With public cloud providers competing to be the favored destination for their customers’ artificial intelligence projects, IBM Corp. has decided to bring its Watson technology out from behind the Big Blue firewall for the first time.

At its flagship Think 2019 conference in San Francisco today, IBM is announcing that Watson will now be available as a portable technology running on any cloud as well as inside customer data centers.

“Because data is so spread out, if you have to start every analytics project by moving data, it’s a waste,” said Rob Thomas (pictured), general manager of IBM Data and AI. “Our thesis is to bring AI to the data wherever it resides.”

Watson won’t run natively on platforms like the Amazon Web Services Inc. and Microsoft Corp. Azure clouds. Rather, customers will need to install IBM Cloud Private for Data container management platform on the target environment and Watson on top of that. As long as customers have a Watson license, they can move the instance to any cloud they prefer.

IBM Cloud Private for Data is part of IBM’s multicloud strategy that’s centered on making customers’ choice of cloud platform transparent. It’s a single environment for managing containers, which are virtual machines that incorporate all the elements an application needs to run.  Cloud Private for Data combines the Kubernetes container orchestration manager with a private image registry, management console and monitoring frameworks.

Battling complexity

While an estimated 80 percent of enterprises currently use or plan to use multiple clouds, the complexity of managing and protecting data tends to increase geometrically with the number of clouds they employ. Like many companies that failed to reach the pantheon of public cloud providers, IBM has adopted a strategy of making it easier for customers to shift workloads back and forth by leaving the data where it is rather than forcing large data transfers. The company continues to sell its own public cloud services.

With the new portability features, customers will be able to run Watson services, including the Watson Studio development environment, Watson Assistant virtual assistant and Watson OpenScale AI management framework, on any cloud as well as on-premises. Portability is enabled by a series of new microservices built for the IBM Cloud Private for Data that can move and scale across cloud environments, IBM said. With the Kubernetes foundation, the microservices can be run on any public, hybrid or multicloud environment.

Lock-in protection

IBM is positioning the move as protection against customer lock-in, despite the fact that customers must work within the Watson framework. As an early entrant in the market for commercial AI products, IBM invented much of its own technology for Watson, a strategy that has been criticized for limiting customer choice.

A year ago, SiliconANGLE sister research firm Wikibon, commented that large, integrated “initiatives like IBM Watson are finding themselves obsolete as their best AI tools are rapidly being duplicated in the open source community.”

But IBM’s Thomas said Watson supports a wide range of open source tools. “Eighty-five percent of the work that happens inside Watson is based on open source, such as Python, R and Tensorflow,” he said. “We’re not supporting all 10,000 open-source tools, but we support anything that has broad adoption.”

Watson began life as a question-answering engine focused on natural language comprehension and information retrieval, but IBM has been gradually expanding its charter into more mainstream machine learning applications and positioning it as a development platform rather than a black-box solution.

In line with that, the company is also today announcing enhancements to its business process management suite that folds intelligent recommendations into existing process automation. Thomas cited the example of an insurance underwriting process that can be made more efficient by incorporating machine learning algorithms that recommend improvements to workflow beyond simple automation.

“One of the biggest inhibitors to improvement is the claims companies don’t approve, often because of false negatives,” he said. “When you bring massive data to this process, you can be much more aggressive about on what you approve.”

Watson integration will be available later this year, IBM said without offering further specifics.

Image: SiliconANGLE