Video: Portrait of a modern multi-cloud data center
IBM Chairman Ginni Rometty issued the clarion call to an arena packed with roughly 20,000 IBM customers during her Think 2018 keynote. Incumbent companies can disrupt their own industries and you don't have to be an Uber to pull it off.
Compared to her keynote at the former World of Watson conference almost 18 months ago, Rometty on this go-round was far more specific, showing concrete examples of how legacy companies are adopting next-generation technologies that changes their businesses.
Among the examples were Maersk, the worldwide shipping giant with a blockchain pilot that will streamline the processing of shipping documents, much as containers streamlined global trade 50 years ago. Or Royal Bank of Canada, which is utilizing artificial intelligence (AI) to deliver new personal financial assistants designed to make its customer relationships more intimate.
Also: IBM's cloud strategy revolves around multi-cloud support, grabbing new workloads |Top cloud providers 2018: How AWS, Microsoft, Google Cloud Platform, IBM Cloud, Oracle, Alibaba stack up IBM turns Watson into a voice assistant for enterprises | TechRepublic: IBM's reactive platform makes it easier for Java developers to build AI apps
Since its coming out party on Jeopardy back in 2011, Watson has been the epicenter of IBM's reboot. The company is making a riverboat gamble that AI will be its -- and its customers' -- futures. AI has morphed into an entire market ecosystem that, as we pointed out earlier this week, includes lots of the brand names, applications, and tools that most enterprises are already using. The case against IBM was that its cognitive computing approach required too long an onramp for delivering results.
Going forward, IBM has changed its branding, and Watson is no longer singularly synonymous with the cognitive flavor of AI. While Cognitive is still very much in IBM's wheelhouse, Watson now stands for any IBM product or service that embeds AI. In so doing, IBM has joined the crowd with its own flavor of making the fruits of AI more accessible.
Among the headlines are IBM Cloud Private for Data and Watson Studio, both of which were unveiled over the past week. If you speak to an IBMer, they get stuck in the non-intuitive ICP jargon for a product the rest of the world would more likely term "Cloud Private." Cloud Private for Data is actually the first purpose-built product on the private cloud platform that was first released last fall.
We have not yet had a chance to look at the innards of Cloud Private for Data. But it is clear that it adheres to cloud-native design principles by deconstructing IBM monolithic data integration products into containers and microservices for running database, federated query, data cataloging, data lifecycle management, predictive modeling, dashboards and reporting, and other services.
Complementing Cloud Private for Data is the newly released Watson Studio, which is a 2.0 rethink of Watson Data Platform and Data Science Experience. It provides a more visually integrated user experience; bundles in formerly a la carte capabilities like Data Refinery (for data preparation) and cataloging; and automates key steps of preparing models.
While it also provides a code-centric alternative and accommodates Jupyter notebooks (which are autopopulated and navigated to at any point of the process), the highlights are the visual flow and automation. The deep learning capabilities are the highlight: Watson Studio can automatically generate complex, multi-layered neural networks and automatically run hyperparameter optimization for tuning models that would otherwise require considerable manual footwork. To that, IBM has also added support for Python programmers who prefer the Anaconda packages to those of Spark (Spark remains the compute engine).
At first glance, the visual approach makes Watson Studio look like IBM's answer to entry-level cloud AI services like Amazon SageMaker and Microsoft Azure ML Studio. But Watson Studio does not restrict you to a curated library of popular algorithms (although like SageMaker, it offers some curated pretrained models, in this case built from Watson). Despite its visual front end with prebuilt automation, Watson Studio is still very much aimed at hard-core data scientists. It is a good step forward in rethinking the data science workflow. As a 1.0 refactoring of previous offerings, there are still plenty of functional gaps to fill. For instance, the visual charting that illustrates model scoring over time only displays accuracy; we presume that adding other parameters like learning rate will follow in an upcoming release).
The challenge for IBM, like any company with a long legacy and large base of customers still running that legacy, is how to onramp them so they can become, in Rometty's terms, disrupters.
We bookended a couple sessions back to back that looked at both sides of the information governance issue. The brunt of IBM's data governance and integration tools stem from acquisitions made 10 - 15 years ago, so it's not surprising that these tools are showing their age. First, we attended one of a series of "user experience" roundtables where a facilitator led existing customers through a discussion of pain points and wish lists. For the InfoSphere Information Server and related tools, customers voiced concern that data lineage and search capabilities were not easily usable. A couple hours later, we caught a session focusing on how IBM is incorporating AI and ML in their next-generation data governance tools, many of them deployed as cloud services, that would address issues like those.
The dilemma IBM faces is how to upgrade these customers, many of whom either lack the budgets to undergo forklift upgrades to the new generation of tools or have other priorities. It would appear that for those on the legacy InfoSphere tools, applying some of this machine learning to the existing portfolio for tracking lineage or searching metadata to make those functions more intuitive would be logical add-ons -- as long as the products are actively supported, of course.
IBM Watson Studio is closely integrated to IBM Cloud Private for Data, but for now they don't run on the same clouds.
In some cases, IBM gets in its own way. Its messaging and positioning was muddled and confusing. With the dearth of press releases, there was no authoritative go-to source for the brunt of IBM's announcements or upgrades, making the process of separating fake news from fact a matter of finding the right expert in a crowd of 20,000.
And then there was ambiguity. The slide, shown above, was prominently displayed at an expo hall keynote about IBM's latest AI offerings. It accurately shows that the new offerings, Cloud Private for Data and Watson Studio, are in fact synergistic. The speaker, IBM general manager of Watson Data & AI Beth Smith, stated that you could work seamlessly across the private and public cloud, but the slide implies that Watson Studio works with Cloud Private for Data in the private cloud. It doesn't tell you that Watson Studio is not available on Cloud Private for Data -- at least not yet (it is only available on the IBM Public Cloud). There may be good reasons for this, such as that the clock ran out before IBM could get an implementation of Watson Studio ready for Cloud Private.
Cloud Private for Data and Watson Studio are good examples of how Design Thinking was applied to produce intuitive workflows. The breakdown in communication points to the need for IBM to extend design thinking practices into the realms of messaging, positioning, and communications with all of its varied constituencies.