In the future, computers may solve complex problems using learned experiences from the surrounding factors, thanks to Intel, the silicon giant that has developed the first neuromorphic and self-learning chip called Loihi.
Chip that mimics human brain
Inspired by the human brain, the researchers taught Loihi’s digital circuits to imitate the functions of the human brain and to operate using the responses from its surrounding environment. Instead of getting trained, the chip learns using data, makes conjectures, and upgrades its intelligence over the time.
Loihi learns the way neurons communicate and learn by observing plastic synapses and spikes, which can be adjusted based on timing. By doing this, it makes machine learning faster and more productive.
The chip is currently a prototype but offers flexible on-chip learning. Besides, compared to other typical spiking neural nets, Loihi’s learning is at 1 million times improved rate.
Its additional features are:
1. It’s fabricated on Intel’s 14 nm process technology.
2. Each neuromorphic core comprises a programmable learning engine.
3. It comes with asynchronous neuromorphic core mesh that supports a number of neural network topologies.
4. Each chip contains a total of 130,000 neurons and 130 million synapses.
The Loihi chip could be potentially used in the automotive, industrial, and personal robotics field. It could help those applications to conduct autonomous operations and learn continuously in an unstructured environment. Besides, compared to general purpose computing system used for training purpose, Loihi is 1,000 times more energy efficient. Researchers will start sharing this chip with leading research institutions and universities from the first half of 2018, with a focus on advancing AI.