Infusing AI into machines

I have been recently reading on machine learning. I found it quite interesting. Machine learning is a sub field of artificial intelligence. As I came across, various algorithms in this field and different applications its became more and more thrilling. These algorithms, though do not give computers the ability to think autonomously, there is great level of automation that could be achieved using these algorithms.

Still there are good number of AI problems that should be solved. Problems like scene analysis and natural language processing still have a wide scope for research. I feel there isn't any comprehensive algorithm for such problems that could be applied in for all cases. Though there are quite some algorithms that work well for specific cases of these problems, i feel only if an algorithm can be applied to all problems of a particular kind then it could be said that the algorithm has infused some intelligence to that machine.

There are some predictions that genetic algorithms, or genetic programming could be the stepping stone for such an algorithm, that could be used to infuse intelligence to machines autonomously. But this is debatable. There are still quite some stuff in this part of science, that i feel isn't science yet. It is still an art. For example choosing a good classifier for a particular classification problem is still a state of art. Such stuff cause doubts, that whether an autonomous AI is a possibility in the near feature.

If such things happen, then lot fictions could become true one day. The MATRIX story of man fighting highly intellectual machines could become true some day. But the future is highly unpredictable. I guess an algorithm to predict the future could be of very high value in the market now. :-)