Mission

Tuna ― one of the ocean’s greatest treasures. The numbers of tuna have dwindled in recent years due to overfishing. Because tuna is generally traded by weight, fishing practices that sacrifice quality in favor of sheer quantity have become the norm, contributing to the overfishing problem the world faces. Overfishing practices damage fish quality, but this issue often goes unidentified. So we created a new technology that could analyze the quality of each and every tuna, with the goal of shifting the market’s focus from quantity to quality. Our aim is to share high quality tuna with the world.

Strategy / Idea

We created TUNA SCOPE, an AI system built to be the successor to Japan’s artisanal legacy of tuna examination. We discovered that all of the vital data was to be found in the cross-section of the tuna’s tail. In markets in Japan, a handful of skilled artisans would spend at least 10 years mastering the ability to determine the quality of each fish by analyzing these cross-sections, and then go on to trade them.  We took a over 5,000 cross-sectional data and expert analyses. The AI had mastered the unexplainable nuances of artisan’s’ skills through deep learning, thereby successfully inheriting the legacy, and Japan's skillful inspection has become available everywhere.

Results & Achievement

The system was introduced into fish factories in Yaizu, Japan. The top ranked tuna was branded as “AI Tuna” and sold at a sushi restaurant in Tokyo. Beginning with Dalian in China and Misaki in Japan, the AI is spreading across the world. Tuna quality-certified by TUNA SCOPE has been launched in Tokyo, NewYork and Singapore, and this will allow people worldwide to taste high quality tuna. The business has been adopted by the Promotion Program for Seafood supported by the Government of Japan. The project became a popular story in over 1,550 media outlets in 55 countries. This could help shift the industry to one that emphasizes quality over quantity. This AI has the potential to help protect the ocean’s precious resources. With it, we can ensure a stable supply of high quality tuna is always available.

Additional Information

The nuances of the skills of tuna examination as it has been carried out for years in Japan were part of an unspoken wealth of knowledge that was difficult to express in words or teach to others. We used deep learning to process an enormous amount of data, effectively compressing the amount of time that is usually necessary to master this craft. The AI examination was measured as 90% accurate compared to human examiners with 35 years of experience. By harnessing AI’s intrinsically creative aspects, we were able transplant an intricate and precious store of human knowledge into a digital system, where it can it be preserved forever. We are also currently considering further applications of the algorithms used in this technology, such as medicine, agriculture and forestry.