IBM presents ‘brain-like’ chip for more environmentally-friendly AI tools
Image credit: iStock
IBM has developed a prototype “brain-like” chip that could lead to less battery-draining artificial intelligence (AI) semiconductors for smartphones.
IBM Research has unveiled a prototype of an analogue AI chip that demonstrates remarkable efficiency and accuracy in performing complex computations for deep neural networks (DNNs).
The new design aims to address concerns regarding the carbon emissions associated with the large number of computers needed to power AI systems, as well as extend the life of smartphones.
The chip's efficiency is down to components that work in a similar way to connections in human brains, according to IBM.
The 14nm CMOS IC is composed of 64 analogue in-memory computing tiles, each with a 256 x 256 crossbar array of synaptic unit cells. This allows the semiconductor to stores weights locally as analogue levels as conductance in phase-change memory, and implements analogue multiply-accumulate calculation.
Currently, most chips are digital, storing information as 0s and 1s. In contrast, IBM's new chip relies on components called memristors, which are analogue and can store a range of numbers.
This makes the chip similar to the way synapses operate in the human brain, which allows it to achieve remarkable performance while consuming little power.
“We believe this to be the highest level of accuracy of any currently reported chips using similar technology,” said IBM.
Thanos Vasilopoulos, a scientist based at IBM’s research lab in Zurich, Switzerland, said the chip's higher energy efficiency would allow larger and more complex workloads to be executed in low-power or battery-constrained environments, such as cars, mobile phones and cameras.
Additionally, cloud providers could use these chips to reduce energy costs and their carbon footprint.
“These advancements suggest that we may be on the cusp of witnessing the emergence of brain-like chips in the near future,” Ferrante Neri, professor of machine learning and artificial intelligence at the University of Surrey, told the BBC.
However, there are still challenges to be overcome, such as the costs of materials and manufacturing difficulties.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.