According to Gartner, traditional computer systems based on legacy semiconductor architecture will hit a digital wall by 2025, forcing changes to new paradigms such as neuromorphic computing. Meanwhile, Emergen Research predicts that by 2027, the worldwide neuromorphic processing market will be worth $11.29 billion. So, what is neuromorphic computing? 

As the name suggests, Neuromorphic Computing works by mimicking the physics of the human brain and the central nervous system's information processing architecture. "It is an intersection of diverse disciplines including neuroscience, machine learning, microelectronics, and computer architecture," said Dr Manan Suri, IIT Delhi, earlier to INDIAai. 

What's so different with the human brain? 

The human brain, unlike traditional computing systems, does not deal in rigid binary terms - in 0 and 1, or if it is 'yes' or 'no'? Compare a human brain with a supercomputer and the very first thing strike - size - a supercomputer fits in an entire room while the brain fits in the size of the head. Adding on, the human brain utilises around 20 watts of energy, whereas the Fugaku, the world's fastest supercomputer, requires 28 megawatts (2.8 × 10^7 watts). Finally, while supercomputers require complex cooling systems, the brain is housed in a bone casing that maintains a constant temperature of 37°C. 

Supercomputers can do calculations with greater speed than our brains, but the systems are still miles away from the point to think creatively, adapting to different situations, recognise people or objects they have never seen before, and somehow adjusting in order to act - the way a human brain does. This is where neuromorphic computing, harnessing techniques of our human brain, comes into the picture. 

Is it really the technology for tomorrow? 

As the demand for high computing power is heating up every single day, the time is ripe to look for possible alternatives. So, it is important to understand how neuromorphic computing can be the best bet forward.

Sustainability domain: Time and again, nature has shown us to get the job done in an energy-efficient manner. "Suppose you simulate the brains of a cat – it could be related to 128,000 processors, with each one having a 1GB memory. Studies show you need at least 100 petaflops to emulate a human brain. To power up this cluster, you would need a mini data centre. There is something inherent in how mammalian brains process that makes them low power and energy-efficient. Researchers and the industry at large should explore bio-inspired neuromorphic structures as we will be looking into the sustainability domain very soon," said Dr Manan Suri. 

Mammalian brains are built in such a way that they can operate at low power levels. This type of structure should strive to create for future machines. "The farther we delve into AI, the more we venture into uncharted terrain in terms of long-term viability," he further added. 

Removes existing bottlenecks: The traditional personal computer architecture faces the von Neumann bottleneck, which limits throughput. In the von Neumann architecture, programs and data are stored in memory; the CPU and memory are separate, and data flows back and forth between them - wasting a lot of time and energy. Both quantum computing and neuromorphic systems have been proposed as solutions; however, keeping in mind the fragile nature of qubits, neuromorphic computing or brain-inspired computing is more likely to be commercialised. 

Expanding the scope: As the power consumption drastically reduces, it will lead to the point where a device might run for years instead of weeks or months on a tiny battery. This would eventually pave the way for using the chips in machines that need to perform computationally complex deep learning operations locally, such as autonomous vehicles, facial recognition security cameras, and military drones. 

Way forward 

The race has already started; Intel and IBM have their own neuromorphic chip named Loihi and TrueNorth, respectively. In 2020, Intel demonstrated a neuromorphic robot that, unlike standard models that require substantial instruction and data, can sense and recognise unknown objects based on just one example. In addition, the Human Brain Project (HBP), a 10-year EU-funded project that began in 2013, aims to increase brain understanding through six research fields, including neuromorphic computing. 

"We always talk about manmade structures like the Eiffel Tower or Burj Khalifa. These are about 800 m and 160 floors in height and are considered civil engineering marvels. In nanoscale memory tech, with 3D semiconductor flash, it has nearly 128 floors of data structures packed and neatly engineered in a few 100 microns. Semiconductor hardware is no hype – this is what we need to focus on for a richer future in AI," says Dr Manan Suri.