This is a pretty broad question, could you specify it more so it shows you have indeed interest in the subject? This like me asking what are the growing areas of Databases and expecting a person to educate you on the major areas of research. Maybe give some areas that you are interested so others can answer it more concisely for you. It doesn't come off well when one doesn't do a bit of research themselves before posing a question.
Like the previous post said, AI is a very broad area, nad very few people are knowledgable enough to give advice on all AI. Better ask about one or to sub-areas (ex.: computational vision, planning, machine learning, natural language processing, etc.)
Actually i am very new to the subject so please give your suggestions according to your level of expertise, like if u know about computational vision , so please tell about that....and if some one knows machine learning then please give your response according to that.
Artificial Intelligence is a very broad area and many areas such as Natural Language Processing, Speech Technology, Speech Recognition, Computer Vision and Image Processing use the expertise of Artificial Intelligence to build systems. A few systems with artificial intelligence are Nuance's Dragon, IBM Watson and Facebook's Face detection from photos. There are many more systems which are artificially intelligent. In terms of research, there is research happening in all the said areas and there are many more areas which I have not listed. Try yourself and explore more. Nothing is short of future if you can innovate and come up with a useful piece of work that will help the humanity.
What you may like to think about is why you have come to artificial intelligence (?). What do you think you can achieve with more knowledge in this field? Artificial Life and AGI are important in AI, but what can they do for society? Are you perhaps part of a group that thinks it is not useful, is that why you ask this question (I feel somewhat lost with your question, like D. Page from the University of Manitoba)? If you would like to work on a question that is really hard, think about the intricacies of human-human communication, with its psychological, sociological, philosophical or religious aspects: can it be reduced to speech recognition of automatic text generation? If so, how? If not, why not?
Mobile Health is one.there are many devices that allows you to collect data from people in real life. But the usefullness of these data has still to be proven.
Computational Sustainability is an area that is increasing in importance and has some very interesting challenges. More details at: http://www.computational-sustainability.org
answering your request more with the eyes of a machine learner (which you might not have asked for, sorry!) -- there is a topic I always wanted to work on but never actually found time for it:
The large number of (strong AI) / machine learning solutions available poses a difficult design decision for the practitioner when dealing with a novel source of data.
Surely
* the kind of learning problem
* the underlying data type (e.g. floats/ strings / integers / graphs / images / more complex objects)
* the amount and type of noise contained in your data
* the availability of labels
* the possibility or even need to learn online, to request novel labels over time
* the dimensionality and number of the data points
* real-time applicability of a ML model / availablity of training time
* many more characteristics which I did not mention here
will exclude a substantial number of learning methods, but still a good choice among the remaining ones is hard to make.
Thus it might be worth considering some form of meta-learning over machine learning problems / data sets / models as a prospective future ML topic!