IBM’s Watson Retreat Highlights AI Shortcomings
Amid reports of declining sales and growing skepticism about the utility of machine learning for complex medical research, IBM will reportedly end sales of its Watson AI software used by pharmaceutical firms for new drug discovery.
The health care news web site STAT reported last week that IBM (NYSE: IBM) is halting development and sales of its Watson AI drug discovery tools, citing disappointing sales. The company will instead shift the focus of its Watson Health offering to “clinical development,” the web site quoted a source said to be familiar with IBM’s strategy.
Health care industry experts note that AI tools such as Watson may be better suited to applications like diagnostic imaging, where they often outperform humans in terms of objectivity.
IBM stressed in a statement it is not discontinuing its Watson drug discovery offering and remains “committed to its continued success for our clients currently using the technology. We are focusing our resources within Watson Health to double down on the adjacent field of clinical development where we see an even greater market need for our data and AI capabilities.”
The company added in an email: “IBM Watson Health continuously looks to refine its portfolio to concentrate on the areas where we can have the greatest impact for our clients, and ultimately, patients. We remain very focused on helping the life sciences industry bring innovative medicines to market effectively and efficiently.”
In its latest quarterly financial statement, IBM executives sought to reposition Watson Health towards data analytics applications as healthcare clients “look to harness data to create actionable insights.”
Indeed, IBM has been signaling for months that it wants to integrate its Watson AI tools more closely with its expanding cloud platform in advance of what IBM CEO Ginni Rometty calls “the second chapter of cloud” adoption in which enterprises begin shifting core business applications to multi-cloud platforms. Foremost among them would be cloud-based analytics tools used for “harnessing that data, learning from it,” Rometty said.
Among those emerging AI tools is Watson Open Scale, designed to help data scientists track the quality of the machine learning models, including whether its accuracy is high enough or whether bias has been introduced into the model.
IBM’s retreat from costly but potentially lucrative healthcare segments like drug discovery did not come as a surprise to industry insiders.
“Reality is starting to catch up with the hype,” one healthcare executive observed. “After years of non-performance, a cynicism is growing in healthcare about the applicability of AI [and machine learning] to solve clinical diagnostic problems—which is perhaps the biggest problem in healthcare. IBM is mostly responsible for that cynicism, but has infected other firms with similar marketing angles.”
Which is not to say AI and machine learning don’t have a place in clinical settings, often outperforming human clinicians. “Particularly in the area of diagnostic imaging where the machines do a hell of a lot better job not bringing their prejudices, confirmation bias’, to the image,” the healthcare source noted.
Nevertheless, AI startups continue to look for ways to extract insights from unstructured data such as scientific research and clinical trials to advance drug discovery. “Developing new drugs is not a trivial task and in most cases, it consumes billions of dollars and takes years before it even becomes available to the market,” said Laurent Fanichet, marketing vice president at Sinequa, the cognitive search and analytics specialist.
“Fortunately, technology exists that is helping pharmaceutical researchers to expedite drug discovery, testing, and clinical trials in an information-driven way,” Fanichet added.
–Editor’s note: This story has been updated to include an IBM statement in response to reports about the status of its drug discovery offering.