IBM Turns Up Heat Under Competition in Artificial Intelligence

By Robert McMillan 

Programmers of artificial intelligence software got a new tool to work with Monday, when International Business Machines Corp. announced a proprietary program known as System ML would be freely available to share and modify through the Apache Software Foundation.

The letters ML stand for " machine learning," a hot technology in Silicon Valley that enables computers to find common patterns in large amounts of data. Machine learning has been used to teach computers tasks such as predicting phrases entered into search engines, recognizing faces in photos and detecting unusual moves in stock prices.

IBM is the third company this year to make available proprietary machine-learning technology under an open-source license. Facebook Inc. in February released portions of its Torch software, while Alphabet Inc.'s Google division earlier this month open-sourced parts of its TensorFlow system.

IBM Vice President of Development Rob Thomas said System ML had gained preliminary acceptance by Apache, a respected open-source organization that manages more than 150 projects. Adoption by Apache was the first step, he hoped, toward System ML's widespread adoption.

"It's an endorsement of sorts that says, 'This has value,'" Mr. Thomas said.

System ML, which was first developed at IBM's Almaden research lab nearly a decade ago, could make it easier for developers to create customized machine-learning software, Mr. Thomas said. It could help a bank, for example, program risk modeling software that would provide early warnings of fraudulent activity. The current version is intended to work with software known as Spark, another Apache project that helps process large amounts of data as it arrives from continuous sources such as smartphones or fitness trackers.

"It's a pretty early-stage project, but it's pretty promising," said Dong-Bang "DB" Tsai, a senior research engineer at Netflix Inc. and Spark developer. He said System ML could make it easier for Netflix to improve its recommendation engine, which uses machine learning to determine which movies a given customer may want to see over, say, Thanksgiving weekend.

IBM decided to open System ML's source code through the Apache Foundation to attract a wider community of programmers and accelerate its development. Mr. Thomas said.

"It's about speed and innovation," he said. "Right now, my [research and development] is limited to my R&D budget, unless we're doing work in open source."

But IBM--like Facebook, Google and others--has another reason to open-source its machine learning code. The move helps it recruit new AI experts, and these days, AI talent is in great demand.

The release of System ML could give IBM an edge against companies including Apple Inc., which hasn't open-sourced its artificial intelligence software, said Brandon Ballinger, a former Google AI engineer who is now working with the University of California, San Francisco, on cardiology research.

Many of the world's top AI experts come from academia, and academics are active users of open-source software, Mr. Ballinger pointed out. "If you're a holdout, like Apple, you're simply not going to attract the best people."

Write to Robert McMillan at [e-mail]

Access Investor Kit for "NIRI Demo Co"

Visit http://www.companyspotlight.com/partner?cp_code=P479&isin=US4592001014

Subscribe to WSJ: http://online.wsj.com?mod=djnwires

(END) Dow Jones Newswires

November 24, 2015 06:14 ET (11:14 GMT)

Copyright (c) 2015 Dow Jones & Company, Inc.