IBM's New Artificial Synapses Drive AI to New Heights

In the last couple of years, IBM has invested tremendously in producing game-changing computer hardware and is currently concentrating on gaining substantial advances regarding AI, to develop more energy efficient and seamlessly deployable technology advancements on smaller devices.

Now IBM is working on designing a new chip, integrated with neural networking, which is recording a 100 percent efficient AI functionality.

Ordinarily, neural networking is integrated with software on a device. This feature allows AI to analyze the data, offer relevant responses, and further learn from it. Neural fibers are the crown jewel of AI growth, imitating how the human brain works.

With this development, IBM has proven that manufacturing key features of a neural net in silicon makes it a 100 times more effective. Chips produced this way can improve and enhance machine learning in the years to come.

The IBM chip imitates the synapses of the neural network that connect individual neurons in a brain. The intensity of these synaptic connections requires to be tuned for the system to perceive and comprehend. Inside an organic brain, this happens in the scheme of connections and responses growing or withering over time. The researchers working on these chipsets illustrate the microelectronic synapses with an approach, inspired by neuroscience, using two models of synapses: short-term ones for computation and long-term ones for memory. According to Michael Schneider, a researcher at that National Institute of Standards and Technology, this arrangement addresses low accuracy that has baffled researchers in their previous efforts to build artificial neural networks in silicon. Schneider is currently researching neurologically inspired computer hardware.

The researchers examined a neural network constructed from the segments of two simple image-recognition tasks: handwriting and color image classification. They discovered that the system was as accurate as a software-based deep neural network, consuming only one percent energy produced.

The design of these chips is still comparatively tactless and clunky; integrating five transistors and three other segments whereas there is always a single transistor on a standard chip. Moreover, few aspects of the system have been tested and examined only in simulations, a conventional technique for validating microchip designs. IBM still has to build and test an entire chip to create a fail-safe methodology. Although the development may yet require time to gain significance, it is a breakthrough biologically inspired step toward a computer with AI logic burned into its core.