This article is part of an Essential Guide, our editor-selected collection of our best articles, videos and other content on this topic. Explore more in this guide:
2. - Data management market and trend analysis by editor Jack Vaughan : Read more in this section
- Data pros face tough data ethics questions
- Commercial cognitive computing not so easy for IBM Watson
- Various NoSQL databases seek to drive out RDBMSs
- Security services provider calms data flood with Hadoop, HBase
- Hadoop's role in modernizing mainframe apps -- or shutting them down
- The Google big data architecture: Not something for everyone to try
- Lucene search and Hadoop combine to surface 'unknown unknowns'
- Storm-on-YARN highlights Hadoop applications dichotomy
- Marketing and advertising driving big data cloud agenda
- Hadoop's evolution sparks new ways to program big data applications
- Bringing data to the masses: 4GLs yield to HTML5 for BI apps
- Big data age requires change in business decision-making process
- Out of the sandbox: Hadoop management tools become more important
Explore other sections in this guide:
- 1. - Business intelligence analysis from editors Craig Stedman and Ed Burns
- 3. - Oracle product and strategy analysis by editor Mark Fontecchio
The IBM Watson supercomputer has garnered a lot of attention in recent years, but it's entering a particularly critical passage now. What happens next could influence the future paths of data analytics generally, and IBM specifically -- for better or for worse.
This week, IBM showed its plan to move Watson forward. Virginia Rometty, the company's chairman, president and CEO, said IBM would invest more than $1 billion in a new business group dedicated to commercializing Watson. That figure includes $100 million for venture investments to create an ecosystem of application developers and other business partners. The challenge, though, will be to take highly technical machine learning software from the lab -- and the game show milieu -- to the business mainstream.
In 2011, Watson was carefully programmed as a specialized cognitive system with distinctive natural-language processing prowess. It was fed a large helping of information, and it went on to gain fame as something of a technology showcase for IBM. In effect, the company proved a machine could learn about human language and knowledge in amounts that were sufficient to beat vaunted Jeopardy! champions Ken Jennings and Brad Rutter in a two-match series of the TV quiz show.
Subsequently, IBM worked to expand Watson's domains of expertise to include healthcare, with a big emphasis on oncology and curing cancer, and, more recently, somewhat more mundane undertakings such as financial planning and customer service.
What happens next could influence the future paths of data analytics generally, and IBM specifically -- for better or for worse.
Watson's ability to learn as it interacts with humans and to predictively infer reasonable possibilities shows promise in analytical applications that challenge humans. IBM calls this cognitive computing, although the process also resembles expert systems work that it and others have long pursued. But replicating Watson's Jeopardy! success on the commercial stage hasn't been easy. For one thing, the questions that the system is called upon to answer are of a different ilk.
Game changer for Watson
"Jeopardy! has very clear parameters," said Adrian Bowles, principal at market research and analysis company Storm Insights Inc. He noted that the Jeopardy! Watson system was sequestered and could not search the Web or other data sources after the games began.
And issues of language aside, Jeopardy! facts can be pretty straightforward. In healthcare applications and other real-world scenarios, Watson will work much differently, Bowles said. In turn, the system will require the organizations that use it to work differently.
Bowles said that Watson provides users with measures of the confidence it has in its own answers. We've seen these confidence measures in weather reports, Gartner Inc.'s IT predictions and data analytics wunderkind Nate Silver's best-seller The Signal and the Noise, but working with confidence indexes is not first nature to many business leaders.
"Watson is a cognitive system that gives answers in context," Bowles said. "It's not just about building a database. It will change the way we think about applications. In a way, it is going to make people think about thinking."
Avoiding the AI graveyard
The IBM Watson technology is pretty heady stuff. And there's the rub. IBM will have to take care not to oversell it. That has happened to similar advanced technology before, going all the way back to the artificial intelligence (AI) systems that sprang up three decades ago. The AI market turned out to be something of a graveyard for overblown technology hopes.
"Making machines that beat humans at chess or a TV game show is much easier than solving problems in the messy real world," said Curt Monash, president of Monash Research and editor and publisher of DBMS2 and other blogs. "Watson doesn't seem to have yet overcome the problem that derailed 1980s AI technology, namely a reliance on small pieces of domain knowledge."
Watson capabilities, such as automated knowledge ingestion, could eventually get the job done, Monash conceded. "But it doesn't seem to be there yet," he said.
For more on IBM technology moves
Learn about DB2 updates known as 'BLU Acceleration'
Watch a brief video on the company's cloud analytics rollout
Read about IBM PureSystems expert integration
What's becoming clear is that IBM itself has to change its way of thinking to successfully bring Watson to market in a way that large numbers of customers will adopt. Signs suggest that Big Blue has gotten that message and is adapting its strategy.
Early Watsons were big honking machines -- supercomputers in name. But the company is quickly reducing Watson's footprint and deploying it as a cloud service. It's also finding ways to incorporate its data analytics technologies into Watson products, and vice versa. And IBM's efforts to build a Watson developer community shows a marked change from AI efforts in the days of yore.
Like Watson combing through the questions to the Jeopardy! answer "Nicholas II was the last ruling czar of this royal family" -- or the answer to the question, "Does this kid have tonsillitis?" -- IBM has to find the right answers to the challenge of commercializing cognitive computing. Will what it's doing be enough? If only IBM Watson could answer that one.