Watson is a traditional AI（symbolic AI) machine. It has a little language capacity (called syntax), so be able to find out what is the subject and objective in a sentence. It searches for the subject to match the objective through its huge database. It has a guessing mechanism (probability) built to deal with uncertain answers. Its advantage is its huge database. Human semantic memory is relatively too small to compare with the huge database.

But Watson comes with the fundamental flaw of symbolic AI. Its does not really understand a question and what its own answer really is about. A good example is: a answer by a person next to the machine is already proved wrong, Watson gives the same wrong anyway for the same question. It processes symbols (words) only, and has zero capacity in process patterns or simple pictures (circle or triangle). If in Jeopardy, the question was presented in picture (car, apple, whatever), Watson shall get a score of zero.