With the recent introduction of Watson Explorer, a set of data exploration and content analytics capabilities that organizations can apply against their own data, IBM is in effect pioneering the development of APIs that continuously learn.
John Gordon, vice president of the IBM Watson group, says that as data is added to an environment, IBM Watson will continuously invoke Watson Explorer via the APIs that IBM has developed for the Watson cloud service.
Intended to be a bridge between cloud-based services and proprietary data that often runs on premises, Watson Explorer provides a virtual data mechanism for employing natural-language queries against existing data without having to physically move it into the cloud, Gordon says.
Watson Explorer is available in an Enterprise Edition that provides search and 360-degree information applications across multiple sources, as well as the ability of qualified customers to license cognitive services from the Watson Developer Cloud. An Advanced Edition includes all of the capabilities of Enterprise Edition plus content mining and content analytics.
IBM says that as it develops new Watson APIs, they will be exposed to developers participating in the IBM Watson Developer Cloud. Current Watson APIs provide access to Question Answer, User Modeling, Relationship Extraction, Message Resonance, Concept Expansion, Language Identification and Machine Translation services.
Gordon says that developers of Watson applications need to think in terms of teaching their APIs how to explore and discover relevant data, test them, then create a feedback loop through which Watson can then learn not only what new data has been added to the environment, but also how that data changes the context of other sets of data.
Gordon says 11 Watson applications are in production, with 100 more in flight. All told, Gordon says, IBM has commitments from developers that put the total pool of Watson applications being built at 3,300. That includes an investment in Pathway Genomics Corp., a clinical laboratory that offers genetic testing services globally, to help deliver a cognitive consumer-facing app based on personal genetics that IBM announced this week, as well as a recently announced set of oil and gas exploration applications being developed by Repsol S.A.
Next up, Gordon says IBM is planning to add support for video and audio data types alongside existing text analytics capabilities. In addition, IBM is planning to add support for a variety of data types associated with Internet of Things applications.
There are, of course, a number of advanced machine learning platforms in various stages of development in the cloud. But in terms of overall richness of the developer environment, IBM Watson cloud services running on top of the IBM Bluemix cloud integration platform appear, for the moment at least, to be the most mature.