Biomimicking the brain

1. What is biomimicry?

Velcro was invented by Swiss engineer Georges de Mestral after noticing how burrs clung to his dog’s fur while in a forest area. He mimicked the hook-and-loop structure of the burrs to create a fastener that could easily attach and detach.

The lotus plant's leaves have a micro- and nanostructured surface that repels water and dirt, keeping them clean. This property has been mimicked in self-cleaning coatings for windows, roof tiles, and fabrics, which repel water and prevent dirt buildup.

Japan’s Shinkansen bullet trains were redesigned to reduce noise and improve energy efficiency by mimicking the streamlined beak of the kingfisher.

These are some fascinating examples of biomimicry.

Biomimicry is the practice of emulating or drawing inspiration from nature's designs, processes, and systems to solve human problems and create sustainable technologies or innovations. By observing how plants, animals, and ecosystems have evolved to adapt and thrive, scientists, engineers, and designers develop solutions that mimic these natural strategies to address challenges in fields like architecture, engineering, medicine, and materials science.

2. Key Aspects of Biomimicry

Nature as Model:

Nature offers models and designs for solving problems. For example, studying how birds fly to create better aircraft designs. Biomimicry focuses on studying how nature has solved complex problems through evolution. It’s about learning from nature, not just copying it.

Sustainability:

Nature has been solving complex problems for billions of years in efficient and sustainable ways. Many of nature’s processes are inherently sustainable, using minimal energy and materials. For example, the Eastgate Centre in Harare, Zimbabwe, is designed based on the ventilation systems of termite mounds. This innovative design helps reduce energy consumption for air conditioning and heating.

Innovation:

The working of biological systems can inspire innovation and development of cutting-edge technologies and products. Biomimicry takes clues from these biological systems and applies in various fields such as architecture (for designing energy-efficient buildings), robotics (for developing robots that mimic animal movements), and medicine (for creating medical devices based on biological systems).

In essence, biomimicry encourages viewing nature as a teacher. Rather than exploiting nature, it promotes learning from it. Nature’s designs are adaptable and resilient, providing solutions that are not only effective but also durable and sustainable in the long term. Biomimicry is about looking to nature as a source of innovation, sustainability, and efficiency, leveraging the wisdom of the natural world to improve human designs and processes.

To sum-up, Biomimicry offers a wealth of innovation by drawing on nature's time-tested strategies. Humans can create sustainable, effective solutions by emulating the ingenuity of the natural world.

3. Mimicking the human brain

Now, our brain is central to all thinking, computation, analysis, cognition and lot more. Humans have always fascinated to have a device that has the capabilities of brain. Let us see how the concept of Biomimicry is being applied to mimic the human brain.

The computers are often likened to a brain.

Colossus was the world’s first large-scale electronic computer. It was developed in 1944 and used thousands of vacuum tubes and was semi-programmable. It was designed for a specific purpose—to break the Lorenz cipher used by the German military during World War II.

Since then, computers have undergone revolutionary developments both in terms of Hardware as well as Software. Artificial Intelligence has evolved significantly to rival the brain’s cognitive abilities. So much so, that we are now talking about how to control and hold AI.

Seems we are very close to having developed artificial brain. But, a more realistic view will show us the challenges and limitations faced by current computers and AI systems.

4. Challenges of current computers and AI systems

Traditional computers which are based on von Neumann architecture, consume a lot of power due to the separation of processing and memory units. Data has to constantly move back and forth between the processor and memory, which is energy-intensive. This also results in delays and inefficiencies because the speed of data transfer limits overall system performance. Also, these computers are reaching physical limits in miniaturizing components. Smaller transistors lead to problems like increased heat generation, power leakage, and signal interference.

AI models, especially deep learning algorithms, require vast computational resources and consume significant energy during both training and inference. Large-scale models, like those used for natural language processing (e.g., GPT-3) or computer vision, often run on massive clusters of GPUs, which consume enormous amounts of power.

Computers are great at performing precise calculations with huge numbers, but when it comes to learning and abstraction, they are nowhere close to brain.

Neural networks are a software approach that mimics how brain learns. A neural network changes when it’s shown lots and lots of examples of what it’s supposed to learn, but it may need to see thousands to millions of examples to achieve the desired results, like how to tell the difference between a chihuahua dog and a blueberry muffin or a furry dog and a mop.

Clearly that’s not how we learn. As humans, we don’t need to see millions of pictures of dogs before we know what a dog is.

Also, our brain just uses about 20 watts of power, while computers need significantly higher power to operate.

Hence, an alternate approach is needed to mimic human brain and its capabilities.

5. What is Neuromorphic Computing?

Neuromorphic computing is a field of computing that seeks to mimic the structure and functioning of the human brain and nervous system in hardware and software. The goal is to create systems that process information more efficiently, flexibly, and adaptively, much like how biological neural networks work. This approach involves creating hardware that closely emulates the behavior of neurons and synapses, which are the fundamental units of biological brains.

6. Key Concepts of Neuromorphic Computing

Neuromorphic systems are designed to replicate the way biological neurons communicate with each other using electrical signals called action potentials via synapses.

Like brain, Neuromorphic hardware works in an event-driven manner. This means that computations are performed only when spikes or events occur leading to lower energy consumption. This is known as Spiking Neural Networks or SNNs. They are able to perform complex computations with much lower power consumption than traditional digital computers.

Neuromorphic systems aim to incorporate the brain’s ability to learn and adapt over time. This is done through mechanisms which are similar to brain’s synaptic plasticity, which allows neural connections over synapses to strengthen or weaken based on experience, much like how learning occurs in the human brain. One such mechanism used in neuromorphic systems is Spike-Timing-Dependent Plasticity (STDP) where the strength of connections between neurons is adjusted based on the timing of spikes, similar to how biological neurons work.

Neuromorphic architectures are being developed to integrate memory and processing in hardware, mimicking how the brain handles memory storage and real-time data processing, allowing for rapid information retrieval and decision-making. This is achieved through memristors.

7. What are Memristors?

Memristors (short for "memory resistors") are a type of non-volatile, passive electronic component that can regulate the flow of electrical current in a circuit and "remember" the amount of charge that has previously passed through them, even after power is turned off. This memory-like behavior is what sets memristors apart from traditional resistors.

The resistance of a memristor changes based on the amount and direction of electrical current that has passed through it. Once the current flow stops, the memristor "remembers" its last resistance value. They can switch between different states (high and low resistance) much faster than traditional storage technologies, like flash memory. This behavior allows the memristor to act as a programmable resistor, whose state depends on its history of voltage and current.

Memristors can be used to build logic gates, which perform computational operations. Since they can store data and process it simultaneously, memristors offer the potential for "in-memory computing," which could dramatically improve computing speed and reduce power consumption.

Memristor technology is still in the development phase and large-scale adoption in commercial products has been slower than expected.

8. Development of Neuromorphic Hardware

Several companies and research institutions are building neuromorphic chips and systems.

IBM’s TrueNorth is one of the earliest large-scale neuromorphic chips that was announced in 2014. It contains 1 million neurons and 256 million synapses. Intel’s neuromorphic chip called Loihi was introduced in 2017. Its 130,000 neurons are each capable of communicating with thousands of others for a total of over 130 million synapses.

Intel has also released Loihi 2 in 2021, a second-generation chip, with improvements in performance and flexibility. IBM released its NorthPole chip in 2023, which is a proof-of-concept for dramatically improving performance by intertwining compute with memory on-chip, thus eliminating the Von Neumann bottleneck. It is designed to achieve speeds about 4,000 times faster than TrueNorth.

Pohoiki Springs, a computer created using 768 Loihi processors, that’s the size of 5 servers and boasts 100 million neurons. That’s similar to brain size of a small mammal like mole.

Intel claimed in 2024 that Hala Point was the world’s largest neuromorphic system. It uses Loihi 2 chips. It is claimed to offer 10x more neuron capacity and up to 12x higher performance. At 1.15 billion neurons, its neuron capacity is roughly equivalent to that of an owl brain or the cortex of a capuchin monkey.

Other efforts at neuromorphic hardware are BrainScaleS developed by the European Human Brain Project, and SpiNNaker developed at the University of Manchester.

Neuromorphic hardware is just emerging, software that can make the best use of it needs time to develop. Still, it’s something to look forward to.

Neuromorphic computing is a key area of collaborative research for universities, government organizations, and companies working in artificial intelligence, neuroscience, and computing. Some of these efforts are: DARPA’s SyNAPSE Program by the U.S. Defense Advanced Research Projects Agency and European Human Brain Project.

9. Key Applications Under Development

Though neuromorphic computing is still evolving, some early applications have shown its potential in specific domains like Robotics and Autonomous Systems, Edge Computing, IoT, Real-time Learning, Healthcare, Brain-Machine Interfaces (BMI) and Energy-efficient AI.

10. Challenges

Creating efficient software that can fully exploit the potential of neuromorphic hardware remains a challenge as Neuromorphic computing requires different algorithms and models compared to traditional AI systems. Other key challenges are Scalability, Adoption Barriers and Interfacing with Traditional Systems.

Neuromorphic computing is a rapidly evolving field with great potential but is still in its early stages. While there have been significant advancements in neuromorphic hardware and software, widespread practical applications and commercial adoption remain limited. Researchers are actively working to overcome technical challenges and improve the efficiency and scalability of neuromorphic systems.

At the end

With tremendous growth of AI, the power consumption and heat generated by the servers running those AI systems is also increasing exponentially. Despite these advancements, the AI systems require long learning cycle. Neuromorphic Computing aims to mimic human brain to solve challenges faced by current computers and AI systems. Like other solutions created using biomimicry, neuromorphic computing can potentially provide a sustainable solution for the problems faced by current computers and AI.