Icon (Close Menu)

Logout

Machine Learning Tops List of Hot, Trendy TechnologiesLock Icon

5 min read

Mummified cats aren’t a natural accompaniment to voice-operated doors.

The leaders of technology company RevUnit made that discovery after moving its offices into the old Farmers Co-op in Bentonville, which retained some of rustic charms. “There were old mummified cats under the floorboards,” Michael Paladino, RevUnit’s chief technology officer and co-founder, said in giving a reporter a quick tour. “It smelled like an old feed store.”

The place has been cleaned up and outfitted with a voice-operated door that impress clients or accommodate people whose hands are too full to grasp the door handle.

The high-tech doors are just a hint at what goes on inside RevUnit, which builds applications to help businesses operate efficiently in areas like productivity, worker engagement and learning.

Paladino has focused on some of today’s trendiest technologies: artificial intelligence and machine learning.

Paladino said that at RevUnit, he generally classifies artificial intelligence with machine learning because of the interrelationship of the two technologies. There are a wealth of applications for which ML/AI serves as the underlying science.

“Machine learning and artificial intelligence enable a lot of different functionality that helps that enterprise,” said Paladino, referring to a business or company. “Those technologies can be used on the front end for data consumption with things like image recognition. They can be used on the back end for understanding the data that you have within your enterprise. All of those things can be used to improve on that employee productivity and engagement.”

Machine learning can work on any business level. RevUnit devises ML/AI-based applications for clients, and massive companies such as Walmart Inc. have invested heavily in machine learning and artificial intelligence.

Big and Small
One of the projects RevUnit is working on — Paladino said he can’t divulge the client — involves combining the technology of computer vision with ML/AI. A client told Paladino it wanted to find out how long its customers were waiting in line to make a purchase and wanted to accomplish that without added manpower.

Paladino and Colin Shaw, RevUnit’s senior machine learning engineer, said achieving that goal was a matter of getting images of people in line, training a computer to count the people, and then combining that learned behavior with the data produced by sales. A trained computer can do the recognition and provide the information without human intervention.

“They spend a lot of time not doing a lot of value-added work in their business by looking at the images,” Shaw said. “We can correlate this with information we can get, the rate of transactions from a register, to see how fast the people are going through the line and how many people are waiting in the line.”

Laurent Desegur, vice president of customer experience engineering at WalmartLabs, spoke at the MobileBeat 2017 technology conference in San Francisco last summer and said machine learning was a critical component of some of the technologies Walmart is testing. Desegur said Walmart is using machine learning to build a model of personalization and recommendation for its customers and bridge the experience between in-store shopping and online shopping.

“Everything we want to do is make sure there is a seamless experience,” Desegur said.

He told the audience that machine learning can help the company weed out abuse of product or service reviews to exclude spam or to make sure comments come from real people. Machine learning is also the technology behind many of the in-store trials Walmart is using, including self-service kiosk towers, drive-up delivery and mobile paying to avoid check-out bottlenecks.

“We want to create a notion of the store of the future,” Desegur said.

Worth the Wait
In another presentation he gave at the France Is AI technology conference, Desegur said artificial intelligence and especially machine learning are only as good as the quality of data provided.

At Walmart, its 140 million weekly customers generate an immense amount of data that the company uses to devise solutions.

Shaw said ML/AI really began to come to the technological forefront five years ago because of a confluence of factors. Computing power improved to allow analysis of the tremendous amount of data being produced, and faster computing led to an increase in academic research on machine learning and artificial intelligence.

“It’s really kind of a new thing, in a way, to see all the progress that is happening,” Shaw said. “Those items all had to converge together. You needed to have a lot of quality data and you needed to have computing resources and you needed to have interest from people creating the tools. For 20-30 years beforehand, you didn’t have those.

“At this point a lot of why this is turning into something noteworthy is because in the last five years, we’ve seen a lot of successes where computers are able to solve problems as well or better than humans can. That wasn’t something that was happening before.”

Shaw said machine learning is all about input. A computer never really becomes intelligent about how many people are in a line; it just learns how to count if showed enough images of people in line.

Paladino said ML/AI has greatly benefited from libraries created by major technology companies such as Google or IBM, which have put the research and data within the reach of people like Shaw and Paladino. RevUnit, which doesn’t have the depth of data or research funds of Google, takes the available data and computer models and then rewrites algorithms for its specific projects.

“One of the reasons that is useful is not because we can’t do that in-house, but one of the hard parts of that problem is being able to train the model,” Shaw said. “That requires vast amounts of data and a lot of computing resources. They’ve already done a lot of the difficult part of model training for us.”

TeachableMachine.WithGoogle.com allows visitors to see how a computer, when given enough data, can then learn to associate that data with a result. The simplistic example at that website is the same concept as ML/AI projects, though the machine learning projects are much more complex.

“Some of that technology has been around for a while,” said Paladino, using email spam filters as an example. “These libraries are allowing us to get access to that functionality much more quickly. We are taking advantage of all the research that has been done. The reason you are seeing so much more of it now is because it is easier to do now than it used to be.”

Send this to a friend