Non-scientific Thinking in Artificial Intelligence


Just for record, this entry is just a non-scientific thinking result based on the course I am taking. So, why not I begin with some comments about that course.

I am doing research work on AI like half year, and to be honest, I did not read any reference book in details, including my favorite one: Artificial Intelligence: A modern Approach. However, after someone did some work in some area, he would have his thinking and arguments on this subject. And unfortunately, me too. Let me start from the definition of AI. Oh, wait, I should mention that the course I take is called Principles of Artificial Intelligence.

Obviously, in different book, there are many similar definitions of AI, and actually, the differences is about the purposes of the certain book. For example, like AIMA, as a reference book, it should build a environment and symbolic system as cornerstones. In contrast, if the book is just for introduction, it is no necessary to make it too academic, the work of that book is showing examples and explaining. Thus, since it is a university course, it should provide a symbolic system as tool. Otherwise, if there is no well-defined foundation, the course eventually would become a product exhibition or a bunch of concepts. AI is not like neither mathematics or physics, it is a new scientific area even though it has serval decades history. Thus, all the history should be clear. Where is AI born? Why is connected to Neural Science and Neural Network? This roughly half century is definitely a fascinating story in whole science history, rather than a timeline. Furthermore, about religion, the words like “we are creations of god” should be avoided because the speaker should stand as where a scientist does. The pure knowledge and academic discussion is very important, because even though people have different religions, they still admit that science is right. Actually, how the religion said is more about philosophy part. And philosophy is the pure evil ghost who makes all science subjects complex. For instance, philosophy provides a uncertain answer when physicists are thinking about universe. Philosophy also makes mathematics more difficult because of logic and natural rules. And for AI, what philosophy does is just mixed everything up about “understanding”, “learning” or “thinking”. Definitely, AI should consider this part eventually, however, before the knowledge and experience on AI and technology is enough, philosophical thinking just makes the situation worse. Alright, the comments about the course is enough. For now on, I am going to share what I think on AI. And again, just for record, it is non-scientific thinking.

A common observed situation in Artificial Intelligence area is

Given a environment \(E\), and a intelligent agent \(A\). \(A\) receives a sequence of perceptions \(\mathcal{P}\) and acts the appropriate actions \(\mathcal{A}\).

The design of the model as above is following the experimental rule: stimulate–react. The problem is scientist cannot touch the most important part, which is “processing”, the way convert perception into action. This limitation leads that there are so many possibilities on so-called “processing” part. And that is why neural science is involved in AI, a group of scientist believe that if they can understand how the human brain works, we can simulate it via “artificial” neurons. And lots of research work shows that biological algorithms are stronger than mathematical-based algorithms in many aspects.

Is the human brain the ultimate brain between all creations? Yes, it is. However, is the human brain beyond the natural laws? I think it is not. Furthermore, the two cornerstones of science: mathematics and physics can describe and explain everything theoretically. There is a possibility that the biological algorithms we followed is not the most natural way. And I am the one who believe in this. I have read a book named “The Future of Artificial Intelligence”, and finally, the author shows how the neural science and his method is powerful. But deep inside, I believe these science can be represented in mathematics naturally.

The reason why I insist on, because I believe in Science, Nature, and what Mathematics give and prove in recent thousands years.