I think my road for research is not very smooth. When I graduated at 2009 from Nanjing University, I spent two years at CMU. At that time, actually I know that I do not know anything about doing research. What I have by that time is some amateur interests and passion for my AI dream, building a machine which can have the same intelligence as human being. However, I have no idea about how to really solve the problem. At first, I was very proud of myself, and think that I can make some big contributions soon. Later, I realize that what I have at that time is purely interest, but no foundation to support such interest, and do not actually have the action to make it real.
I start doing research in the computer vision field, which seems to be very interesting to me because I am awed by how the human eyes can enjoy the visual world so abundantly. However, when I go deep into the research field, and know the state-of-the-art method for dealing with the problem, I was somehow disappointed because the field can only achieve the understanding of a baby. It can never achieve the level of human being. One more thing made me sick at that time is that I notice that all the papers are about mathematical equation. I had some thought by that time, does our brain need so many mathematics to achieve our common skills, such as vision?
Actually, my longest dream is to build a real AI machine. My philosophy and belief is that I can only build such machines by mimic how the brain works. Neural network seems to be the first choice. The basic unit of the model is one unit which is a sigmoid function of the input, which actually does the very similar thing as the neurons in our brain. The hope is to be able to train the whole parameters (the weights within the connections) for the big network. However, this seems to be too difficult to solve. Until recently, Hinton invented the Deep Belief Network, where he can train the network layer by layer and construct such huge network by stacking many Restricted Boltzmann Machine from one to the other. The framework seems to open a new window for the AI dream. DBN is actually very similar to how the brain works, and it actually has been proven to be effective, for example, it has beaten the SVM on the raw input pixel value of the digit database MNIST.
Anyway, these two days, I spent some time, reading related papers, listening to the talks given by those pioneers who are working in the DBN, it is really exciting to embark on this new land. I wish I can really spend some time, learning some real stuff, but not just junk. Let me learn things just out of my interest. I am on the road of a PhD. If that is so, do everything I can to understand things well enough.
There is much to learn, but I know God has given me wisdom and He truly has blessed me and make my learning be so effectively, I will praise Him and sing song to Him because He is worthy of that. Thank you Lord! I am going to explore the world that you have created, give me a humble heart in front of you and give me a thankful heart also, because I know future is in your control. May glories be all to yours! Amen
