Nonsense-machine intelligence
Human Thinking has rich backgrounds and is conducive to solving problems with rich backgrounds. Machine thinking is based on several computing components and the logic for solving the current problem. The solving logic is responsible for organizing the running order of these components. People are responsible for expressing this solution logic.
The specific process of thinking generally goes through the stages of analysis, synthesis, comparison, abstraction, and generalization.
The process of thinking is expressed by the specific forms of concepts, judgments, and reasoning. You have me and me in this specific form of thinking and the previous thought process. The form is easy to grasp. The "commands" we can send to machines tend to be "specific forms ". Language is a form. Language always tries to describe this abstract world. It is costly to describe a world in terms of concepts. Therefore, machines use existing concepts, specified judgment logic, and reasoning processes. For example, for the reasoning sequence a-> B-> C-> D, to launch D, people only need to tell machine C and the reasoning steps from C to D. Machines do not have information about B or even. The direct reason that the machine does not know a and B is that no one tells it. The root cause is that description B is more complex than description C, and description A is more complex. Although the machine only needs to know the information of C for a specific problem D, the problem is much simpler, but in the inference sequence of A-> B-> C-> D, rich information is lost step by step. The work that can be done based on C's information is already quite limited.
-- 2009-4-0