Sunday, November 8, 2009

Brave New World

The computer revolution in late twentieth century has raised the possibility of (so far) unthinkable proposition. "Can the artificial intelligence simulating human brain be built?" Though computers can outsmart humans in brute computing power, humans still beat computers in finding solutions in a complex situation, like in a game of chess. Can we build computers with lesser computing power which have 'humane' intelligence?

The field of artificial intelligence has advanced so much in recent years that we are led to think that 'humane' intelligence may indeed be possible. The traditional computers work in hard wired way. They just do what is told to them in a step by step manner (remember those algorithm flow charts?) . If the programmer misses out even a minute detail, they get lost. But intelligent computers (or programs) are built in such a way that once a mistake is made it's remembered and alternate paths are sought. Every time the goal is reached the path is reinforced. It's like throwing a stone at a mango. The traditional computer have an equation built in by the programmer to calculate the force and direction based on the weight of the stone, and the height of the mango. But AI learns the trick by initially making wild guess. At successive throws, if the stone gets near to mango, the action gets reinforced. So it learns by trial and error just like a human child. If a wind blows from any direction , the traditional computer keeps on making bad throws, but AI quickly learns to adjust the throws. Computers already exist, which can learn this way. It's only matter of time when they 'learn' more complicated tasks. They will also have to be provided with better means of input of data, to serve as sensory organs.

It'll be interesting to watch these AI's grow and see what characteristics they develop. They will have their sense organs, different from ours and other tools such as lasers, ultrasound probes etc. Will they develop into an individual? In other words what characteristics are of an individual? Will they have a sense of 'Me' or of an integrated continuous being? I think they will develop such a notion sooner or later. They develop a notion that the collection of processors, sensors and other mechanical parts that make them is a collection they have unique and immediate access to and will (with the help of others referring to them as yours ) call them 'his'. Can AI have a mind? This is more difficult to answer because mind, by some scientist is considered more than the manifestation of neural network, and it involves the property of material itself, and is influenced by quantum mechanics. But certain characteristics of mind have to be noted That it doesn't reside in any single part as a single entity. It's also a collection of thoughts in different parts of brain. I may 'change' my mind when a different part of brain comes up with a more calculated solution. So I think when the AI acquires a sense of individualism itself, it'll have a primitive mind which will gradually develop.

How will AI' s thought be? Our own thoughts are unlike the lines in a book, fragmentary. If we are recollecting a tour, we think randomly and incompletely about the boat, the smell of the food, the noise etc. Can we make the computer think randomly like this? Traditional computers think in well formed complete sentences. But the formula to make an AI is its ability to think 'parallel and with 'fuzzy logic'.

Can they take over from Us?

Obviously. Thinking intelligent AI's will also be able to capable of building another AI. So they will build a community of AI's. They will be a better and more powerful species than us. Unlike us, each generation can design a superior AI for the next. So their evolution will be faster. And they are going to replace humans on earth. It sounds like a terrible thought. But we can consider AI's as our own children though not biologically. They are our intellectual creation. In the evolution of life, this will be another revolution, since the invention of sex ( 1 billion years ago). AI's in a broad sense can be considered living. Living beings need not necessarily have the proteins, & all those stuff, the only prerequisite is that they should make copies of themselves. AI's will make other AI's though not exact copies. 'Mutation' will be frequent event & giant leaps forward in evolution will be an occasional event. For first time the direction of evolution will be designed by purpose.

What purpose will AI have in mind in living and making new AI's ? Or what will be their purpose in life? Let's examine what is our purpose of life. All religions will come forward and say- to go to heaven. We live this life as an intermediate step or a qualifier round. Though it's a fancy stuff, all is not well in heaven too. We go to heaven, enjoy all comfort there, live there, for what? So the eternal question remains unanswered. Science neither answers the question. Every process in the universe goes on without a purpose, design or direction. Purpose is an invention of mind which came much later than the universe. Evolution engraved it in the mind of living things because only those with a purpose and will to live survived and passed on the genes for same qualities. We have to instill a sense of purpose in AI too so they will struggle hard to survive.

Should we be alarmed at these species.

Definitely not. Our ethnocentricity prompts us to view all other powerful species with suspect. I actually regard AI as our own children. not biological but intellectual. It's like human species taking a sudden mass mutation & changing its form. If we name AI's Roboman, the alien feel is considerably reduced.

So is it going to be true?

Probably. but thinking about such a probability itself changes our perspective about life, purpose, and future. I hope this dream come true, AI's build a bright new future, design successive generation with increased intelligence, good qualities and devoid of any evil tendencies. They 'll make a heaven on earth- a long unfulfilled dream of all men.