Cognitive Computing: The Myth and the Reality
When IBM Watson burst on the scene a few years ago by famously winning Jeopardy! people swooned. Here was a machine that seemed to live up to the promise of a science fiction future. IBM purported to show us all an artificial intelligence (AI) that could understand human language, sift through massive amounts of data, and provide answers to questions.
From a marketing standpoint this is a gold standard for enterprise software. Companies quickly signed on to the promise, looking to Watson for answers for everything from insights in their CRM data to finding a cure for disease.
The reality, however, is much different.
We regularly hear from clients struggling with failed Watson implementations that the IBM system is truly a front end for IBM Business Services. This means consultants will descend on the project and try to use a series of APIs to connect into different technologies. The result is a hodgepodge of different systems, none of which truly do the job. In some cases, six-months or a year can go by with massive amounts of time and money invested, but no real results to show.
This is why when speaking with CNBC, Social Capital CEO and founder Chamath Palihapitiya called Watson “a joke” and attributed its success more to the company’s sales and marketing teams than to its technology.
This isn’t only the case in the enterprise, but across all of Watson’s product lines. It’s why STAT recently reported that the Watson Oncology product is of little use to most doctors.
IBM offered grand promises for its Oncology product, saying the AI will help doctors find new insights and new treatments for cancer. But the reality is that it’s only as good as the humans behind it, and the AI technology hasn’t supported the delivery of new treatments. Some of those using the system who were interviewed by STAT pointed out that the expensive AI system usually just tells them what they already know.
The Problem with Big Promises
The truth is, IBM Watson does have some solid technology behind it, but the technology is like a 1000-piece jigsaw puzzle: it’s a menu of APIs. Building the final application is resource-intensive, and few ever get to see the value.
This isn’t a small problem for the industry. Because of IBM’s size and position, along with the huge marketing budget behind Watson, people have begun associating cognitive computing with the Watson offering. When companies become frustrated with the results the danger is that they will associate that frustration with the underlying technology. The truth is, cognitive search and cognitive computing have great power and capabilities when executed properly.
The Solution: Start with Smaller Successes and ROI
Forrester Research points out that it’s cognitive search that underlies AI applications by augmenting those applications with relevant information from multiple sources. The true value lies in connecting the information across multiple data sources that contain structured and unstructured information. When cognitive search can find links within that data, it can give users the insights they need to make informed decisions.
The problem isn’t with cognitive computing: it’s with the promises that IBM is making to its customers. The answer lies in starting with smaller successes and then growing those within the enterprise. That allows each project to stay focused on returning value, building trust, learning, and expanding on the successes.
Only then can cognitive search live up to its promise: giving intelligence to the people who can make the most of it.