The chip’s components work in a way similar to synapses in human brains.

Tech corporation IBM has unveiled a new “prototype” of an analog AI chip that works like a human brain and performs complex computations in various deep neural networks (DNN) tasks.

The chip promises more. IBM says the state-of-the-art chip can make artificial intelligence remarkably efficient and less battery-draining for computers and smartphones.

Introducing the chip in a paper published by IBM Research, the company said: “The fully integrated chip features 64 AIMC cores interconnected via an on-chip communication network. It also implements the digital activation functions and additional processing involved in individual convolutional layers and long short-term memory units.”

Reinventing ways in which AI is computed

The new AI chip is developed in IBM’s Albany NanoTech Complex and comprises 64 analog in-memory compute cores. By borrowing key features of how neural networks run in biological brains, IBM explains that it has embedded the chip with compact, time-based analog-to-digital converters in each tile or core to transition between the analog and digital worlds.

Each tile (or core) is also integrated with lightweight digital processing units that perform simple nonlinear neuronal activation functions and scaling operations, explained IBM in a blog published on August 10.

A replacement for current digital chips?

In the future, IBM’s prototype chip could replace the current chips powering heavy AI applications in computers and phones. “A global digital processing unit is integrated into the middle of the chip that implements more complex operations that are critical for the execution of certain types of neural networks,” further said the blog.

With more and more foundation models and generative AI tools entering the market, the performance and energy efficiency of traditional computing methods that these models run on are at a testing limit.

IBM wants to bridge that gap. The company says that many of the chips being developed today have a split in their memory and processing units, thus slowing down computation. “This means the AI models are typically stored in a discrete memory location, and computational tasks require constantly shuffling data between the memory and processing units.”

Speaking to BBC, Thanos Vasilopoulos, a scientist based at IBM’s research lab in Switzerland, compared the human brain to traditional computers and said that the former “is able to achieve remarkable performance while consuming little power.”

He said that the superior energy efficiency (of the IBM chip) would mean “large and more complex workloads could be executed in low power or battery-constrained environments”, for example, cars, mobile phones, and cameras.

“Additionally, cloud providers will be able to use these chips to reduce energy costs and their carbon footprint,” he added.

A rendering of IBM's analog AI chip

the original story appeared here: