|
Toshiba recently announced they have developed new artificial intelligence (AI) technology that can suppress performance degradation even when used on different devices, such as smartphones and automated guided vehicles (AGV). The calculation volume is adjusted according to the data processing performance of different devices. This is said to enable more efficient AI development, with the goal of practical application by 2023.
For example, AI that detects people from images is used in smartphones, surveillance cameras, AGVs, etc. However, depending on the device, computational capabilities differ, and in the past AI was developed by manually adjusting calculation volume according to the device. With Toshiba's newly developed technology, calculation volume can be adjusted to suppress accuracy degradation.
If calculation volume is reduced, AI performance also decreases. However, in Toshiba's case, when experimenting with image classification, it is said that even when the calculation volume is reduced to one-third, the accuracy only decreased by 2.1%, which is currently at a world-leading level.
