Google usually returns pretty good search results, but only if there are pages that include the words from your query or some synonyms. Google doesn't understand the query, it only tries to match the words from that query. Sometimes Google can answer questions like "How tall was Albert Einstein?" or "What's the real name of Al Pacino?", but it can't go beyond simple facts.
Amit Singhal, Senior Vice President at Google, says that Google works on building "a huge knowledge graph of interconnected entities and their attributes". Freebase was just the starting point: Google's knowledge graph has 200 million entities, while Freebase only has 22 million entities. The graph is actually an encyclopedia with structured information obtained from the web. This will help Google understand your queries, provide answers to complex questions and find more relevant results.
Right now, Google only uses the graph to show a list of related searches for singers, actors, painters etc. As Google improves its infrastructure, the knowledge graph will be used more and more.
Death E-mails Google - New Yorker (blog)
3 hours ago