interest and had the opportunity to focus on research, things did not go smoothly. His advisor, Longuet-Higgins, changed his research direction.

At that time, Marvin Minsky, regarded as a 'master' in the AI academic community, pointed out the limitations of the perceptron concept and delivered a fatal critique. The overall trust in artificial neural networks and perceptrons in academia plummeted. Many AI researchers abandoned this topic or chose other AI approaches. This period is later described as the 'AI winter.'

Although his advisor left, Hinton did not. He chose once again the lonely path. He focused on mimicking how the human brain learns and processes information.

This was a path far from the mainstream AI research of the time. He immersed himself alone in his interests. Through repeated experiments, he refined his theories. At the end of this 41-year solitary path awaited the 'AI revolution.'

41 Years of Solitary Research... Changing the Course of AI History

The ImageNet competition, which started in 2010, is a contest that focuses on technologies for classifying and recognizing images. Wikipedia

The year 2012 is recorded as a turning point in AI history. Geoffrey Hinton's research team at the University of Toronto won the ImageNet Large Scale Visual Recognition Challenge with overwhelming results.

This competition tested how accurately computers could recognize objects in images. Until then, most participating teams used traditional computer vision techniques. Hinton's team was different. They used 'Deep Learning.'

The artificial neural network developed by Hinton's team using deep learning was called 'AlexNet.' AlexNet's error rate was about 16%. The second-place team recorded 26%. A 10% difference in error rate in this field was revolutionary. Until then, performance improvements were about 1-2% annually, but this result led to several years' worth of progress at once.

This victory meant more than just winning a competition. It was the moment that proved the potential of artificial neural networks, which had been ignored by mainstream academia for decades. At 64 years old, Hinton finally demonstrated that his long-held conviction was correct. This triggered a rapid paradigm shift in AI research worldwide toward deep learning. It was the moment when the persistence of a scientist who had pursued solitary research for over 40 years finally shone.

How He Overcame the Barrier of Mathematics: It's Okay Not to Be Perfect

Professor Geoffrey Hinton. Screenshot from the University of Toronto, Canada website

How did Hinton, who once changed majors several times due to fear and difficulty with mathematics, overcome the obstacle of math? Although he did not fully understand mathematical algorithms, he developed a unique strategy.

'Let's accept the parts that are hard to understand as they are for now. Then move on to the next step!' Sometimes moving forward, sometimes backward, he jumped between steps. Through endless trial and error, he used intuition deeper than anyone else and developed abstract thinking skills.

Thomas Edison, who did not receive formal schooling, is remembered as the 'King of Inventors' through relentless experiments and trial and error. Steve Jobs, who led the computer and digital revolution, was actually far from coding and computer engineering. He had intuition and insight. It seems that a person who achieves great accomplishments in a field does not have to know everything about that field.

Hinton also endured physical hardships. In his youth, he injured his back while helping his mother at home and had to live with pain for life. Sitting was difficult, so when working or attending seminars, he had to stand or lie down. When taking the bus, he had to lie down in the very back seat, and especially boarding airplanes caused him extreme pain.

However, Hinton never lost his optimistic attitude and sense of humor. Above all, his obsession with one subject never wavered. What might have seemed like an outdated research topic to others, he quietly walked his own path.

Hinton's story leaves us an important lesson. To become the best in a field, it is not necessary to perfectly know everything about that field. Sometimes, acknowledging one's limitations and creatively overcoming them is more important. The paradox of a scientist who lacked mathematical talent but sparked a revolution in the mathematically intensive field of artificial intelligence is another great legacy left by Geoffrey Hinton.