China's 'Darwin Monkey' supercomputer, developed by Zhejiang University & Alibaba, now emulates a macaque brain with 2 billion artificial neurons, aiming for breakthroughs in AGI and neuroscience with high energy efficiency.
Darwin Monkey or 'Wukong' features over 2 billion artificial neurons and more than 100 billion synapses — similar to the neural structure of ...
Darwin3, the largest neuromorphic chip currently available, can support up to 2.35 million neurons and 100 million synapses. It presents an efficient neuro.
Beijing, PRC - Chinese scientists have revealed "Darwin Monkey," also called "Wukong," a neuromorphic supercomputer built to mimic the neural architecture of a macaque brain. Created by researchers at Zhejiang University and Zhejiang Lab-a collaborative effort between the Zhejiang provincial government and Alibaba Group-this platform boasts more than 2 billion artificial neurons and 100 billion synapses. It marks a noteworthy leap forward in brain-inspired computing, nearly equalling the neuron count of a macaque.
The Darwin Monkey system incorporates the Darwin 3 processor, which employs brain-inspired instruction sets and on-chip learning algorithms. This arrangement harnesses spiking neural networks (SNNs), a shift away from conventional artificial neural networks that depend on classic computing principles and continuous binary data handling. SNNs emulate how biological neurons convey signals by reacting to electrical stimuli and processing information through discrete "spikes" or bursts of activity, firing only when sufficient electrical input has accumulated.
The supercomputer's architecture is powered by 960 Darwin III neuromorphic chips, each capable of supporting up to 2.35 million spiking neurons. Despite its intricacy and processing prowess, the Darwin Monkey is reported to draw roughly 2,000 watts of power-about the same as a typical electric kettle or hairdryer. This level of energy efficiency contrasts sharply with that of conventional high-performance computing systems.
For perspective, the former leader in neuromorphic computing, Intel's Hala Point system, contains 1.15 billion artificial neurons and 128 billion artificial synapses across 140,544 processing cores, delivering an estimated 20 petaops (20 quadrillion operations per second). While exact peta-operation figures for Darwin Monkey have not been disclosed, its capacity to host a larger tally of artificial neurons positions it as a formidable contender in the arena.
Researchers outline two main goals for Darwin Monkey: serving as a simulation platform for neuroscientists and acting as a stepping stone toward Artificial General Intelligence (AGI). AGI denotes AI systems that exhibit human-like intelligence and reasoning abilities. The system has already shown competence in cognitive tasks such as logical reasoning, content generation, and mathematical problem-solving, utilizing an AI model developed by the Chinese AI startup DeepSeek.
In addition, the platform is being applied to broader brain-science investigations to emulate animal brains with varying neural complexities, including those of zebrafish and mice. This work builds on earlier projects, like the launch of Darwin Mouse ("Mickey") in September 2020, which featured 120 million artificial neurons and mirrored the neural structure of a mouse.
The emergence of Darwin Monkey highlights a worldwide movement toward weaving neuroscience concepts into advanced computing architectures. By copying the energy-efficient, event-driven processing of biological brains, neuromorphic systems aim to overcome several constraints of traditional Von Neumann designs, especially in domains that demand intricate pattern recognition, adaptive learning, and low-power operation.
This progress aligns with a broader international emphasis on creating more efficient and biologically plausible AI systems. The capacity to simulate extensive neural structures at this scale could grant neuroscientists unprecedented tools for probing brain function, potentially hastening breakthroughs in neurological disorder research and the comprehension of cognitive mechanisms. For AI, platforms like Darwin Monkey embody a hardware-software co-design paradigm that may pave the way for more sophisticated and embodied forms of artificial intelligence.
References:
The original article from Interesting Engineering is a brief but factually correct announcement of China's "Darwin Monkey" neuromorphic computer. Its core claims regarding the system's name, neuron count (over 2 billion), and synapse count (100 billion) are corroborated by a more detailed report from Live Science, which appears to be based on primary research published in the National Science Review.
While accurate, the original article is superficial and omits significant context provided by other sources. It fails to mention:
The original article functions as an accurate headline summary. However, its lack of sourcing and comparative data prevents it from offering a comprehensive understanding of the technological development, a task more successfully accomplished by the Live Science report.
12 листопада 2025 р.
Darwin Monkey or 'Wukong' features over 2 billion artificial neurons and more than 100 billion synapses — similar to the neural structure of ...
Darwin3, the largest neuromorphic chip currently available, can support up to 2.35 million neurons and 100 million synapses. It presents an efficient neuro.
Related Questions