This is why it's always a good idea to check the questions, but it's misleading to say that all polls are so flawed. And, besides, the entire point is that if you ask the population as a whole you are far more likely to get a better representation of the population than from your own circle of friends and family, which, however wide, is simply not going to do the same job.
Polls are flawed, but they are less flawed than other methods. And there *are* ways of making them tell you something far more useful than you could get by other methods. Make the sample size random, but representative. Make the questions non-leading. Ask a reasonable number of people, but not necessarily everyone. Perform sophisticated statistical techniques to analyse the data, rather than just make it all up. And so on. For the public, the key, too, is to take a collection of polls rather than any individual one: look for trends, pay attention to what different companies are doing, and the like. It's hard work, and there are always limitations, but it's still worth it within the constraints of sampling.
Unfortunately, it seems that there's a conceit of familiarity, along the lines of regarding our own circle of friends as large, diverse, representative, etc, all of which are demonstrably false.