送交者: insight 于 2006-2-22, 19:22:59:
The bottom-up/top-down models, professions, outsourcing and learning
1 The top-down model works on the principle that the answer to a problem is deterministic and there is a finite-step algorithm to reach that answer if we start from certain known premises or initial conditions. However, as implied by the Godel’s incompleteness theorem, if we start with finite knowledge, then we end up with finite knowable predictions. Due to the deterministic nature of these calculations, nothing new can be generated, except that one starts from the known and ends up with the knowable. This leads to the conclusion that artificial processes and systems based on the top-down model, such as axiomatic systems and all man-made machines, including computers, can not generate anything new and innovative.
2 Since the early days of industrialisation, there has been a trend towards using top-down-based machines and processes to replace human muscle and mental powers which are mechanical and deductive in nature. The professions, such as Engineering, Medicine, Law and Accountancy, own much of their extraordinary growth to the top-down model and the deductive power of the left brain. Obviously they can be readily automated by machine-based top-down processes, such as computers and also outsourced to third-world countries as the global communication network becomes well developed.
3 The bottom-up model is based on the idea that nothing is pre-designed and programmed using any finite-step algorithm and the outcome of the bottom-up process can be non-deterministic. Therefore this model is not restricted by the Godel’s theorem in that it can freely generate novel conjectures and intuitions. That is to say, it can generate new knowledge with finite known knowledge, with the help of a formal language. This is basically a inductive process largely involving the right brain.
4 All the innovative and creative processes rely on the bottom-up model, which requires a culturally conducive environment to flourish. That cultural environment must also be based on the bottom-up model itself as a bottom-up process can not be built on top of a top-down process. For example, human consciousness can not be built on a computer that is algorithmic. More often than not, third-world countries are run by authoritarian governments with long tradition of repressive cultures. Therefore, it is highly unlikely that these countries are capable of generating innovation and creativity in all aspects of human endeavors with high impacts. This is shown to be valid if one examines the cultural impacts these countries currently have on the world. As a result, innovative and creative activities can not be outsourced easily to third-world countries on a massive scale. It should also be noted that the so-called “incremental research” in Science and Engineering, which mechanically extends the known to the knowable and predictable, is actually a top-down process and can be easily outsourced to these countries, in an otherwise more advanced and subtly deceptive manner.
5 On the other hand, learning is a process of personal knowledge creation by the learner himself, in which understanding is the key milestone for true learning. And understanding, unlike mechanical information processing, can only be achieved through the generation of mental models and conceptual leaps, itself a bottom-up process. This means that mere mechanical processing of information based on the top-down model without understanding is not true learning. Therefore, learning is a process which is similar to the creation of human global knowledge in scientific research, except that the former only creates personal new knowledge. Therefore learning also requires the processes of conjecture and refutation, as suggested by Popper in his study of scientific discovery. Furthermore the so-called “machine learning” research is a misguided adventure, as a result of the misunderstanding of learning, or even worse, it is simply pseudo science on which large amounts of research funds have been wasted.