The silicon giant Intel has been working on the idea of comparing machines with the human brain. As a result, Intel Labs has developed the first self-learning and neuromorphic chip called Loihi.
Loihi’s digital circuits are designed to mimic the brain functions and learn how to operate using feedback from the surrounding environment. The chip uses data to learn, make inferences, and improve its intelligence over the time–one doesn’t need to train Loihi traditionally.
By trying to work like the brain, Loihi makes machine learning faster and more efficient. By observing spikes and plastic synapses that can be modulated based on timing, it learns how neurons communicate and learn.
The latest Loihi test chip, which is still under testing, offers flexible on-chip learning. The researchers have also demonstrated learning at a 1 million times improved rate compared to other typical spiking neural nets. It’s few highlights are:
- Based on Intel’s 14nm process
- Total 130,000 neurons and 130 million synapses
- Each neuromorphic core has a learning engine
- Fully asynchronous neuromorphic many core mesh
The potential applications of Intel Loihi chip would be in the automotive, industrial, and personal robotics field. There, an application would be able to learn continuously in an unstructured environment. Moreover, Loihi is also 1,000 times more energy efficient when compared to normal computing needed for typical training systems.
Did you find this AI advancement in the form of Intel Loihi exciting? Don’t forget to share your views with us.