There are few structures quite as complex and exquisite as the brain, so it’s little wonder chip designers have used it to inspire their latest creations. In May 2017, R&D hub, Imec, showed the world its first neuromorphic chip. It can learn, create its own music, and even achieves artificial intelligence that’s energy efficient.
The chip is based on OxRAM technology and is light enough on power to be integrated into smart sensors that will give the Internet of Things new life. By drawing parallels between what it observes and has experienced, the chip can learn rules as it creates. It will contribute to health monitoring, traffic control, and energy. The scientists who developed it hope that the new algorithms will push the computing world into neuromorphic territories.
The idea for neuromorphic computing developed in the 80’s, but the cloud has made it a reality. Machine learning depends on access to high volumes of data, which your average CPU or smartphone can’t handle. By creating neuron equivalents that can communicate at any time, neuromorphic chip creators such as IBM have brought energy efficiency down to a meager 70 milliwatts even though their chips contain five times the transistors of a standard processor.
Stanford scientists joined the neuromorphic race in March 2017 when they, too, created a “brainstorm” chip, but it hasn’t yet become commercially viable. The university’s Neurogrid is already being used in neuroscience research.
One of the most challenging aspects of neuromorphic development has been creating algorithms that can process and pick up information. To support his challenge, Nengo has given developers an easy to build AI algorithm. One day, artificially intelligent systems are expected to be a constant part of our lives, and neuromorphics are a titanic step towards that goal.