Watson peaked early. IBM’s superstar artificial intelligence (AI) dazzled audiences and trounced its human competition on Jeopardy in 2011. IBM (NYSE: IBM) promised a bright future for Watson, transforming lives and work, particularly in medicine.
In retrospect, Jeopardy was a high point for Watson’s career. The technology never fulfilled its promise, and IBM struggled through the decade, a mainframe business in a world moving to the cloud. Recently, new AI superstars eclipsed Watson’s fading glory: ChatGPT, a revitalized Microsoft Bing and Google Bard.
But Watson is back. Last month, IBM launched its new AI strategy with a cool new monicker— watsonx —and a plan to integrate AI with the company’s strengths, delivering multi-cloud enterprise solutions and services tailored to big business needs. Other companies shine in offering consumer services, but IBM is built for the enterprise, providing unmatched understanding of customers’ business needs and integration with existing technology, and governance and security, the company says.
Tarun Chopra, IBM VP of Product Management, Data and AI, now oversees IBM’s AI initiatives after years heading up the company’s mainframe business. We spoke with Chopra about IBM’s AI future and its checkered past.
“We have learned a lot over the last 10-plus years about what it takes to scale AI in enterprise clients’ environments,” Chopra said.
A tarnished crown jewel
Watson Health was supposed to be the crown jewel of IBM’s Watson strategy, but instead, it was a spectacular failure. The company invested $5 billion and 7,000 workers in the platform designed to improve cancer treatment, but got negative reviews from MD Anderson Cancer Center and Sloan Kettering. IBM sold Watson Health for $1 billion to a private equity firm last year.
Nonetheless, IBM says Watson has had many successes in customer service, supply chain, financial planning, risk and compliance, advertising, IT, video and security at scale. The company claims 70% of global banks and 13 of the top 14 systems integrators used Watson.
During the Covid pandemic, the pharmacy chain CVS used Watson for vaccine scheduling. Call center operators couldn’t come into the office and call volume ballooned, so CVS deployed Watson to automate processes and get vaccine appointments done, Chopra said.
IBM’s business background makes it well-suited to match AI to enterprise needs for skills, cost, regulatory compliance, and working with existing investments and infrastructure.
Watsonx uses new technology to streamline AI processes. With previous AI technology, you had to build a new model for every problem, Chopra explained. New AI tools, such as Microsoft-backed OpenAI’s ChatGPT and Google Bard, use “foundation models” to build models that can be reused and customized to speed up time to value.
For instance, ChatGPT uses a Large Language Model (LLM), which, as the name suggests, is optimized for language, Chopra said. An LLM is expensive to create, but once done, delivers big benefits.
A foundation trilogy
Foundational models were a big part of IBM’s major AI announcement last month. IBM introduced watsonx.ai, set to be available next month, with foundation models including fm.code, for generating code through natural language, to improve developer productivity and enhance IT automation; fm.NLP, a set of LLMs for specific industries, which can be quickly customized for client data; and fm.geospatial, built on climate data and sensing to help companies plan for natural disasters, sustainability and other geophysical processes that could effect business.
Consumer conversational AI’s shortcomings are well known, including the instance of a lawyer who used ChatGPT to prepare a court filing, only to face penalties when it turned out that the AI simply made up facts and citations. Samsung banned its staff from using conversational AI after an accidental data leak. And Google recently warned its staff about using conversational AI, despite Google’s being one of conversational AI’s biggest champions, offering one of the leading AIs, Bard.
Unlike ChatGPT and Google Bard, IBM’s foundational models are designed for the enterprise and regulatory ready.
But there’s more to IBM’s AI strategy. The company is partnering with Hugging Face to use the latter's open-source libraries, models and data sets.
Watsonx.data is a data store built on the Lakehouse open data management architecture – combining the best qualities of data lakes and data warehousing. It will be delivered next month and can be managed on-premises and in the cloud.
Additionally, IBM will provide access to GPUs for AI infrastructure, and, importantly, launch the IBM Consulting Center of Excellence for Generative AI, with 1,000 consultants.
Multi-cloud advantage
Holger Mueller, VP and principal analyst at Constellation Research, is impressed. “IBM is definitely a player in AI,” he said.
The company offers freedom from lock-in to a single cloud platform, with Red Hat as a multi-cloud enabler, Mueller said.
And IBM’s consulting arm gives the company a powerful advantage. “Enterprises need the skills,” Mueller said. “They need to look outside for help. There are just not enough data scientists and developers in the world. And enterprises need strategic advisors.”
Patrick Moorhead, founder, CEO and chief analyst at Moor Insights & Strategy, also sees IBM as a strong AI contender. He said IBM has been partnering aggressively with public cloud providers since Arvind Krishna took over as CEO in 2020, including partnering with Amazon Web Services (AWS) for software distribution. (AWS no longer markets using the term “mainframe replacement,” Moorhead noted.)
IBM’s AI will be particularly attractive for highly regulated industries, such as finance, healthcare, transportation and sensitive government agencies such as the CIA, Moorhead added.
And IBM’s reputation is another major advantage for the company; it’s been doing business with banks, insurance companies, government agencies and healthcare for more than 60 years. These companies trust IBM. And trust is something that can't be bought.