Neuromorphic engineering is the art of getting inspiration from the realm of neuroscience in order to build better electronic systems. It is an highly interdisciplinary field, mixing electrical engineering, computer science, neuroscience and mathematics.
Indeed, it is well known that Moore’s law, which says that the number of transistors that can be placed in an integrated circuit doubles approximately every two years, will soon become just impossible to follow. Of course, one solution is to move away from the traditional CMOS fabrication processes and to go for new technologies like CMOL or memristors, which I will also talk about in this blog. But still, even with these new technologies, fabrications processes will become more and more unreliable, which means the yield of integrated circuits will drop drastically.
A revolution is therefore also needed at the architecture level. We need new architectures that can cope with the inherent yield problem of these new fabrication processes. And also which are low-power, because we love putting these things into our smartphones and tablets. This is where neuromorphic engineering kicks in. Indeed, we already know a system that is massively parallel, low-power, and defects tolerant : the brain. Neuromorphic engineering proposes to replicate the architecture of the human brain on silicon chips, emulating neurons and synaptic connections between them. The product will be a new class of integrated circuits that works in a massively parallel way, are defect tolerant, and consumes much less power than the current integrated circuits.
Of course, there are challenges, and these will also be discussed intensively on the blog. Configuring and using these kind of neuromorphic chips is probably the most difficult part, even more than actually making the chips. But if we can go beyond these challenges, neuromorphic engineering can really create a revolution in our daily life. Think about possible applications like systems that can learn from their mistakes, virtual agents that can act as personal assistant, completely autonomous robots … the list is long. These possible future applications, as well as already existing neuromorphic systems, will also be discussed on the blog.