The Land of Analog, Japan's 'AI Genius' [AI Error Note]

Another Godfather of AI, Kunihiko Fukushima
Japan Sheds Its Image as a "Digital Laggard" to Focus on AI
A Clear Contribution to the Modern Explosive AI Revolution

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

The wave of artificial intelligence (AI) is sweeping even the Nobel Prizes. Following last year's Nobel Prize in Physics, the Chemistry Prize was also awarded to figures in the AI field. Out of the three Nobel Prizes in science, AI took two, excluding the Physiology or Medicine Prize.

The Physics Prize was awarded to researchers who established the foundation of AI machine learning using artificial neural networks. They are Professor John Hopfield of Princeton University, USA, and Professor Geoffrey Hinton of the University of Toronto, Canada. They were recognized for their achievements in developing algorithms such as machine learning, which are the basis of modern AI.

The Chemistry Prize went to those who contributed to protein structure prediction and design using AI. The joint winners were Demis Hassabis, CEO of Google DeepMind, researcher Dr. John Jumper, and Professor David Baker of the University of Washington, USA. The name DeepMind will be especially familiar. It is the developer of the AI Go program 'AlphaGo,' which played against 9-dan player Lee Sedol in 2016.

Until now, the Nobel Prize has strongly recognized human creativity and scientific discovery. Considering this, the fact that AI researchers are listed among Nobel laureates is evidence that AI has deeply entered human life.

The Era of 'AI Star Scholars'... Another Name to Remember: Kunihiko Fukushima

The Land of Analog, Japan's 'AI Genius' [AI Error Note]

Kunihiko Fukushima. Franklin Institute website

As AI enters everyday life, the names of AI researchers are becoming widely known. The media has even coined the nickname 'The Four Giants of AI.'

Including Professor Hinton, who won the Nobel Prize this time, they are Professor Yoshua Bengio of Montreal, Yann LeCun, Meta's Chief AI Scientist and New York University professor, and Andrew Ng of Stanford University, USA. Except for Professor Hinton, the other three jointly received the Turing Award in 2018, known as the 'Nobel Prize of Computer Science.'

As the AI revolution accelerates and its influence expands, they will be mentioned more frequently and more often. However, it is necessary to remember a person who devoted himself to AI research before them and laid the foundation for today's AI revolution but is not widely known: Kunihiko Fukushima.

As you might guess from his name, he is Japanese. If you hold the stereotype that 'Japan still uses fax machines and prefers stamps and cash,' you might be especially surprised. He was another AI pioneer born in Japan.

The Boy Who Played with Industrial Waste as Toys

A fictional image of Kunihiko Fukushima's childhood generated by ChatGPT (DALL·E 3).

Fukushima was born in 1936 in Taiwan. At that time, Taiwan was Japanese territory. He grew up there until Japan's defeat in World War II. With the end of the war, he fled back to Japan with his family. Having left all their property behind, young Fukushima had no proper toys. Occasionally, small transformers or electric motors given by his uncle served as playthings. However, electricity and wires fueled Fukushima's imagination.

After earning a Ph.D. in Electrical Engineering from Kyoto University in 1966, Fukushima devoted himself to research. In 1979, he announced an innovative neural network architecture called the 'Neocognitron.' Simply put, the Neocognitron is a system modeled after how humans recognize objects.

For example, imagine a red apple in front of us. First, we see simple lines and shapes, then recognize the round form, and finally identify it as 'an apple.'

The brain's complex neural network recognizes images, compares similar sizes, shapes, and past experiences to classify and recognize. We recognize an apple regardless of its size, color, or position. This is why we can distinguish dogs from cats and lions from tigers.

‘Neocognitron’ as the Foundation of Convolutional Neural Networks (CNN)

The brain recognizes images after classifying them step by step. It identifies multiple layers of features such as size, color, and shape, and integrates past memories and experiences. The image is a capture from a video by the My Franklin Association explaining the operation of the Neocognitron.

The Neocognitron became the prototype of Convolutional Neural Networks (CNN), a core technology in today's computer vision. Fukushima's work connecting electrical engineering and neuroscience laid the foundation for various practical applications, from facial recognition and cancer diagnosis to flood prediction.

The Neocognitron did not receive much acclaim at the time. It was the era known as the 'AI winter.' Most AI researchers were focused on developing rule-based systems and logical reasoning. Neural network research was considered impractical and unfeasible. Papers with 'neural network' in the title were automatically rejected. Even published papers received little attention.

What clipped Fukushima's wings most was the hardware technology level of the time. Training large-scale neural networks like the Neocognitron efficiently required substantial computing resources. Massive amounts of data and computational power were needed, but neither was easily accessible then.

The then state-of-the-art Intel 8086 processor operated at only 4.77 to 10 MHz, with memory measured in kilobytes. It was far slower than the computing power of today's smartphones we carry in our pockets. Processing complex neural network computations could take days or even weeks.

His research was arguably ahead of its time. The Neocognitron's potential was proven and its value shone long after he completed his research.

Fukushima's case shows that technological revolutions cannot be achieved by individual genius or effort alone. Two elements are necessary for a true technological revolution. First is an innovative idea, such as Fukushima's Neocognitron. Second is hardware technology capable of implementing it. The explosive increase in computing power opened possibilities to experiment with and utilize innovative ideas.

Nevertheless, "I am a Lucky Man"

Since falling behind in the IT industry after the 1990s, Japan has recently revived as a new powerhouse in the AI era by attracting large-scale data centers and global big tech AI research hubs. OpenAI, Microsoft, and Google have chosen Japan as an AI testbed. An image virtually created of Japan's future using ChatGPT (DALL·E 3).

There is more to learn from Fukushima than AI alone. He did not become discouraged or resigned because his research was not recognized at the time.

On the contrary, he said he was 'a lucky man' just to have had the environment to conduct such research. In the 1970s, when Fukushima devoted himself to research, Japan was experiencing rapid economic growth. It was able to invest heavily in new science and technology. The research institute he belonged to was the NHK Research Institute. You are probably familiar with NHK. Yes, it is Japan's broadcasting company. It was a time when broadcasters had enough money to sponsor scientific research. Fukushima said in a media interview, "It was a great fortune to be able to focus solely on neural network research at the NHK Research Institute."

His pure passion for research, unwavering conviction, and humble acceptance of all circumstances show us what a true researcher looks like.

The benefits of the AI revolution we enjoy today were made possible by the dedication of pioneers like Fukushima who were ahead of their time. Sometimes their pioneering insights may not receive proper recognition in their era. But unextinguishable passion eventually shines in history. Let us remember Kunihiko Fukushima, the unsung hero of today's AI revolution, his name and achievements.