 Hello everyone, this is Alice Gao. In the previous video, I discussed four definitions of artificial intelligence. In this video, I will choose one definition for this course and discuss some reasons for this choice. First, I have a question for you. If you were an artificial intelligence researcher, which of the four definitions would you use and why? This question does not have a correct answer. As I mentioned, all the definitions are valid and researchers have studied all four definitions in the past. The purpose of this question is to let you think about the four definitions critically. Pause the video, prepare your answer and your justifications, then keep watching. Here is the quote-unquote official answer. This course will focus on the rational agent definition, building a system that acts rationally. This definition uses rationality as the benchmark and aims to model behavior rather than thoughts. Now that I've given you the official answer, let me give you some justifications for it. Part one, why should we care about behavior rather than thoughts and reasoning? First, rational behavior is more general than rational thoughts. Thinking correctly and performing correct inference is only one way to achieve rational behavior. For example, in real life, we often face questions that do not have a correct answer. Nevertheless, we still have to take action. In this case, correct thinking cannot help us since it does not exist. Consider another example. When we face immediate danger, we often rely on reflexes to avoid trouble. In this case, we do not have the time to think and reason. We must act quickly to survive. These two examples show that rational behavior is more general than rational thoughts. Part two, why should we use rationality rather than humans as the benchmark? I will discuss two reasons. First, humans often act in ways that we do not consider to be intelligent. You may find this funny, but you also know that it is true. Let me recommend a book that I like. It's called Predictably Irrational by Dana Rayleigh. This book describes several ways in which humans behave irrationally, but in a predictable way. It's a fun read and I highly recommend it. Second, rationality is mathematically well-defined and completely general. Since we can define rationality mathematically, we can explore it scientifically. We can develop theoretical models, analyze the models to produce predictions, then perform experiments to verify these predictions. In contrast, if we use humans as the benchmark, it would be challenging to define and scientifically study human behavior. These are two reasons why we should use rationality rather than humans as the benchmark. Let me describe an analogy to help you understand the difference between the two benchmarks. Let's think about how we created the modern aircraft. Approach one, we observe that birds can fly and all of the birds have wings. We can hypothesize that the wings must have allowed the birds to fly. Given that, we can build aircrafts to mimic birds with wings. Approach two, we'll try to understand the principles that allow the birds to fly. This idea led to the field of aerodynamics, which eventually led to the development of the modern aircraft. This analogy mimics our discussion regarding humans versus rationality. Mimicking human behavior is like developing flying machines by mimicking birds with wings. On the other hand, modeling rational behavior is like understanding the principles behind flying and using that to create the modern aircraft. That's everything on choosing one definition of artificial intelligence. Let me summarize. After watching this video, you should be able to do the following. Give some reasons why we should choose the rational agent definition over the other definitions of artificial intelligence. Thank you very much for watching. I will see you in the next video. Bye for now.