An IDC survey of businesses out earlier this week found that for about 1 in 4 enterprises deploying AI today, half of all AI projects will fail. Why they fail can be related to external factors like a lack of talent necessary to properly deploy systems, but sometimes companies implode without the influence of outside forces.
In her experience with enterprise customers, Cloudera general manager of machine learning Hilary Mason said the raw ingredients necessary to succeed with AI go beyond data, straight to an organization’s ability to execute against a goal.
“If you take away one thing from what I’m saying here today, the dirty secret of applied machine learning, especially in the enterprise context, is that you set out to solve one problem, you try it, and then you solve a simpler problem that gets you to the same business outcome,” Mason said onstage at VentureBeat’s Transform conference Thursday. “And that creative ability to sort of realize what’s valuable and what the technology can handle, what you can do on a sustained basis — that’s what gets sabotaged by folks who may be managing or championing a project without really understanding what they’re actually doing.
“The other way I see this go wrong from an executive point of view frankly is just land grabs and people wanting to own something, therefore not doing what’s best for the entire organization just doing what’s best for their piece of it.”
In addition to acting as a general manager at Cloudera, Mason is founder and director of Fast Forward Labs, an applied machine learning research group created in 2014 and acquired by Cloudera in 2017.
Fast Forward Labs routinely writes about useful applications of advanced systems or approaches like meta learning or transfer learning. Mason and the lab also share predictions.
In a conversation with VentureBeat in late 2018 for 2019 AI predictions, Mason predicted more product managers would get involved in AI planning.
She objected to the presumption that data science is too abstract for product managers to understand.
“I think that’s wrong, and that the people who are best positioned in an organization should recognize where the opportunities lie are not the data scientists and the technologists themselves, but they are the people who actually own the product, who live with it every day, who work with the customers and clients — the people who are integrating all of that information to build a mental model of decision of where that product and where that business is going to go. Those are the people who are best positioned to recognize where AI should fit into it,” she said.
In her talk, Mason also urged businesses to embrace “boring” machine learning. That is, rather than chasing a trendy approach to AI or grand solution, businesses should look to solutions that might not be all that exciting to talk about but move the needle.
“When I say we need to make machine learning boring, we need to make it about results and outcomes and about using the simple-as-possible approach that will suffice to get you those results and outcomes. If it happens to be mathematically cool, boy will I be happy, but that’s really not the point,” she said.