Get ready to be Mindblown! Neuromorphic Computing & Climate Change?

Neuroscientists everywhere are nerding out over neuromorphic computing (sometimes referred to as neuromorphic engineering), and it's not a new concept. It's been around since the 1980s - but what does this translate to in layman's terms and how quickly is it going to change our lives? Neuromorphic computing is the grouping of designing and architecting systems that mimic the powerful information processing strategies and architecture of the human brain. While many people may remark that clearly computers process information and data faster than people - what's most useful about neuromorphic systems is leveraging the modelling of the human brain for: plasticity (ability to change), resilience and data evolution.

The human brain's most notable features include the prefrontal cortex and the neurological reality of being able to adjust to new information and build new neuronal pathways. This ability to plan, think and consider beyond instinct is precisely what differentiates humans from animals. The fact that people can recover after traumatic brain injuries and how learning and development continues through life are just two fascinating examples of neural nuances that could benefit computing methodologies. What if computing didn't require ongoing programming and instead computation included an implicit expectation that the foundational engineering and programming could actually evolve?

Question: How is this different from neural networks or machine learning? Neural networks are inspired by the behavior of neurons, but their structures are inherently limited. Think about image processing - when you play a video game there is an implicit GPU-latency (graphical processing unit) - that means that each frame you visually perceive requires time for a CPU to process inputs and render a new frame or visual. Neural networks take a huge amount of time for programming and deep learning training - AND there is a high computational cost (think energy consumption)!

Question: So - how does it work? Neuromorphic chips can utilize artificial neurons created out of silicon! From a hardware perspective this means using memristive devices - picture devices that could "remember" and hold previously utilized voltages and currents - this memory retention doesn't rely on an external power source. Analog circuits can be designed with the human brain as a roadmap - this means that not only will processes emulate neuronal pathways but the actual structures and hardware circuitry changes. It's easy to start considering robotics, and even metaverses, in this context, as well as the creation of potentially sentient artificial intelligence. Ethical considerations abound in this space and regulatory compliance mechanisms will certainly be imperative to formulate effective guardrails for safety and security concerns.

In recent articles I've referenced theories about the 5th industrial revolution - neuromorphic computing is one immensely useful way to better structure inevitable human-machine collaboration. As we consider the many concerns in the world at present - how can neuromorphic engineering help us function better as a society? The answer to that question largely depends on your personal opinion of what are the most important issues facing the globe.

Climate Change is front and center in most discussions of global crises. There are many arguments against blockchain and neural network technologies citing that such processes require data mining, consume massive amounts of energy, and exacerbate environmental concerns. Neuromorphic strategies could present a balance between the demand for faster data processing speeds while mitigating the effects of energy consumptive practices! Efficient neuromorphic engineering could result in processing data in a more energy preserving manner.

A tangible example is that one of the frustrations with many public sector transformation projects is the requirement that citizens' data cannot leave an edge device - such as a cellphone. Legislative mechanisms raise red flags with data processing that happens in the cloud and cite privacy concerns. From a technologist's perspective this can be frustrating - neuromorphic computing is one possible solution to address challenges with processing on edge - such as biometric validation on a device that meets certain levels of assurance for fraud controls. Processing on edge devices is in many ways more energy efficient and wouldn't rely on external data centers.

There is much more to be explored on this topic but hopefully this little blurb has piqued your interest to go do some research on neuromorphic engineering - and ironically will result in the formulation of some new neural pathways in your own mind!

References:

www.robotshop.com

www.nature.com

www.iopscience.iop.org