IBM announced recently that they are making it easier for developers to jump into machine learning. At the World of Watson conference, the IBM Cloud-based Watson Data Platform with Machine Learning was launched. At the core of it all was the concept of the interconnectivity of data made possible through APIs.
IBM Lead Architect for Streams Product, Mike Spicer, summarized the advantages during an AMA session. Watson, he said, “brings in historical data, builds a picture of what’s happening in the real world, and applies models to predict outcome in real time. All within 10 milliseconds.”
David Kenny General Manager, IBM Watson, called Watson, a “cognitive assistant” and clarified that it is not Artificial Intelligence, but an Intelligence Assistant. Because, he said, Watson analyzes data but does not draw conclusions. Man and machine, not machine replacing man. Humans are still needed to make decisions.
Speeds are startling. Watson can suck data into the cloud at 10 gigabytes per second. The Watson UI allows collaboration between data professionals, but the company claims that it is so easy to use that non-engineers can use Watson to visualize and share insights. And Watson’s value grows over time; as teams teach Watson, that learning becomes available to the entire company.
Watson’s machine learning, stressed IBM CEO Ginni Rometty, will remain proprietary for each business. “Your data goes to school,” she said, “learns what it needs to, then goes back home to your business.”
Laura Bennett, Development Manager for Watson Application Starter Kits, demonstrated the new Developers Starter Kits, a collection of REST APIs, SDKs, and samples that use cognitive computing to solve complex problems.The SDKs have been expanded and users can now code in variety of languages, including SQL, Python, R, Java, and Scala, Node.js, i/OS or Android. In addition, more than 20 ecosystem partner APIs are available to extend the platform services including Twilio, Box, IoT, Insights for Twitter, Weather Insights, Cloudant DB, and Facebook Messenger.
Starter kits are robust, including not only the expected documentation and samples, but models pre-trained for common use cases, notebooks, utilities, bundled APIs, and code templates available on GitHub.
For example, the News Intelligence Kit uses AlchemyLanguage, a collection of APIs that analyze text through natural language processing, AlchemyDataNews, a query that scours the world’s news sources and blogs like a database, and Tone Analyer, a tool that gauges the reaction of commenters. This Kit allows users to process Web and social media to not only understand important topics but how people feel about those topics. You can launch the app or check out the code on GitHub here.