AI and Robots: What is AI (1)

This is a make up post!

The image you see here is from the famous chess match between Garry Kasparov, a world champion, and Deep Blue, a super computer built by IBM, that took place in 1997. The computer was able to "beat" the world champion. Despite the argument that this match was not "fair" to Kasparov, at least there's one thing we could all agree: Deep Blue certainly showed signs of Artificial Intelligence. But, what really is Artificial Intelligence? Is it simply intelligence created artificially?

In order to understand what is Artificial Intelligence, we might have to first define intelligence. But what is intelligence? Do the bacterias on the keyboard I am typing on have intelligence? Are the rose bushes in my garden (which adds more to my yard work load) intelligent? How about the butterfly sitting on the rose pedal? How about the planet called earth we all live on? Different people might have different answers. I don't want to get too philosophical here, so I'll simply define intelligence as the ability to reason and learn. I know this is still rather vague. Under my definition, all those things I mentioned above could still be categorized as intelligent beings. If you are still not satisfied, you are also welcome to read and modify the wikipedia page on intelligence.

So how should we define Artificial Intelligence then? Again, many people would give you very different definitions. Russel and Norvig summarized all the different definitions into four categories in the book "Artificial Intelligence: A Modern Approach":

Systems that think like humans	Systems that think rationally
Systems that act like humans	Systems that act rationally

The top ones focus on the ability to reason while the bottom ones emphasize on behavior. The left ones relate AI to human performance while the right ones only measure rationality. All these categories have their merits, but I personally lean toward the last one: Systems that act rationally.

When we try to create Artificial Intelligence, it is easy to try to model after human beings. Why? Because first of all, human beings are intelligent beings (arguably, there are stupid people too). Furthermore, we certainly understand ourselves easier than say, the white mice used in scientific experiments. We try to understand our reasoning and logic behind our thinking and behaviors, and then try to apply the same kind of ideas to an AI agent and have the agent mimic us. There are certainly still many things we don't know about ourselves, and the research in AI is actually a great way to try to understand ourselves better (individually or socially).

It is necessary here to mention the famous "Turing Test". Alan Turing (shown in the picture on the right) is often considered to be the father of modern computer science. The "Turing Award" named after him is considered the Nobel Prize in computing. In a 1950 paper he proposed an operational definition of measuring artificial intelligence: The computer passes the test if a human interrogator, after posing some written questions, cannot tell whether the written responses come from a person or not (Russell & Norvig). You can think of this in terms of a chat window. If you think who you are chatting with is human, but the other entity chatting with you is in fact a computer program, then this program would have passed the "Turing Test". (You can check out this chatbot if you know a little bit of Chinese.) Interestingly, some "flirting chatbots" are reported fooling lonely Russians into giving out their financial information. Can we say these AI agents passed the "Turing Test"?

There are other intelligent species on earth too, and many times we learn from them because they might do better in certain areas. Many AI researchers also get inspired by biological beings and develop AI algorithms accordingly to solve problems related to human. In my opinion, AI is really the study and expansion of human intelligence.

However, human also do stupid things. We pollute the world we live in, we destroy forests, and people get killed in wars or genocides. Sometimes we are also irrational; we let emotion take over and let that affect our judgment. Therefore, the rationality approach to the definition of AI has good reasons. So what is thinking and behaving rationally? Let me give you two examples and then you can decide yourself.

The first example comes from Russell and Norvig's book. If you see someone you know across the street, you look to the left and look to the right and made sure there is no traffic nearby before proceeding to cross the street. Meanwhile, at 33,000 feet, a cargo door falls off a passing airliner and flattens you before you make it to the other side. Have you acted rationally?

The second example comes from the novel "I, Robot" by Isaac Asimov (made into film in 2004 starring Will Smith). In this story, the robots came to the conclusion that in order to protect human from self-destruction, it is necessary for robots to take over. Have these robots acted rationally?

[To be continued...]


[Quote from the I, Robot movie (2004)]

Detective Del Spooner: Human beings have dreams. Even dogs have dreams, but not you, you are just a machine. An imitation of life. Can a robot write a symphony? Can a robot turn a... canvas into a beautiful masterpiece?
Sonny: Can *you*?