Former STRL Researcher Fukushima Kunihiko Receives the Bower Award and Prize for Achievement in Science from the Franklin Institute

Fukushima Kunihiko*, former researcher at NHK Science & Technology Research Laboratories (STRL), received The Bower Award and Prize for Achievement in Science from the Franklin Institute in the U.S. on April 29. The Bower Award is said to be one of the world’s most prestigious, and receiving it places Dr. Fukushima in the ranks of past winners, who include prominent researchers and Nobel laureates. In 1979, Dr. Fukushima developed a neural network model for pattern recognition, neocognitron, at NHK STRL. His work is highly valued as the invention of the basic structure of deep neural network, a core technology of current artificial intelligence (AI).
On the opportunity of this reward, we asked Dr. Fukushima about the neocognitron and the future of AI.

― “Congratulations on receiving the Bower Award, Dr. Fukushima. At the time you conceived of the neocognitron, did you imagine the application of AI in such a wide range of fields as we see today?”

Fukushima: “No, not at all. Back then, the neocognitron and AI were in two entirely separate fields. The AI processing at that time was based on IF-THEN rules. I was developing the neocognitron because I wanted to build a neural network model on a computer to understand what happens in an actual brain.”

― “Advances in computer technology have made your idea of modeling neural networks of the brain useful in the field of AI. How do you think research on neural networks and AI technology will evolve from now?”

Fukushima: “Seeing with our eyes is visual information processing. I think it also important to investigate the mechanism of information processing for other senses, such as hearing and touch. Human motivation and creativity emerge from some regions of the brain, but very little is known about their neurophysiological mechanisms.”

“It is currently possible to simultaneously observe about 10 neurons, but simultaneous observation of more than 10 billion neurons in the brain and elucidation of their interactions is still a challenge. I think that AI could be a powerful tool for investigating the mechanism of information processing performed by interactions of a large number of neurons. I therefore feel it very attractive if we could make neural network models of these processes. If we apply the results to current issues in AI, AI will evolve into a more advanced stage.”

― “Recently, some people have been saying that AI takes work away from humans. Do you think AI will have the broad capabilities of humans in the not-so-distant future?”

Fukushima: “I think that remains a difficult challenge. AI that is superior to humans in specific domains has already been achieved, but considering the elucidation of the mechanisms of interactions of huge numbers of neurons I mentioned earlier, I have to wonder when AI will exceed human capabilities. It may take centuries, or never happen at all.” (laughs)

― “Your research continues even now. Could you tell us something about the approach to research?”

Fukushima: “It’s important to bring out your own creativity, without simply imitating the currently popular research. In order to get new ideas, it is important to pay attention to the periphery of your subject as well as your own work. At the same time, you have to maintain focus on your own objectives and not just let your mind wander aimlessly.”

― “Congratulations on your award, Dr. Fukushima, and thank you for sharing your valuable insights today.”

Fukushima Kunihiko speaking on the neocognitron and AI
Doing research at STRL (1976)

* Currently a Senior Research Scientist at the Fuzzy Logic Systems Institute