Okay let’s say that you’re walking down the street in the night and you suddenly hear an alarm and then you look across the street and see a man in a mask climbing out of a jewelry store window with jewelry in his bag. So what goes through your head? There are 2 different possibilities and let us see which one the human brain recognizes as true:
A) Well you assume that the man is a thief, it hardly takes time to do that but you arrive at the conclusion that he’s a thief.
B) There’s an equally plausible case that the man is the owner of the shop and had gone to costume party dressed like that and while returning home he saw his store window broken, hence lacking his key he went in through the window to investigate and then decided to protect his valuables by carrying them home in a black bag.
So which one is true? A or B, you will definitely choose A; why? Is the decision you have just made logical? No it isn’t either of the cases are equally likely to be true and there can be a 100 other scenarios for the same event coming up with an explanation which points out that the man is innocent, so why did your brain say that A is true? It’s due to something called world knowledge or experience you have learned over the years that such a person is most likely to be a thief as its an observation you have made over time.
Let there be an event E which is your observation of the man. Now, C is the conclusion that he is a thief (option A), you know through experience that if E is true then C is true. This is sort of reasoning is called deductive reasoning where the validity of one event is based upon the result of the observation made for another event. However there is a fine grey line over here, there simply isn’t 100% certainty in our daily life. So how do we deal with it? Well we use a diluted version of our logic in our life i.e. if the observation E is true then C/A ought to be true.
This is what common sense is all about; making inferences based upon data from observations and a repository of data tucked away somewhere in the brain. It deals with logic, probability and statistics but the key element is the data, it doesn’t matter how refined a technique you’re using when you don’t have data to use it upon.
This is the greatest problem facing AI today; how do you feed something like this into a machine?