AI Semiconductor That Reads Fast and Slow Signals Together Unveiled

According to AI TIMES, KAIST researchers said on Sept. 29 that they had developed an AI semiconductor technology that stacks semiconductor devices with different response speeds in multiple layers to read both newly arrived changes and the flow of earlier changes at the same time. In model validation that reflected measurements from actual devices, it identified video signals with different speeds with more than 90% accuracy. The core of the research is that it turns the slow movement of ions from a weakness into a device that reads traces of previous signals. The researchers made ion-containing materials into hard, thin films to implement devices with different response times, then layered them so they could separately process short changes and the changes that follow. The approach is suited to handling motions and body signals that need to be distinguished over time, such as walking and a brief arm swing, inside small devices. However, the media said that performance when applied to actual wearable devices and the effect on power savings still need additional verification.