IBM officials recently told a Bloomberg journalist that the company is preparing a version of its Watson natural-language question-answering software for access via smartphones and tablets: Watson 2.0. A previous Explorer article discussed Watson's public debut as a contestant competing against human players on the quiz show Jeopardy. Another previous Explorer article described efforts to teach Watson to function as a physician's assistant in clinical settings but suggested that IBM has reasons—the long development road map and the abundant existing artificial-intelligence (AI) tools for clinical medicine—to seek opportunities for Watson that are completely unrelated to medicine.

Watson's 2011 appearance on Jeopardy implies roles for Watson in casual queries about any topic. But the Bloomberg article indicates that IBM is now focused on developing applications of Watson 2.0 for professional users—for example, farmers who ask advice about when to plant. Such professional users can expect Watson 2.0 to take advantage of image recognition via smartphone and tablet cameras and of the location, position, and motion data reported by sensors and GPS receivers in portable-devices.

Bloomberg interviews indicate that IBM expects that eventually a patient will be able to describe personal medical symptoms and receive recommendations from Watson, with no physician or other professional intermediary required. But IBM continues to portray personal applications of Watson as longer term and not currently under development.

IBM's Bernie Meyerson told Bloomberg that "the power it takes to make Watson work is dropping down like a stone," a reference to ongoing rapid progress in semiconductor manufacturing. Because any voice phone—not just smartphones and cellular phones—can access intelligent speech services, the cost of operating Watson is governed by the outlook for server-operating costs rather than by the outlook for handset capabilities.

Potential Business Models for Watson 2.0

Many stakeholders expect that a commercialized Watson will be more capable than Apple's Siri. If so, Watson 2.0 will likely require IBM to expend more data-center resources per Watson query than Apple devotes to a Siri query. Hence, commercializing Watson depends on IBM's reducing the cost of running Watson at the data center and crafting a business model that collects more revenue than it spends on data-center operations. Here are some attributes of potential business models for Watson 2.0:

• IBM may seek advertising and affiliate-merchandising revenue, a business model similar to Apple's initial business model for Siri. However, professional users will likely prefer Watson's responses to be unbiased, uninfluenced by an advertiser, and unaccompanied by annoying paid messages.
• A professional focus for Watson 2.0 may entail subscription access comparable to access to premium news services such as those of the Nikkei and the Financial Times.
• Another possibility is a wholesale IBM service delivered to brands that sell handsets, mobile services, or retail apps or all of the above. Similarly, Siri makes use of Wolfram Alpha's wholesale service.
• Different versions of Watson 2.0 may focus on different professional domains; for example, one version may only answer questions about financial services (Citibank is training Watson for this purpose). Potential exists for the creation of other domain-specific personal-assistant agents, such as for legal research, building codes, information-technology (IT) implementations, and so on.
• To compete effectively against free-to-user natural-language applications such as Siri, Voice Search (Google's app for various smartphones), and Wolfram Alpha, Watson 2.0 may need IBM to implement a "freemium" business model in which some service is free and premium service carries a charge.

Concluding Perspectives

Recent communications about Watson 2.0 may reflect a strategy at IBM that is still in flux. Watson's future competitors are already available and will likely improve. The fact that Watson's public debut was on a game show suggests that question-answering software's destiny is to make a broad contribution to everyday life, not just to play narrow roles in specific professional domains. IBM or another organization can in theory develop a general-purpose personal-assistant agent with intermediate costs and capabilities—that is, one that can deliver a query response less expensively than today's Watson can but is more capable than Apple's Siri or Google's Voice Search.