Google is again trying to redefine what it means to search for information on the Web and get closer to delivering on its mission statement to “organize the world’s information” and make it accessible. It may also get us closer to realizing the dream of artificial intelligence, or AI, some researchers say.
In the first generation of Web search, Google’s search engine matched the keywords that people typed into the search box with the words written on pages across the Web. Google’s algorithm, or mathematical formula, then showed the best and most authoritative matches on the search results page.
But there is “ambiguity in words,” said Danny Hillis, an artificial intelligence pioneer who co-founded Metaweb, the semantic search company Google acquired in 2010. “Jaguar” could mean an animal or a car, and search engines don’t know whether a Web page is discussing one or the other. Or if someone searches for “dictatorship in Syria,” the search engine might not know it should pull up news articles about the Syrian government that don’t include the word “dictatorship,” even though they would be useful to the searcher.
Search engines today are like “librarians that don’t know how to read the language in the books,” he said.
But with semantic knowledge, the computer system knows the meaning of ambiguous terms. “It knows that the Redskins is a football team in Washington and knows what businesses are located there,” he said.
When it can do that, the system shows that it can “think about the meaning of things, similar to how the human brain works when it makes thousands of associations at once.”
Among other things, the new technology could help Google to identify more websites that have been able to game its system. Google has said it repeatedly changes its ranking formula because some websites that want to artificially rise up in Google’s search-results rankings have fooled the system into thinking that their Web pages are useful, even when they aren’t. If Google’s search engine understands the meaning of the words on such sites, it could more easily detect poorer-quality content, Mr. Hillis said.
Could the new Google system become the foundation for how robots speak back to humans?
Mr. Hillis said that the founders of artificial intelligence always have believed that “someday computers would have to do this,” meaning to understand the meaning of words and their associations, he said. Now the “pieces are coming together” and Google’s search engine can “take a step toward artificial intelligence,” he said.
But much is still missing. Traits such as “planning, having goals, the ability to reason and make inferences,” Mr. Hillis said. Also: emotion.
What’s Google’s view? Peter Norvig, a director of research at Google and an artificial-intelligence expert, has repeatedly said that the company is more interested building applications that help people accomplish tasks than being able to have deep conversations with talking robots.
Amit Singhal, a top Google search executive, says his “dream” is to build the computer from TV series “Star Trek,” which regularly helps starship crew members do a wide variety of research, diagnose problems with the ship, and locate people, among many, many other things.

The Cosmos News Astronomy&Space Videos