MIT Technology Review: IBM's Watson supercomputer used two artificial intelligence techniques—natural language interpretation and deep statistical analysis of unstructured text—to become a champion on the television game show Jeopardy! Since then, the company has commercialized Watson for use in medicine and has made some of the computer's functionality available to developers via a cloud application programming interface (API). Now IBM has announced that it is adding translation, text-to-speech, and speech-to-text functionality to Watson as well as the API. Those functions are developed by repeatedly feeding the computer a massive amount of data and training the software to recognize complex and abstract patterns. This "deep learning" has advanced significantly in recent years but is still very specialized. IBM hopes that combining deep learning with Watson's other capabilities will produce better results.
Skip Nav Destination
© 2015 American Institute of Physics

IBM adds deep-learning capability to Watson Free
10 July 2015
DOI:https://doi.org/10.1063/PT.5.029029
Content License:FreeView
EISSN:1945-0699
Q&A: Tam O’Shaughnessy honors Sally Ride’s courage and character
Jenessa Duncombe
Ballooning in Albuquerque: What’s so special?
Michael Anand
Comments on early space controversies
W. David Cummings; Louis J. Lanzerotti