Current chip technology uses billions of small electronic components called transistors to do the computation we are all used to. From self-driving cars to watching the latest cat videos on YouTube, all is done on the same horde of transistors. However, we reached the limit of how many of those transistors we can place on the same chip and how small we can make them.
That is why a new electronic component comes to life right now. It behaves like a neuron and will make it possible to build computers that are more efficient than current computers based on transistors. Or maybe even more efficient than the human brain.
The human thinking computing paradigm
The human brain is the result of millions or even billions of years of evolution. It is the most advanced “piece of technology” we know of, and we just recently started exploring it and how it works. We don’t yet know all of the mechanisms that work in creating who we are or how we think.
In current chip architectures, all processing is done in a serial manner — that is one calculation after another. Parallelism in binary chips means, basically, adding more chips (or cores). There is a limit, however, to how many of those cores we can add, and adding more cores means more dollars spent.
The human brain does things differently. Each neuron of our brain can be connected to any other neuron. That doesn’t look like much until you find out that we have approximately 86 billion neurons in the average human brain. That means each neuron has 86 billion possible connections.
What is mind-boggling is the fact that in the human brain computations (or decisions) can be done in parallel. Our brain can do up to 100 billion operations, in parallel, each second of our life.
More so, those operations are not binary in nature, but analogical. Our brain doesn’t work in 1 or 0, it works in current, voltage, and chemical substances (which are analog by nature). That means each operation done in the human brain is equivalent to a few thousand done by binary computers! This is based on only the things we know about how our brain works. We still have much to learn.
New components, new computing
Creating a brain-like machine was impossible practically, until 2009 when the 4th fundamental passive electronic component was found — the memristor. You can read more about this in The Future of Artificial Intelligence — Neuromorphic Computing.
What is important about memristors is the fact that they can behave in a way that is very similar to the way neurons work. Some of them also have a very analog nature, as they can have hundreds of resistive states. Comparatively, transistors usually have only two, on and off. More famously this is known from computers as binary, 0 and 1 — like in The Matrix.
And remember, all of this digital revolution had at its center one thing: the transistor. It all started with the transistor and from hardware development. Even if the world is now “run by software” as some people would say, that software is still run on hardware — on that little electronic component called the transistor.
Now we are at the beginning of a new age and a new revolution. The memristor will be the central element of this revolution, just like the transistor once was. It will enable Highly Parallel and Low-Power computing units that will use a completely different paradigm than the current processors. It will be the comeback and revolution of analog computation — a type of computation that was pushed under the carpet for far too long.
And the fact is that we need this new type of computing paradigm. Quantum computers are not yet there and will not be for a long time, so we cannot count on them. Moore’s Law is dying, and there is no cure.
Be a smart car and use your brain
This little pet project was built by CyberSwarm’s team and is using an astonishing number of only 16 memristors (i.e. Neurons) to follow a line and make decisions based on a QR code.
You can see this car in action here.
All of the decision-making in the video is done using only 16 neurons, instead of a billion transistors, like in the example of current processing units.
You can only imagine what this car could do using even a million memristors/neurons.
There is a lot of work to be done and breakthroughs to be found, but this clearly exemplifies the difference between a computing unit working on memristors and a classical, binary CPU.
