Poll have proven to be more reliable the closer the date of an election event. Voter polls more than three months away from any election should be considered more of a snapshot in time rather than a prediction of which way voters will eventually cast their ballots. Public polling data is one of the most important sources of information that journalists can use to create a reporting narrative, and that campaign managers can use to broaden a candidate's appeal.
Polls have historically been very good at predicting political elections, as long as they are done properly. The exception was the declaration that Dewey defeated Truman, but that was based on bad survey construction, not using surveys as a methodology. They are used across types of sciences and have proven to be good.
Polls represent a small subsection of people. They represent people who don't mind being polled and who happened to be at the right place at the right time to take a poll. Plus, pollers can purposely poll people in a particular area that may be inclined to vote a certain way.
I think polls give a small sample of how people will vote. I think they need to be taken with a grain of salt. These polls don't necessarily cover all demographics or enough people. They also can't account for people changing their minds from one week to the next. I think polls are fun, but they shouldn't be taken too seriously.