Sep 1, 2011

Human-Inspired Intelligence in Machines: Is the current research on the right path?

Bayesian learning, artificial neural networks etc. come under machine intelligence tools, but none of the programs is actually intelligent. None of the human-inspired machines is as efficient and intelligent as the human brain which is just a mass of protoplasm that can contemplate the universe and even contemplate itself contemplating the universe.
More than 60 years have passed since the beginning of research on AI, but what we have got now are machines which cannot even identify the differences between a cat and a dog, which a two-year old kid could do with almost no effort. This paper discusses the fundamental flaws in the current research trends on human-inspired intelligence in machines and proposes a few right directions towards achieving truly intelligent machines.

This is an excerpt from an article that I wrote for the IET WATW `09. You're right. It didn't get through. But anyway try reading it.
Read more here.. HIIIM
Also read my other article on BC Some Thoughts on ON INTELLIGENCE

Aug 19, 2011

Beginners' nightmare in research

If you're a student who has just started becoming a researcher you probably have good knowledge on the fundamentals of the subject, but not much on the research that has been going on for many many years on the subject. In junior and high schools, much time is spent on teaching the fundamentals. They are not to be blamed since teaching all advancements would be a tough load on them. So naturally, at the end of the course the student has a list of conceived ideas and problems relating to the subject, a part of which he would attempt to solve using his own ideas after starting his research career. The interest in the subject as well as the knowledge on the tools he'd learned would drive him to approach a conceived or existing problem without much reading on the existing literature. Believe me, it is a natural inclination especially after a solution to some problem has been so successful for him. After solving the problem creatively the student either directly submits the solution for publication in a journal or conference dealing with that particular subject or digs the internet for materials to include in the literature review section of the paper. This is the point where he may get a shocking sting. And that is when he finds his very same solution to have already been given by some other author decades ago. "Holy.... What the..." It feels really bad to be in such a situation. The entire work done for validating the solution goes waste.

There can be two situations. The functional architecture of the brain is the same for every person and hence the approach of several persons to solving some problem may be very similar and close to each other. This kind of approach is what would have been used by ancient people to solve seemingly direct, simple problems. For instance, the universal solution found by man to climb to a higher platform is the stair. The ladder is the portable version of the stair. Now, if they had had a patenting system then, it would have resulted in conflicts. Sometimes, this kind of a situation arises after solving something without referring the literature. The other kind of situation is when there has already been a patent or publication of a highly creative solution. It happens.

Sometimes it is quite not possible to look at the entire research previously done on a particular topic and most of the time an idea strikes instantly which urges for immediate validation that people don't feel like searching for instances of the same idea in the literature. The process of research becomes familiar with time. We will change our methodology by making extensive reading on the problem before attempting to solve it. But, it is not wrong or 'wasted' to attempt to solve something without knowing its history or reading the previous solutions. If such an attempt results in a novel solution it is great, but even if the solution is found to already exist there is nothing to worry about for such an experience will be beneficent in some ways. When you find your solution to have already been discovered by someone earlier and has been recognized, then it means you're normal and you have the ability to come up with something on your own that had once been considered novel. This gives confidence in yourself that you can bring out novel solutions to new problems if you try. The next thing is, you will get a chance to find out what you're supposed to add to such a solution that you've found, to present as an entry for publication or patenting. The residue will tell you what subtle or critical aspects you generally miss in your solution or the presentation.

So, it's not too bad to find yourself in such situations, though it will not be very easy to take the sting. Just look at the bright side and move on. If you're not in agreement with the ladder part of the story, that is, if you don't believe that such simple repetitions are made in practical situations, I will post some of my own experiences in the posts to come.

All the above should matter only to a student starting a research career. Professional researchers concentrate on much converged areas and they always try to bring out something very new for which they update themselves with the current research.    

Jul 21, 2009

Simulating the olfactory sense

The blackboard (yes, we have one) is an active area of consciousness in the brain where we create the so-called mental images of the senses. It's the thalamus that I'm speaking about. The neocortex is the center of computation inside the brain that interprets stimuli from various senses, encodes them and generates actions. The structure of neocortex has been studied for a long time, but how it works is still not fully understood. This conforms to the fact that one cannot understand the working of complex programs in a computer just by looking at how the transistors are connected at the lowest level (I'm actually not sure, but I think it will be a monumental task to do so). All the senses from the sense organs (except one) enter the neocortex through the thalamus. It has been proposed that the thalamus does denoising and encoding on the raw data. Also, it performs some kind of voting that selects one of the many hypotheses put down by various areas of the neocortex when interpreting stimuli. The thalamic nuclei are reciprocally connected to the neocortex such that there is feedback between them. Many have proposed that the thalamus functions as an active blackboard which helps us to mentally imagine or more precisely, mentally simulate a specific pattern as if it has occurred. This can happen because since the cortex and the thalamus are reciprocally connected and the thalamus gets the stimuli directly from the sense organs, for mental imaging, the cortex can send a stimuli pattern back to the thalamus so that it will be like the thalamus actually getting the stimuli from the senses. But anyway, they are differentiable. I mean, we know that we're only imagining such patterns and that they are not real.
The following is entirely my proposition. We know that we can mentally imagine an image of any object that we have seen before with ease. For example, imagine a wooden chair. We can rotate, translate or anything inside the brain. Similarly, we can imagine any known voice, example our parents'. We can also mentally picture the feelings of touch as well as taste. This is directly the result of the feedback between the neocortex and the thalamic nuclei. There is one sense left that is directly sent to the neocortex without being sent through the thalamus and that is the olfactory sense or the sense of smell. Now, I think this is why we cannot bring a mental picture of any specific smell in our brain. Why the olfactory sense has been denied this, is a mystery. What would have demanded for such a setup during evolution? We can find it out if we think. Just think.

"This is why we cannot bring a mental picture of any specific smell in our brain. Why the olfactory sense has been denied this, is a mystery."
The computational architecture of the neocortex can be understood only with the inclusion of the thalamic activity which is the center for consciousness and mental imaging. Machine intelligence cannot be fully realized without consciousness. But even the recent HTM (Hierarchical Temporal memory) has not taken into account the blackboard and hence the consciousness. 
The human brain keeps fascinating me so much day by day that I have now started to think why it cannot be God.

Apr 27, 2009

Merely a product of imagination?

You believe that you exist among different individuals of your own species as well as others and that there exists a world which you perceive using your senses. You perceive the visual aspect of the entire world using only the retinal signals captured by your eyes. Now, imagine that you were born blind and you still are. In that case, you should be able to agree that you may not be able to visualize or visually feel any part of the world as it presents itself now. And so is the case with all your other senses – hearing, feeling, smelling, etc. If all your senses had been switched off since your birth you would be like a rock now. From this discussion we understand that it’s our senses that show us this world and make us believe its existence.
We come to the next part now. I think you are aware that you can use your brain to simulate your sense organs’ functions. For instance, close your eyes and try to visualize a chair. You should have been able to do it with no difficulty at all. Even you can feel the chair with your hands in your imagination.  Coming to the hearing sense, you can simulate it as well. Imagine that the chair is falling down and hitting the floor. You should be able to hear it in a way close to reality. For a particular sense, the ‘close to reality’ (CTR) factor will be high if that sense organ is idle in reality. That is, for example if you close your eyes and imagine things, the CTR is high and if you be in a calm, noiseless environment the CTR of imaginatively hearing will be high. You can simulate all the senses in a similar manner, but simulating the smell is quite difficult (close to impossibility in my case). But that will not be discussed now. So, we’ve come to understand that our brain can be voluntarily made to falsely activate our sense organs.
The next part of discussion is to show you that there exist two different things – the brain and the mind. By brain, I mean the physical organ. There is a conceptual entity called mind, which indeed exists, that is so very complicated. The mind itself is composed to two parts – the conscious one and the subconscious one. The one that we use voluntarily everyday and which we can control is the conscious mind. The subconscious mind works by itself and aids the conscious one in several ways. Both of them are interconnected and dependent on each other. The following can help you understand it better. You must have had experiences in which you had forgotten a person’s name or some detail and had tried to recollect it the whole day. But in the middle of a night you would have woken up suddenly after having found that name or detail strike your mind. Now ask yourself how you can get the answer at some time when you were not even trying to search for it. And who did the searching for you at that time? It is obviously something on which you don’t have your full control. And that is the subconscious mind, we’re talking about. Understand that the mind structure is even more complicated than we believe it to be.

To be continued..