This is what I think is the big challenge in achieving ArtificialIntelligence. Making computers do stuff that we (HumanBeings) do without thinking.
WilliamGibson in his book Idoru, quotes an unknown source (paraphrased) If AI is ever achieved, it will come from the direction we least expected it. I think it may be achieved through adding small pieces of ArtificialCommonSense until someday the whole gets bigger than the sum of the parts. -- AalbertTorsius
It would be interesting to find out how much ArtificialCommonSense has been programmed into Spirit and Opportunity Mars rovers - Spirit was able to reboot itself multiple times and re-establish contact with earth after its MarsSpiritSoftwareProblem. They have to function highly autonomously, only receiving commands from earth a few times a day and that with many minutes delay. The environment may seem like just rocks but there is radiation, chemicals, temperature, duststorms, unpredictable terrain - all kinds of variables to stay on top of while carrying out a complex mission. The coding involved must be brilliant. Just speculating here but one of the reasons for the Flash memory problem could be more priority was given to ArtificialIntelligence and CommonSense reasoning than more mundane things like memory and file management which could be assumed to be taken care of by the VxWorks OS and standard code libraries. I did read that the memory system was given relatively less hours of testing.
See ISBN 0893912131 Formal Theories of the CommonSense World
A real-world implementation of ArtificialCommonSense is the CycMergedOntology. See also SummaryOfCyc.
See also GuiAsConversation