Today searches and search advertising relies on keywords, but that will change. Google makes its revenue from bidding on these keywords, but changed in search and search advertising will force the change as the media moves from keywords and into the world of natural language. "Keyword bidding will go away, replaced by bidding on natural language queries," Mohajer said.
He said today a query responds to the word "hotel," but it also can simultaneously respond to other words in the same natural language query -- such as in the sentence "Show" me "three" hotels in "Seattle" for "Saturday," "June 13." And serve the information to the user based on previous queries for an airline flight to Seattle several days prior.
Mohajer has no intention of leaving this bidding advertising platform for Google and Bing to build. He said "you can assume we will build many layers over time, and have partnerships along the way."
Some of the partnerships will connect inanimate objects that recognize search commands to make computational decisions, he said. An espresso machine will make a drink based on a voice command, but it also will let you know when the system is out of milk to make the latte. The same machine will provide a weather update, provide the status of an airline flight and road conditions to the airport.
Hound processes very long and complicated voice queries such as "What is the population and capital for Iceland, and Italy, and the area in square miles for each?" The answers: The population is 389,904 in Iceland, and 60,340,328 for Italy, and the capitals are … .
On my request, Mohajer asked Hound, which is only available today on devices running Android, "Where were Abraham Lincoln and John Lennon born?" in an effort to ask an off-the-wall question the technology might not have heard in the past. The answers: Hodgenville, and Liverpool, respectively.
Mohajer had a vision. "If we spoke to computers they would talk back to us," he said, but financial investors wanted to see a product in less than 10 years. So the team, at Stanford during this time, built the music recognition system Soundhound as a jumping off point to better understand natural language through voice search.