Google Inc (NASDAQ:GOOG)’s interest in artificial intelligence was apparent considering the organization’s acquisition of firms such as DeepMind, Boston Dynamics, Flutter and Nest Labs.
The company recently released a research paper titled “A Neural Conversational Model” detailing how its artificial intelligence venture shaped up. Google researchers Oriol Vinyals and Quoc V. Le were the authors of the paper published in Arxiv, a leading repository for academic research. The firm’s researchers programmed an advanced version of “chatbot” that learns how to answer questions as per examples stored in it. The bot does not simply output answers from a database. It can form new answers from new questions. The machine also can have different types of conversations including those about the meaning of life.
Google’s researchers queried the bot about everything ranging from monotonous IT questions to questions related to life. As per Le, the machine has the unique capacity to analyze existing conversations. A machine learning approach is used rather than using rules to build a conversational engine. The machine learns from data instead of hand-coding the rules.
It harnesses neural networks that are information systems that mirror how the neurons in the human brain are interlinked. Even though this concept has been existent for a while, it has been implemented only recently, thanks to a huge jump in computational power.
One flaw of the model is that it only delivers simple, short, occasionally unsatisfying answers to questions. Also, the model lacks a consistent personality.
The machine response were sourced from a database of movie scripts in which it utilizes logic and reasoning to give suitable answers to the questions that are being asked from it. Researchers have leveraged cognitive computing that involves teaching computers how the human brain functions.
Sources: businessinsider.in, ubergizmo.com
