Beijing time, August 8 evening news, IBM released a new computing architecture based on the characteristics of the human brain on Wednesday evening. This architecture allows next-generation applications to match the efficiency of the human brain in perception, cognition, and action.
In 2011, IBM developed the "neurosynaptic computer chip" which can simulate the abilities of human brain cognition and activities. Last year, with the help of the world's second-largest supercomputer, IBM successfully simulated 53 billion neurons.
This time, IBM has taken a step further in this field by introducing a new software ecosystem and programming model that simulates the low power consumption and compact size features of the human brain through programming chips.
The ecosystem includes multi-threaded software simulators, neuron patterns that support extensive neural computation, relevant programs, as well as modules representing the blueprint of neural networks.
IBM stated that this ecosystem will bring highly parameterized neural models to the market and form basic information processing units for brain-like computing. Such systems have the ability to recognize, perceive, and even act upon spatial, temporal, and multimodal environments.
IBM's long-term goal is to create a computer chip system equipped with tens of billions of neurons and trillions of synapses, with energy consumption not exceeding 1 kilowatt and a volume of less than 2 liters. In short, everything is modeled after the human brain. (Li Ming)