Google hopes to answer complex questions
Google makes searching the internet easy with its simple and user-friendly design, but the popular search engine is far more complicated that it appears, with intricate algorithms behind every search and intricate software such as Machine Learning underneath every advertisement and Google+ friend suggestion.
Yet Google is unable to answer complex questions, something its developers are working diligently to achieve, said Google Fellow Jeff Dean in a recent interview with The Puget Business Journal. Dean outlined some of the problems he and his colleagues are trying to overcome at Google Headquarters.
Dean explained that Google is aiming to be as accurate at answering questions and meeting demands as a human, but the search engine is not yet there. He offered booking flights as an example. He explained that if he said to a human, "Please book me a trip to Washington DC," the human would answer with further questions regarding lodging, layovers and other details of the itinerary. If a search engine is to answer the question as thoroughly, it needs to break it down into smaller chunks.
The problem isn't that Google doesn't have the correct answer, it's that the search engine has hundreds of answers that are correct to varying degrees, making it nearly impossible to sift through all these answers and determine which is the most suitable.
Google is the world’s most popular search engine, with over five billion searches per day, according to Statistics Brain. But at this point, Google only understands words, not sentences or phrases. Dean hopes that one day Google will be able to understand sentences on a "deeper human level", and even be able to provide answers from multiple sources.
It might not be long. Google can now answer straightforward questions that it has stored on its database. For example, if you typed in "the President of the United States", you would get the answer "Barack Obama".
Dean pointed out a few more interesting features his team is working on, such as voice recognition, which could one day lead to voice-activated searching. They’re also working on using image recognition for Google Street View, trying to recognise street numbers that they could then place on Google Maps.
Posted by Dylan Brown