As far as I know, computers can only be as smart as humans make them. We can program them to "think," as we have when creating a super computer that can beat the best of chess players. However, if you ask a chess playing programmed computer to do math, it will be unable to compute it if it does not have the right software.
Even computer AI characters don't think in the common sense. They are programmed to do certain actions under certain situations. By no means is this reason; it's all in the code. The code however can mimic thinking so that from the user's perspective, it looks as though the AI character is thinking.
Why is this important? Well, considering that I'm making a game (based off the lego labyrinth), I will eventually have to create AI characters. Luckily, I'm using a program called Virtools that allows me to use various scripts to create characters, so I don't need to code from scratch (I only know basic Java, HTML, and a little bit of C++).
Simulating human thinking can be tricky though since no human is ever completely correct. Sometimes, even the best thinkers make mistakes. However, what level of mistakes is considered human-like? Make a computer character do too many mistakes and the player will think the AI is flawed. Make a character be too smart and the game will become impossible. I guess the only way I'll be able to find the balance is by simple trial and error, which is also time consuming.
Gah, too much thinking, I'm only human.