Will Intelligent Machines Replace Or Complement Human Workers?

February 29, 2016 Jeff Kelly

sfeatured-machine-handA recent article in Canadian Business tells the story of a company called Landr and how it uses machine learning to master music tracks without the aid of human engineers. As the article explains:

“Landr deploys big data analytics and machine learning to analyze a track and determine how to adjust parameters such as compression and equalization to make the song sound as pleasing and dynamic as possible to the ear.”

The company has been at it since mid-2014 and has mastered over 1.5 million songs. It’s a great equalizer for smaller bands trying to break through that don’t have the cash to hire professional recording engineers.

While Landr presents another good example of a real-world machine learning application, it also raises a touchy subject that has professionals across industries as well as policy makers concerned—will machine learning and artificial intelligence ultimately serve to complement human workers or replace them?

The article quotes Steve Albini, a recording engineer who’s worked with Nirvana and the Pixies (two of my favorite bands, by the way). He is adamant that technology can never replicate the work of experienced engineers. “There is literally no way it can be automated any more than you could automate getting your haircut,” Albini wrote in an email to Canadian Business.

Whether intelligent machines will replace or complement human workers is a question Erik Brynjolfsson and Andrew McAfee tackle in their book, The Second Machine Age. In it, they take the totally reasonable position that intelligent machines will replace humans for some tasks; but in most cases, intelligence machines will complement human workers, making them more productive in the process.

As to which tasks fall into which category (complement or replace), Brynjolfsson and McAfee contend that the biggest determinant is not whether a task is complex (like mastering a song) or simple (like cutting someone’s hair). Rather, the likelihood of intelligent machines replacing or complementing humans for a given task is much more dependent on whether that task is routine or not. In the short- and mid-term at least, intelligent machines are much more likely to complement human workers in non-routine tasks, whether complex or simple, than replace them. Write Brynjolfsson and McAfee:

“Routine clerical work like processing payments is easier to automate than handling customers questions. At present, machines are not very good at walking up stairs, picking up a paper clip from the floor, or reading the emotional cues of a frustrated customer.” —The Second Machine Age, pg. 138

The last example is telling. It is true that machines aren’t particularly adept at reading emotional cues, at least not yet. There is speech analytics software that aims to discern emotion in spoken language, but its accuracy needs improvement before it can replicate the capabilities of a human reading emotional cues in real-time. Today, in early 2016, if it was my company I’d rather have an experienced customer service rep on the phone dealing with an unhappy customer than a machine, like Apple’s Siri. (Again, interactive voice recognition software exists and is in widespread use, but anyone who’s called a customer support line and dealt with an IVR knows how frustratingly inept they often are.)

But that doesn’t mean machine learning doesn’t have a role to play in this call center scenario. Even better than just a customer service rep trying to talk the customer down would be a customer service rep aided by an intelligent application that suggests special offers or discounts for the angry customer based on that customer’s past buying behavior, likelihood to churn, total lifetime value and other relevant data points. These types of data-driven applications have been around for some time, but with the introduction of true machine-learning supporting them, the potential for improved accuracy is enormous.

But the Landr example adds an interesting twist to the replace vs. complement debate. There are some tasks, as Brynjolfsson and McAfee explain, that intelligent machines can’t do as well as human workers. However, they can do the tasks well enough at a lower price, satisfying the majority of consumers. I’m using the word consumers broadly to mean any persons or applications that consumes the output of a given task. In these cases, there is still a market for the higher quality work of humans, but the majority of the market is served by good enough work produced by intelligent machines at a much lower price.

This sounds like an apt description of what is happening to the music business. The Pixies of the world will likely continue to pay for the talents of all-star recording engineers like Steve Albini, while bands with limited cash will more likely opt for services from companies like Landr. Landr Chief Creative Officer Justin Evans concedes as much: “I don’t think I’m going to ever argue we’re better than the best mastering engineer in the world, but I do know that we are perceptively close.”

This is good news, of course, for Steve Albini and his fellow all-star recording engineers, who won’t have any trouble finding work. It is also good news for struggling garage bands, who previously were priced out of the market for recording engineers but who now have a lower cost alternative. It’s also good news for Landr, which has tapped into a real need in the market. But, it’s not so good news for all the other recording engineers out there who are now competing with intelligent machines.

What do you think? Is machine learning an either/or debate, or will we see a spectrum of impacts on a use-case by use-case basis? Leave a comment or tweet me at @jeffreyfkelly and let us know what you think.

About the Author

Jeff Kelly

Jeff Kelly is a Principal Product Marketing Manager at Pivotal Software. He spends his time learning and writing about how leading enterprises are tapping the cloud, data and modern application development to transform how the world builds software. Prior to joining Pivotal, Jeff was the lead industry analyst covering Big Data analytics at Wikibon, an open source research and advisory firm. Before that, Jeff covered data warehousing, business analytics and other IT topics as a reporter and editor at TechTarget. He received his B.A. in American studies from Providence College and his M.A. in journalism from Northeastern University.

Follow on Google Plus Follow on Twitter More Content by Jeff Kelly
Previous
How To Scalably Extract Insight From Large Natural Language Documents
How To Scalably Extract Insight From Large Natural Language Documents

In this post, one of Pivotal’s data scientists, Scott Hajek, explains how Greenplum Database (either the op...

Next
Build Newsletter: Connected Car Dominates #MWC16
Build Newsletter: Connected Car Dominates #MWC16

In this week’s BUILD Newsletter, we provide highlights from Mobile World Congress. The conference has never...

×

Subscribe to our Newsletter

Thank you!
Error - something went wrong!