Key Takeaways
1. FLEXI is a new series of flexible AI chips developed by researchers from Tsinghua University and Peking University, thinner than a human hair.
2. The chips use a compute-in-memory design, allowing on-device data processing that reduces energy consumption to less than 1% of traditional silicon.
3. FLEXI is made from low-temperature polycrystalline silicon on a flexible plastic substrate, enabling it to withstand over 40,000 bends and be folded down to a radius of one millimeter without performance loss.
4. In trials, FLEXI demonstrated high accuracy in medical applications, detecting irregular heartbeats with 99.2% accuracy and monitoring physical activities with 97.4% accuracy.
5. With production costs under $1 per unit, FLEXI has the potential to enable affordable smart textiles and AI-integrated clothing for widespread use.
The future of smart wearables has faced challenges due to inflexible, energy-consuming chips. However, researchers from Tsinghua University and Peking University have developed a solution called FLEXI, a new series of flexible AI chips that are thinner than a single human hair.
A New Approach to Processing
This research, featured in the journal Nature, introduces a major shift in design: compute-in-memory. Unlike traditional wearables that need to send data to external processors or cloud systems, FLEXI can handle data directly on its own circuitry. The researchers claim that this on-device processing reduces the energy required for moving data between memory and processors, enabling the chip to use less than 1% of the energy consumed by regular silicon.
Innovative Physical Design
The design of the chip is also impressive. It is made from low-temperature polycrystalline silicon (LTPS) mounted on a flexible plastic substrate, allowing it to be stretched, twisted, and even crumpled. In tests, FLEXI endured over 40,000 bends and was able to fold down to a radius of just one millimeter without losing performance.
Promising Medical Applications
Trials conducted with volunteers revealed the chip’s potential in medical applications. FLEXI detected irregular heartbeats (arrhythmia) with an accuracy of 99.2% and monitored physical activities such as walking and cycling with a 97.4% accuracy.
FLEXI is also aimed at wide-scale usage. With production costs expected to be under $1 per unit, this innovation could lead to smart textiles and AI-integrated clothing that are currently too expensive. While the research team plans to add more advanced sensors in the future, FLEXI is already a high-quality, low-energy option for the next wave of adaptable electronic devices.
Source:
Link