This article has been excerpted from the original which was published in full at Forbes.com
on September 7th, 2012 by Trefis.
International Business Machines is trying to put a voice-based analytics engine powered by Watson in the hands of smartphone users. The aim is to drive its business analytics division to $16 billion in sales by 2015. Watson is an artificial intelligence system capable of answering questions posed in natural language. It was developed by IBM’s DeepQA project and is one of IBM’s top supercomputers. The offering is likely to resemble Apple’s Siri, but on major steroids, for corporate users. IBM hopes to answer complex questions with its application that focus on finance, health care and telecommunications.
Machine Learning and Power Consumption Key To Success
Analytics that run on a supercomputer consume a lot of power and one of IBM’s biggest challenges is to deliver Watson’s processing power on smartphones. Even with the computation occurring at a data center, a Watson smartphone application would still consume too much power to be practical today. Watson is also based on machine learning and to master a subject, it takes a significant amount of time and data.
For example, researchers are trying to “teach” Watson oncology by feeding the machine with answers to relevant questions. Then, to make the machine “understand”, they ask it to answer similar questions by analyzing documents, websites and books at 66 million pages a second. After several iterations and teaching the machine what answers are accurate, it is expected that Watson will develop enough expertise to assist a doctor in a practical situation.
IBM also has an image recognition software which can be added to the analytics app in the future. It already uses image recognition in a retail analytics application and can use this expertise to add eyes to Watson.
In a potential health-related scenario, a patient can access Watson through a smartphone device to explain his or her problems and symptoms. And based on the symptoms, Watson could provide an initial recommendation as well as access the patient’s medical records to make adjustments to the answer depending on factors like pregnancy or diabetes. It could also provide information about nearby medical facilities, if needed, and contact healthcare facilities.