This New Yorker article explores the development and application of BakeryScan, an artificial intelligence system initially designed to automate the checkout process in Japanese bakeries. Originating from the founder Hisashi Kambe‘s interest in computers, BakeryScan was created to address the challenge of identifying various pastries without barcodes in bakeries offering a wide range of products.
This article traces the history of computer vision, emphasizing the breakthrough in 2012 with AlexNet, a neural network that utilized deep learning for image recognition. While deep learning has become popular, it requires large amounts of specialized data, making it impractical for certain tasks without extensive training datasets. BakeryScan, developed by Kambe and his team, took a different approach, using a hand-engineered system that involved a meticulous process of photographing, analyzing, and categorizing pastries based on visual features.
BakeryScan initially aimed to streamline the checkout process in bakeries with a vast array of pastry varieties. The system went through years of development, incorporating various algorithms to achieve a high level of accuracy under diverse bakery conditions.
Then an unexpected application of BakeryScan’s technology to medical research arose. A doctor at the Louis Pasteur Center for Medical Research recognized the potential of using BakeryScan for identifying and measuring cancer cells under a microscope. BRAIN adapted BakeryScan to this medical domain, leading to the creation of Cyto-Aiscan, a system capable of analyzing entire microscope slides for cancerous cells.
BakeryScan’s core technology, now known as AI-Scan, has been further applied in various fields beyond bakeries, including pill identification, art analysis, and quality control in manufacturing.
To conclude, this article points to the uniqueness of BakeryScan’s development approach, contrasting it with the dominant trend of deep learning in modern AI, and how it continues to evolve, addressing challenges such as the impact of COVID-19 on pastry shop practices.
