So far IBM (NYSE:IBM) hasn't managed to join the three major cloud service providers - Amazon (NASDAQ:AMZN), Google (NASDAQ:GOOG), and Microsoft (NASDAQ:MSFT) - but that could change as the Artificial Intelligence ("AI") technology of IBM Watson heads to the cloud.
IBM announced the addition of five new beta services to the Watson Developer Cloud: speech-to-text, text-to-speech for both English and Spanish, visual recognition, concept insights, and tradeoff analytics. The five new services join the family of available Watson services for "cognitive apps" developers, which include language identification and machine translation.
Watson technology has been making headlines since 2011, when the AI system outperformed the best two players in the world, Brad Rutter and Ken Jennings, in a Jeopardy! contest. Jeopardy! was chosen to demonstrate the capabilities of Watson because it isn't a narrow and well-defined task like most current AI applications, but on the contrary involves most of the subtleties and intricacies of natural language processing.
The knowledge base of Watson isn't hard-coded by programmers, but assembled by automatically reading Wikipedia and a wide range of natural language documents, and combining a huge number of individual inferences into factual knowledge. The first commercial applications of Watson, in 2013, have been for assisting health care providers in patient treatment decisions.
The automated machine learning technology of Watson is at the forefront of AI research, but IBM is extracting from research results practical techniques that can already be used in commercial applications. For example:
Speech to Text is a cloud-based, real-time service that uses low latency speech recognition capabilities to convert speech into text for voice-controlled mobile applications, transcription services, and more. Transcriptions are continuously sent back to the client, and retroactively corrected as more speech is heard, helping the system learn. The service is based on more than 50 years of speech research at IBM. It uses state-of-the-art algorithms based on convolutional neural networks or "deep learning." Using these algorithms, the Watson team has published the best accuracy results (10.4% word error rate vs. 12.5% for the second best as of today.
Using similar deep learning methods based on state-of-the-art neural network technology, the visual recognition service classifies input images by assigning them to categories according to probability scores. For example, a picture of a sport scene can be included in the soccer or baseball category according to the respective probability scores computed by Watson image analysis. Automated recognition permits not only to organize large collections of digital images but also, for example, to analyze and cross-link consumers' preferences based on online pictures.
In related news, Watson AI technology will power Verse, the recently announced IBM cloud email system and collaboration suite, now tested by selected enterprise users and soon to be open for a public beta. Verse will be a competitor to Google Apps and Microsoft Outlook.com, free for individual users and small businesses while large companies will pay depending on the number of users and the volume of data. MIT Technology Review reports that after the public launch, expected at the end of March, Verse will include a personal assistant based on Watson technology, which will learn how to categorize email based on importance.