Experimental feature

Listen to this article

00:00
00:00
Experimental feature

IBM is set to give a concerted push on Wednesday to the artificial intelligence system it is relying on to help revive its flagging growth and give it a new edge over rivals in the computing industry.

The US tech company plans to officially open a base for the system, known as Watson, in New York’s Silicon Alley neighbourhood and show off a range of examples in which the technology is already being applied in industries like finance, healthcare and retailing.

However, its slow start in commercialising the technology and the complexity of applying it in practical business situations has drawn scepticism from some AI experts.

Watson was first shown off winning the US TV quiz show Jeopardy nearly four years ago, demonstrating an ability to understand language that many people working in the field considered years ahead of previous systems.

The attention-grabbing stunt has also drawn scepticism, with some experts arguing that the system’s use of massive number-crunching to make statistical guesses amounted to a clever parlour trick rather than a true breakthrough technology.

“Watson won Jeopardy, but does it know it won,” said Oren Etzioni, head of an AI research effort funded by Microsoft co-founder Paul Allen. “Would you let Watson be your doctor?” The technology was “a stunning PR success”, he said, but: “If you take it any further, it falls flat on its face.”

IBM executives, however, have argued that Watson’s value lies in acting as a decision-support system for humans rather than a pure AI replacement, making it widely applicable in many general business applications.

The system is already generating “significant revenues”, said Stephen Gold, head of marketing for Watson, though IBM did not reveal sales numbers. Ginni Rometty, chief executive, has prodded staffers to accelerate commercialisation of the system as she looks for ways to rekindle IBM’s growth, making Watson a key test of her leadership.

The Watson technology is “a great toolkit” of software that needs a “team of experts to customise it” for different uses, making it slow to implement, said Tim Tuttle, chief executive of Expect Labs, an AI start-up in San Francisco. The system also needs to be “trained” on millions of documents for each field in which it is applied, he added.

IBM executives concede that making practical uses out of Watson’s head-turning Jeopardy win has been an arduous process, though they claim that a greater use of automation has been reducing the amount of human effort needed to apply it.

“There’s still work to do, certainly,” said Mr Gold. “The effort expended earlier on was greater than it is today, and a year from now it will be even less,” he said, with the time it takes for companies to write applications for the technology falling by then from months to “weeks or even days”.

Be alerted on International Business Machines Corp

Copyright The Financial Times Limited 2018. All rights reserved.