Artificial Intelligence


                    The one Phrase that comes to everybody's mind when we say Artificial Intelligence is "Terminator" , all credit goes to Hollywood movies like  "I Robot" , "Wall-e", "2001: a space Odyssey" and obviously "Star Wars", thought its very interesting and pretty cool stuff and all , its Science Fiction , I am going to be talking about the current state of Artificial Intelligence , its history ,its possible future and how far all of these movies might be from our present situation. 
                   Hollywood directors and story writers have captured all our imagination by making amazing movies that in a sense inspires the scientific community to realize the future technology but actually how far are we from the real life Star Wars , Lets see : if you imagined the ultimate future to be filled with flying cars , super intelligent computers , lasers and all kinds of cool gadgets , that is way down the line , how much down the line ? , Lets say our current best technology neural networks does this :


Is this what true Artificial Intelligence is ?

                    Why is it that even though , the most important thing we all are so dependent on , the Computer  cannot describe a video or an image which even a 3 year old kid can ? To begin answering this question i have to talk about the approach scientists are taking to solve this problem
I have to mention that there are two broad categories of approaches scientists are branching on to solve this complicated problem , humankind has ever faced : one is from an Engineering perspective : where we break down the problem into easier and smaller sub problems and solve individual problems and integrate the solutions to solve the bigger problem (kind of makes sense , but if you think about it , this will be very slow and painful to go about , i ll explain why later) , and the other approach is  the biological approach , where we try to understand how our brain does all these intelligent and complicated tasks so seamlessly and try to replicate it.

                    Talking about the progress in both the approaches : In the engineering approach The leading scientist at Facebook's cutting edge AI Research lab , Yann LeCunn : Director of the lab , who is leading the research efforts at Facebook using Deep Convolution Neural Networks which they think is the key to solving AI , Lets see what this does  At its best, it can recognize different objects in a given image, from an engineering standpoint their best system can process three of below images per second , you might be thinking “hey wait a second!!! , Deep Convolution Neural Networks sounds way more cool , for just a simple Object Recognizer” , to be fair these neural networks can be trained to do a variety of things like , recognize text , convert speech to text and vice versa , compose music ,  though these things are good at doing a variety of things , once a neural network is trained to do a specific task like identifying objects in images , playing games , playing .

The First artificial neural network was invented in 1958 by psychologist Frank Rosenblatt called Perceptron , it was intended to model how the human brain processed visual data and learned to recognize objects. Other researchers have since used similar ANN(Artificial Neural Networks) to study human cognition , This technology was realized theoretically in 50's but computers back then lacked the compute power to implement the technology.

Theoretical Model of a simple neural network

Perceptron is the digital equivalent of its biological counterpart the actual human brain neuron , they have typical on and off state and multiple perceptrons are placed next to each other to form layers and all these layers are stacked on top of each other , all the layers in between the first and last layers are called hidden layers , the whole arrangement together is called a neural network , there are many different types of neural networks , you can check out a neural networks in action in this video , this neural network classifies a hand written digit.

Each one of the tiny white dots you see is a digital neuron , and we need  millions of these just classify handwritten digits effectively, at this scale we would need neurons of gazillion number to get a computer to do even simple tasks intelligently which kind of gives me a vibe that this not the way to go as brain does it much more efficiently.

There is a lot of hype in the media and in the Engineering AI scientific community about Deep Convolution Network and its capabilities , this technology has been able to do something none of our previous technologies have not been able to do , few examples of technologies that we see a lot today like Self Driving cars , the face detection in all our fancy smartphone camera's , are all gifts of neural networks , This technology is good at Pattern Recognition , but the actual neural networks in our brain are far more better at Pattern Recognition the reason being that they are not just recognizing patterns but they try to predict patterns and whenever they are wrong they make adjustments to themselves all the time , without us even noticing , our brains do a plethora of computation to just see through our eyes and analyse the image ,our brain forms this 3D model of the world around us by processing all the visual , auditory and olfactory data coming in from our senses , the reason why we don't realize all of this is that all of these computations are processed by our Subconscious which is Powerful at Computation and Prediction but lousy at reasoning , creativity , drawing inferences , making conclusions , connecting the dots , all of these things are performed by our Conscious Mind that is  us , we are what we are because of these two systems perfectly working in sync with each other , and when they don't work in sync then we are brain defects and personality disorders.

Though these neural networks seem pretty complicated , they are a simplified version of there biological counterparts the real Neuron , recent research in Neuroscience have identified how a neuron can have 4 different kind of states in which it can be and understand its different kinds of dendrites and its morphology , though these neurons are functionally pretty simple on its own but when you put a 100 billions of these together they perform miracles(intelligence) , which brings me back to the topic of the second approach to AI : The Biological approach : where we try to actually design the human brain on the actual principles on which the human brain works ,  this approach is less famous in silicon valley and gets way less screen time in the media , the reason being , the human brain is way too complicated , and we still don’t know a lot about how the brain works , Not to be misconstrued here , We know a lot about the brain , more than ever before , but its all in bits and pieces , there is very small focus given to constructing a unified theory of its core inner workings  , but fortunately this trend is changing there is a lot of focus given to Computation Neuroscience (its interdisciplinary study of cognition and computation in the brain) a lot of universities are covering great strides in developing better models to understand brain like University of Washington , University of California Berkeley , Carnegie Melon University ,Boston University,  University of Chicago , University of Illinois Urbana Champaign , University of Massachusetts and there silicon valley companies who are leading research in these fields like Numenta : a true propeller of research in development of intelligent systems  

Where is my Personal Walking Talking Robot , you ask ?

Not Quite Yet I say.

                     Turns out understanding how the Brain works ? , is easier said than done! , it is a fact that the human brain has an estimated 100 billion neurons(brain cells)  , and that constitutes to a giant gazillion synapses(connections between neurons) a number higher than the estimated number of stars present in the visible universe , now that's a really really big number , just a small pinhead size of neural tissue contains hundreds of thousands of neurons and millions of connections , all of these neurons are are arranged in hierarchical layered structure in the human NeoCortex , the NeoCortex is assumed to be the source of Intelligence and sense of conscious that we all have , Neo Cortex is a typically a mammalian characteristic , so are only mammal intelligent , no there are many other animals intelligent but just no as much as us , otherwise they would be ruling earth not us , neocortex has been evolving over millions of years and the human neocortex is the epitome of natures crowning glory , it is so big in human that our skulls had to be extended in size to fit it in , and in females pelvic adjustments were made to facilitate child birth for these big heads , typically neocortex  is the giant convoluted bubble-gummy thing you see in every brain image you have ever seen , it has 6 layers and is about 2.5 - 3 mm thick , all your personality , your memories , your intelligence , your knowledge resides here , in laymen's terms everything that is you is here, if you want know more about how all of this works read my another blog dedicated to NeoCortex over here : .

                      Even though the biological approach is the right way to go about this , the way things are going in the field to solve the problem is not directed well enough, we are taking a bottom first approach rather than top first approach let me tell you what am talking about , We literally have thousand's of research papers presented at Conferences in US alone but all of these are related to only very specific microscopic parts of the brain and theories about these microscopic parts and no effort is made to integrate this theory with existing holistic theory of brain , this is like trying understand  why is sun so hot without knowing any concept of Gravitation Force and Fusion Reaction , if everybody knows and agrees on Gravitation Force and Fusion Reaction , suddenly the whole world will start to make sense and science would move much more faster pace than before , although recently in the past decade a good number of companies and research institutes have come up with great theories and scientific proof's and are building solid provable fundamentals of brain theory for everybody to delve deeper into leading the innovation is , a company co-founded Jeff Hawkins , creator of Palm Pilot and number of other mobile computing gadgets back in the 2000's who took a giant leap from mobile computing and internet world to Brain Theory and Neuroscience after realizing how important this things is , please do checkout some of the videos  at to know more about them.