Sunday, March 15, 2015

The Brain Chip

The human brain has long fascinated scientists, some of it’s inner workings are still a mystery even today and some researchers have devoted their entire lives to studying how this organ works.  Many compare the human brain to a computer but in reality it’s like trying to compare oranges and apples, although some functions performed by both humans and computers, such as solving mathematical problems, lead to the same results, the processes and the internal workings are completely different.

A computer is composed of mainly two distinct functionalities, processing power and memory, this concept was created 70 years ago and remains basically the same today.
 Even though microprocessors are getting more powerful and smaller, the truth is that they will soon stumble upon a physical limitation where they cannot become any smaller.  Microchip designers working in close collaboration with Cognitive scientists have been looking at how the human brain works to enable them to emulate it on newer hardware.  Leading the race in trying to come up with a cost-effective design for producing Neuromorphic computing power or in layman’s terms, the “Brain Chip” is a team of IBM researchers working on the SyNAPSE project, who, since the 1980s, have been busy trying to understand how neurons and synapses, two key components of how the human brain operates, can be mimicked onto a computer design.  Their ideas and results although promising are still not cost effective and it might take some more time before we see such computing power in everyday applications.
Researchers have in the meantime been designing software that thinks and learns just like a human brain does.  Neural networks can mimic how humans learn and take decisions based on their experience.  Nonetheless such software programs are still limited by the fact that they are running on traditional hardware made up of simple processing and memory functionalities.  Neural nets are today helping in time series predictions.  Traditional software packages were unable to solve any problems in these fields just a few years ago.  Neural nets work by analysing past performance of a particular model and learn from the examples it is presented with.  Once the neural network has finished ‘learning’ it is able to form non linear connections between variables even when there is an amount of noise in the training data itself.  The trained neural network is later used to make predictions about a particular type of event and just like a human learns from it’s mistakes and life experiences, neural networks are able to adjust their own models and therefore ‘learn’ based on the error deviances between it’s own predictions and reality.  As time passes a neural network’s ability to generate precise predictions becomes more accurate.  Software neural nets are today achieving a high degree of success in areas such as self driving cars, weather forecasts, financial risk analysis and even medical diagnosis.
In the near future computer architecture will change completely and processing power will increase exponentially if the Brain Chip becomes an economically feasible reality.  Imagine such a chip working with powerful neural networks, the present barriers limiting artificial intelligence will be broken and computers that think like a human being will become a an everyday reality and no longer remain in the realm of science fiction writers.
During August 2014, a major breakthrough was reported by IBM’s team working in close collaboration with the United States Defense department, as they created a prototype brain chip called TrueNorth.  This processor mimics one million neurones and 256 million synapses , this chip encodes data and processes it in a similar way as the human brain.  The adult human brain in comparison has an average of 85 billion neurones which dwarfs the TrueNorth processor, nevertheless we must not forget that in less than 30 years computers have evolved rapidly and if Brain Chips are to start developing with the same rate, a time when computer processors have enough neurons as a human brain might not be far away after all.
The TrueNorth prototype processor is the first step towards the creation of an economically viable production of such brain chip processors for everyday use.  IBM is at present testing this chip’s capabilities and writing specific software powerful enough to exploit all this computing power before launching it into the market.
Such a processor will accelerate research in artificial intelligence and will pave the way for AI software to be embedded in everyday machines, we might start interacting with our devices in more humanlike ways.  It won’t be long before common medical sensors such as a pacemaker might have such a chip embedded to diagnose problems even before a heart attack gets underway, your smartphone might start learning to anticipate what you will do next such as change to vibrate automatically during a meeting or tell a caller that you are driving and cannot take a call after the chip learns your everyday patterns.  Self driving cars might benefit a lot as well and will probably be the next market to be revolutionalized as cars will start learning to drive by themselves from their human owners or at least take over if anything happens to it’s human driver or if the car senses an imminent crash.
Such advances in technology will undoubtedly lead to more controversial debates and ethical considerations of Artificial Intelligence.  For example if a self driving car had to injure a pedestrian, would the human driver still be held liable?  Recently Elon Musk, founder of Tesla Motors and Stephen Hawking, a leading astrophysicist, released statements cautioning us of the risks which humanity might face if we continue developing human-like intelligence in computer systems.  As with anything else in life such technology may be used less noble intentions as well, it is up to us to decide how we use technology in our possession, after all such A.I. applications may also be used to help humanity progress by helping out research in medicine or in homeland security by identifying potential terrorists trying to sneak into a country with a false passport.

Original Article Scan:

Notice: Article was originally published on the Sunday Times of Malta, TechSunday 15th February 2015.  Written by Ian Vella.  Article is being republished here only for information purposes and copyright is shared between the author and editor therefore republication is not allowed unless written consent is obtained by all parties.

No comments:

Post a Comment

Your IP address (computer identification) is being recorded for authentication purposes only. Any comments related to abuse / drug-use / racism / violence and illegal / crime will be reported.