Scientifics, both in Europe and in the USA, wish to make new computers by imitating human brain. This is, according to them, the best way to have both better and more efficient computers and also a better understanding of the human brain.
This summary is based on an article from the economist “neuromorphic computing, the machine of a new soul”. I read it a long time ago, remembered it was quite passionnating and I want to share it with you. Sorry if I did mistakes while trying to rewrite it. This is far much better for you to read the original article and also because I was bored finishing this post (I also need to read it again before any attempt to write something about this subject.
This is the work and aim of scientists who called themselves neuromorphic engineers. They want to build a computer that has the three main characteristics of the human brain.
1. low power consumption (the human brain is estimated to have a consumption of around 20 watts whereas actual supercomputers require megawatts)
2. fault tolerance (losing even just one tiny transistor should not be a problem for the whole computing system as it is actually for microprocessors with a x86 architecture or ARM architecture)
3. A lack of need to be programmed (the human brain learns by itself and does not need to follow a fixed and imposed path ruled by algorithms implemented by programmers).
With such differences, it is easy to understand that the difference between actual computers and brains are huge. There is a gap hard to fill. But neuromorphic computing’s aim is reduce such a gap. Mainly, by creating a real analogy between artificial computers and brains. This goal also requires to all gaps in neuroscientists understanding of the human organ. I would be at pain to understand how we can speak of an equivalent electrical consumption of brain expressed in watt, but anyway, I am going to trust this amazing article I read in the Economist. The beginning is to create artificial cell brains and to connect them in various ways to try to mimic what happens naturally in the brain.
Neuroscientists can currently understand both high and low level of bain’s anatomy, I mean for one hand the level of nerve cells and how they work, and on the other hand the global architecture of the human brain. For every emotion, every move, neuroscientists know which parts of the brain are concerned. But how neurons are organized and worked at the intermediate level of brain is obscure. The America’s BRAIN initiative is to achieve a global understanding and mapping of the human brain. The European human brain project is more advanced that what they do in the USA. A first example of an like brain computer is awaited for 2023.
We need to explain the difference between a analog computer and a digital one. The digital computer, well-known and part of every computer, smartphone or tablet is processing informations as a series of ones and zeros represented in binary way by either the absence “0” of a voltage or the presence (1) of a voltage. This is how, almost every contemporary computers are working nowadays.
But we have to stress the fact that at the beginning of computer science, several computers were analog machine. Different variety and intensity of voltage is used to process information. This is because of possible errors during the voltage manipulation that digital computers won over their analog peers. But this is true to say that analog computers are closer to human brains in their way of functioning and processing information. SpiNNaker is a project to create an analog computer with around 1m processors able to model about 1% of the human brain in real time.
Actual computers are processing information by shuttling relatively few large block of data around under the control of a central clock (this is where the Hz, Mhz, and Ghz, frequencies of the central clock mean something). Analog processors spit out lots of tiny spike of information as and when it suits them.
You can find the original article here … Because my article is very bad …. sorry… and two other articles about artificial intelligence
