{"id":13219,"date":"2024-12-12T22:00:00","date_gmt":"2024-12-12T22:00:00","guid":{"rendered":"https:\/\/modernsciences.org\/staging\/4414\/?p=13219"},"modified":"2024-11-29T07:52:13","modified_gmt":"2024-11-29T07:52:13","slug":"japanese-ai-pioneers-history-shuninchi-amari-kunihiko-fukushima-december-2024","status":"publish","type":"post","link":"https:\/\/modernsciences.org\/staging\/4414\/japanese-ai-pioneers-history-shuninchi-amari-kunihiko-fukushima-december-2024\/","title":{"rendered":"Japanese scientists were pioneers of AI, yet they\u2019re being written out of its history"},"content":{"rendered":"\n
\n\n Hansun Hsiung<\/a>, Durham University<\/a><\/em><\/span>\n\n

The announcement of the artificial intelligence researchers John Hopfield and Geoffrey Hinton<\/a> as this year\u2019s Nobel laureates in physics spurred celebration and consternation<\/a> over the status of AI in science and society. In Japan, however, another feeling dominates: frustration<\/a>. <\/p>\n\n

\u201cJapanese researchers should also have won,\u201d an editorial in the Asahi Shimbun newspaper<\/a> proclaimed. Congratulating Hopfield and Hinton<\/a>, the Japanese Neural Network Society added pointedly: \u201cWe must not forget the role played by pioneer Japanese researchers in erecting the foundations of neural network research.\u201d <\/p>\n\n

Neural networks are at the centre of contemporary AI. They are models for machines to learn independently through structures that, if often only loosely, are inspired by the human brain. <\/p>\n\n

So who are these pioneering Japanese AI researchers?<\/p>\n\n

In 1967, Shun\u2019ichi Amari proposed a method of adaptive pattern classification<\/a>, which enables neural networks to self-adjust the way they categorise patterns, through exposure to repeated training examples. Amari\u2019s research anticipated a similar method known as \u201cbackpropagation,\u201d one of Hinton\u2019s key contributions<\/a> to the field.<\/p>\n\n

In 1972, Amari outlined a learning algorithm<\/a> (a set of rules for carrying out a particular task) that was mathematically equivalent to Hopfield\u2019s 1982 paper<\/a> cited by the Nobel<\/a> on associative memory, which allowed neural networks to recognise patterns despite partial or corrupted inputs. <\/p>\n\n

The North American researchers were working separately to groups in Japan, coming to their conclusions independently.<\/p>\n\n

Later, in 1979, Kunihiko Fukushima created the world\u2019s first multilayer convolutional neural network<\/a>. This technology has been the backbone of the recent boom in deep learning, an AI approach which has given rise to neural networks that learn without supervision, through more complex architectures. If this year\u2019s Nobel was for \u201cfoundational discoveries and inventions that enable machine learning with artificial neural networks,\u201d why not award Amari and Fukushima?<\/p>\n\n

One-sided perspectives<\/h2>\n\n

The AI community itself has been debating this question. There are cogent arguments as to why Hopfield and Hinton better fit the Nobel \u201cphysics\u201d category, and why national balance mattered, given the peace prize went to Japan\u2019s Nihon Hidanky\u014d. <\/p>\n\n

Why, then, should we still be worried?<\/p>\n\n

The answer lies in the risks of historical one-sidededness. Our standard account of artificial neural networks is a North Atlantic-based \u2013 and, overwhelmingly, North American \u2013 history. AI experienced a period of rapid development in the 1950s and 1960s. <\/p>\n\n

\n