For your sample, pick.. let me say... 30 people. You got 30 people but you did not know they were all KKK members and now your pool says that everyone in the US is racist! Let's try to improve, but I can not have an interviewer in each city so I do phone interviews. Now looks like it will work, but in my country a good deal of people do not have phones. Oops, now my pool sees just a biased portion of the population (excluded because they have no phone). Let's try once more, spend more and put an interviewer in state capitals. But still I'm getting a biased sample because it represents only the opinions of those living in capital cities (and they have a different view of living in the countryside).
Short version: Do not come insulting me wanting to force me to believe that a sample of half a dozen people is enough to deduce the opinion of a whole country with millions of people with the most varied possible opinions.