Commonsense knowledge problem

Imagine you are a child and you know a lot of things that no one really taught you, like that ice cream is cold or that water is wet - this is called "common sense knowledge". It's the stuff that seems obvious to us because we have experienced it in our daily lives. However, when we try to teach common sense knowledge to a computer or machine, it's not so easy.

For example, imagine you are trying to teach a robot how to make a sandwich. You might show the robot a picture of a sandwich and explain the steps: bread, meat, lettuce, tomato, bread. But how do you explain what "bread" is? How do you explain that you need to use a knife to spread mayonnaise on it? These are things that we take for granted because we have years of experience making and eating sandwiches. But for a robot, it's like trying to explain colors to someone who has been blind their entire life.

So the common sense knowledge problem is the challenge of trying to teach machines the things that we as humans know without even thinking about it - how to tie shoelaces or pour a glass of water, for example. It's a hard problem because common sense knowledge is often implicit - we don't even know we know it until we need to explain it to someone else.