This article was automatically translated from the original Turkish version.
Scientists and engineers have drawn inspiration from nature to enhance the capabilities of computing systems as technology continues to evolve. For instance, the biomimicry approach aims to apply designs developed by nature over millions of years of evolution to technological solutions. Similarly, artificial photosynthesis draws inspiration from the process by which plants convert sunlight into energy, aiming to use solar energy more efficiently and address energy storage challenges. Artificial neural networks were developed based on the human brain’s method of information processing, enabling learning from complex datasets and making predictions. One such example is neuromorphic computing. Neuromorphic computing employs an engineering methodology based on the activity of the human brain. This approach may yield more efficient results than traditional architectures such as the von Neumann architecture, which has been highly beneficial in conventional hardware design. Neuromorphic computing is sometimes referred to as neuromorphic engineering. The term encompasses the design of both hardware and software computing elements. What makes this field compelling is its goal to recreate the complex architecture and functionality of the human brain in artificial systems. It is well known that the human brain possesses sophisticated abilities such as perception, decision making, pattern recognition, learning, and adaptation. Traditional computing relies on binary logic (0s and 1s) and sequential processing. A light switch in our home perfectly exemplifies the binary logic of traditional computing: lights are either on (1) or off (0). This example forms the foundation of how computers process data. On the other hand, following a recipe is analogous to sequential processing: when cooking, we follow steps in a specific order—for example, preparing ingredients first, then cooking, and finally serving. This mirrors the traditional computing model in which instructions are processed in a fixed sequence. Neuromorphic computing, by contrast, is inspired by the brain’s neural networks. Neuromorphic systems emulate the brain’s efficiency by using parallel processing and interconnected nodes to handle complex tasks.
Figure 1: The Human Brain and Its Working Principle
Figure 2: Fundamental Principle of Biology-Based Neuromorphic Device Design
Figure 1 illustrates how the human brain functions. Our brain hosts 86 billion neurons that transmit and relay information. These neurons communicate with each other through connections called synapses. Information transfer between neurons occurs via these synapses, enabling the brain to perform functions such as thinking, learning, and memory. Figure 2 depicts a device (NPU) designed to mimic the functioning of the human brain. This device contains artificial neural networks (ANNs) and artificial synapses. Artificial neural networks replicate the function of neurons in our brain, while artificial synapses replicate the function of synapses in our brain. These devices typically have two or three terminals (connection points).
Mechanisms of Neuron and Synapse Mimicry
Neurons and synapses are known as the fundamental building blocks that enable information flow within the brain. Neurons communicate with each other through chemical and electrical signals. Synapses are specialized connection points that make this communication possible. Neurons and synapses are far more flexible, adaptive, and energy-efficient information processors than traditional computer systems. Neuromorphic computing achieves this by designing circuits that emulate the behavior of neurons and synapses. This approach enables machines to process information like the human brain and contributes to advancements in artificial intelligence and cognitive computing.
So far, we have discussed what neuromorphic computing is and its biological sources of inspiration. Like any technology, neuromorphic computing has its own advantages and disadvantages. What are the benefits and drawbacks of neuromorphic computing?
What Are the Advantages and Disadvantages of Neuromorphic Computing?
History of Neuromorphic Computing
1936 – Mathematician and computer scientist Alan Turing developed a mathematical theorem proving that a computer could perform any mathematical computation.
1948 – Turing wrote an article titled “Intelligent Machines,” in which he described cognitive models based on human neurons.
1949 – Canadian psychologist Donald Hebb made a groundbreaking contribution to neuroscience by establishing a link between synaptic plasticity and learning.
1950 – Turing developed the Turing Test, still regarded as the standard test for general artificial intelligence (AGI).
1958 – The U.S. Navy developed the perceptron, inspired by biological neural networks, for image recognition. However, due to limited knowledge about brain function at the time, it failed to achieve its intended functionality. Nevertheless, the perceptron is recognized as a precursor to neuromorphic computing.
1980 – Neuromorphic computing, in its current form, was first proposed by Caltech professor Carver Mead. Mead argued that if the nervous system’s operation were fully understood, computers could replicate everything performed by the human nervous system.
2013 – Henry Markram launched the Human Brain Project (HBP) with the goal of creating an artificial human brain. The HBP, initiated to better understand the human brain and apply this knowledge to medicine and technology, has a 10-year timeline. Additionally, over 500 scientists and 140 universities across Europe are working on this project.
2014 – IBM developed the TrueNorth neuromorphic chip, which consumes significantly less power than traditional von Neumann hardware. This chip is used in visual object recognition.
2018 – Intel developed the Loihi neuromorphic chip, with applications in robotics and gesture/smell recognition.
Some examples in the field of neuromorphic computing:
Tianjic Chip Using Computational Approach
Figure 1: Tianjic Chip
This chip, developed by Chinese scientists, has been used to power an autonomous bicycle capable of tracking a person, navigating obstacles, and responding to voice commands. It contains 40,000 neurons and 10 million synapses. It also delivers 160 times better performance and 120,000 times greater efficiency compared to a similar GPU.
Today, neuromorphic computing is one of the most exciting developments in computer science. However, these current efforts represent only a beginning toward even more advanced future technologies. In the future, this technology may lead to systems that fully replicate the complexity and efficiency of the human brain. This could expand the boundaries of computers and artificial intelligence and perhaps transform our lives in ways we have never previously imagined.
Blog Author
Afranur Sude KAYAOĞLU - Yükselen Yıldız Scholar