As machine learning becomes more accessible through avenues such as Intel Corp.’s BigDL and IBM Corp. opening Watson’s core machine learning components up to businesses, some developers and industry insiders are cautioning against getting too dazzled by the potential without considering the human role.

However much data those programs can process, in the end, “what you do with the results of algorithms is key,” said Jean-Francois Puget, Ph.D. (pictured), distinguished engineer, machine learning and optimization, IBM Analytics, at IBM.

Puget spoke with Dave Vellante (@dvellante) and Stu Miniman (@stu), co-hosts of theCUBE, SiliconANGLE Media’s mobile live streaming studio, at the IBM Machine Learning Launch Event in New York, NY. He offered his perspective on machine learning and its applications. (*Disclosure below.)

“For most people, machine learning equals machine learning algorithms,” Puget said. “When you look at newspapers or blogs, social media, it’s all about algorithms. Our view [is] that sure, you need algorithms for machine learning, but you need steps before you run algorithms, and after.”

Puget explained that the “before steps” include getting the data and transforming it to make the data usable for machine learning. “And then you run algorithms — these produce models — and then you need to move your models into a production environment,” he said.

With the results from these processes, practical uses, such as detection of credit card fraud and similar pattern of behavior, can be examined and used to adapt policies.

Models, machines and mankind

Puget was also careful not to be overwhelmed by the possibilities of such applications, saying that machine learning is “overhyped, maybe, but it’s also moving very quickly,” he acknowledged. “Five years ago, nobody spoke about deep learning. Now it’s everywhere. Who knows what will happen next year? So our take is to support open-source, to support open-source packages. We don’t know which one will win in the future, we don’t know even if one will be enough for all needs. We believe one size does not fit all.”

As Puget noted, machine learning was incredibly helpful in forming the models of behavior for algorithms to process, “but a model does not tell you what to do.” The probabilities shared by such analysis were only a means to more informed decisions, in Puget’s perspective, and not a source that should be consumed without additional external input.

“The idea is to use machine learning predictions as yet another input for making decisions,” he said.

And while he didn’t feel that the technology to replace human decision-making was here, or that it would exist anytime soon, Puget did feel that “we can certainly make some professions more efficient, more productive, with machine learning.”

Watch the complete video interview below, and be sure to check out more of SiliconANGLE and theCUBE’s coverage of the IBM Machine Learning Launch Event 2017 NYC. (*Disclosure: TheCUBE is a media partner at the conference. Neither IBM nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo by SiliconANGLE