Thought it will be a good idea to share with others and get their opinion on this.
In this article, I will be referring to intelligence as behavior, actions of humans broadly. Humans are no doubt the most intelligent living beings on earth as we have been successful in disturbing the balance of nature resulting in too many humans on earth and many endangered species. So what makes us smarter than the other species is our complex brain. Researchers have not been successful till date to develop a unified theory for brain or intelligence. In the following sections I summarize some of the facts I learnt from literature and my views about the debatable questions.
About the Human Brain
Breaking down intelligence into constituents I believe is impossible. But the ability to store, process, imagine, communicate do contribute to intelligence. The book Phantoms in the Brain by V.S. Ramachandran is a very good attempt to explain some functionalities of brain. The brain is divided into different regions and each part of body is mapped to some region of the brain. Brain is basically a neural network with about 100 billions neurons. The connectivity/interaction between these neurons determines the human behavior. The neural network is not static as the neurons and the connectivity keep changing. The neural network also adapts to things we do repeatedly and so we get better at things by practice. So can all the functionalities of the human brain broken down to logical rules ? Specifically , can we have a mapping from a set of input to a set of outputs? I will say No. I think some part of it is random. Like we may behave differently in two exactly same situations. Cutting out the random part, can we theoretically synthesize intelligence by logical rules? I think it is possible but not practical because the variable space will be too large. In the next section I discuss the approaches to model intelligence on computers.
Research in Artificial Intelligence
The field of Artificial Intelligence tries to make computers work like humans do. The two main approaches to model human intelligence are:
- Knowledge-based
- Behavior-based
Both these approaches cannot independently capture the essence of intelligence. Theoretically, knowledge-based approach is the right way to go once we are successful in simulating the neuron behavior on a computer. But we do not exactly know the working, connectivity of neurons as yet. So designing application specific systems based on the behavior-based approach or a combined strategy seems like a fruitful approach to add some levels of intelligence to computers.
3 comments:
nice post.
very briefly, perhaps we can consider intelligence as the ability to manipulate concepts. This in turn is the basis of learning - it is how one obtains over a period of time, a set response to a set stimulus (not talking of behavior here). If we define intelligence this way, then it is something possessed by not just humans, but also other species. After all, intelligence is the very necessity of survival :)
I consider intelligence as separate from knowledge, maybe more like what one does (action), given the knowledge...but this raises another question: is it possible to be unconscious of the knowledge that dictates a particular action, at the time of performing that action? For example, you can (if you want) but usually don't, tell yourself that you're brushing your teeth at the very moment you're doing so (by 'tell' I mean think and be simultaneously aware of that thought); is it likewise with some/many of the other species? It is easy to include animals in this category because their brains like ours also have a complex network of neurons; but are they capable of 'telling' themselves that they're performing some action at the moment of doing so? Consider the heliotropic behavior of sunflowers...how and why? When did they 'learn' to exhibit this property? When we studied about plants in school, the focus was on the causality of such plant characteristics - plants need light for photosynthesis, therefore those in the dark room will shrink and die while those outside will flourish...yes, that makes sense but what tells the sunflower to move with the sun? Surely, it's getting light even if it doesn't do so!
I think, that when a life-governing system is learning to respond to a stimulus, it involves the use of intelligence. Therefore, it is not a human-specific trait...and it is certainly not just a bowl of interconnected neurons firing in a mathematically clonable manner. It is more than that. Awareness (whatever that may be, since I have not a clue about it at the moment) has a lot to do with intelligence.
Since intelligence (alongwith awareness) is responsible for learning, therefore intelligence cannot be learned :) until the 'learner' (insert your favorite non-human robot/AI agent here) possesses this thing called awareness. Unfortunately, the moment we researchers read this word, we feel the urge to 'mathematically model' awareness :)
It's going to be a while before we have truly intelligent man-made machines.
Hi Harsh,
That line about mathematically modeling awareness caught my eye.. :) Yeah I am never clear of the terms consciousness, unconsciousness , awareness etc etc
Well yes for sure we are way too far from modeling human intelligence, and I also feel happy about it.. wonder how it would be machines could think like humans ..
@Harsh,
Forgot to mention, I was expecting to read some of your "pravachan" at your blog:)
Post a Comment