Why Isn’t Machine Learning Living up to the Hype?
An obsession with tasks is leading to marginal returns on tech investments. Now is the time to rethink our approaches to machine learning.
When chief information officers think about their organizations and where machine learning might be deployed, the process often begins with an inventory of tasks.
The CIOs and department leaders identify routine, repeatable processes that humans can pass off to computers. Then the operations and IT teams set up targeted programs to make those tasks more efficient.
As legendary CIO Paul Strassmann has pointed out -- not without controversy -- it’s a piecemeal approach that has become standard practice in most businesses. It’s leading CIOs down a path of marginal returns and surprisingly limited innovation.
Strassmann’s career includes serving as NASA’s CIO from 2001 to 2003 and serving in an equivalent role in the Pentagon before that. As far back as 1998 he has been on record suggesting software should be seen as a storehouse of knowledge and experience in an enterprise -- what he calls “knowledge capital.” Software should not be the equivalent of a new forklift.
A new forklift does a job faster and better. But it does not learn or improve with every use. It doesn’t learn how it fits into the workflows of the business where it’s used, or how its work fits with the work of other machines. An even faster and better forklift is eventually bought, and the formerly new forklift is scrapped. All the use put into the scrapped forklift is lost, because obviously the machine never had the ability retain that knowledge capital. Strassman argues too many companies use enterprise technology this way, using it and then replacing it, rather than using it as a store for knowledge capital that becomes smarter and smarter.
That’s true for machine learning as well. It’s used as a tool to make tasks more efficient and faster, but it is not used enough as a store of knowledge capital not only for that task, but for how that task and others fit together, and can fit together better.
CIOs planning their organization’s evolution to machine learning, along with machine learning developers, need to dust off their Strassmann books.
More learning
CIOs should push to empower machines to do more learning, better, ahead of the task. This requires rethinking how machines take in data. Businesses should not think of themselves as a collection of tasks, but rather view their operations as brought to life by streams of data that run through workflows made up of those tasks. The tasks are just the muscles of the corporate body. Data is the blood flow and nervous system.
Focusing on how to turn that data into useful information and unique insights horizontally across the organization, no matter the task, is where CIOs can get a competitive edge and expand the return on machine learning investments. Deploy a smarter system for how data is ingested and interpreted by machines, and it will inevitably introduce greater efficiency and accuracy to the many tasks it touches. The goal is to move from a one to one benefit, to a one to many benefits.
Slow on the uptake
CIOs are having a tough time persuading skeptical business leaders to deploy machine-based intelligence in their organizations, and appropriately so. Enterprise tech marketers say the words “machine learning” very easily. But it’s harder to back those words with sustained, high quality results. Business leaders want more show, less tell.
A recent CFA Institute survey found that in the financial world, only 10% of investment professionals use machine learning. Instead they rely on traditional spreadsheets and desktop data tools. Across industries, only 50% of large businesses have artificial intelligence strategies. About 80% of enterprise businesses that have rolled out artificial intelligence or machine learning projects report stalled progress. And CIOs will continue to have a hard time modernizing their organizations and showing a return on the investment, if the effort remains task oriented.
As a team from Deloitte Australia writes, “if our social and economic systems persist in framing work in terms of tasks completed, and to value labor in terms of its ability to prosecute these tasks -- then we can expect AI & ML solutions to continue to be used as they often are today: as cost-cutting enablers, substitutes for humans instead of partners with humans.”
The question should be: How will the entire organization benefit from smarter data systems that pervade across workflows? And if humans are not spending their time collecting and sorting data, what else can they be doing to add value to the organization?
Kevin Walkup is President and COO of Harmonate, a data services firm serving private funds.
About the Author
You May Also Like