3 machine learning myths and how to overcome them

To succeed with machine learning, you’ll have to get past these common misperceptions about the technology

istock 1151224850
istock

Artificial intelligence (AI) and machine learning (ML) have established themselves among early adopters with clear benefits that range from improved operational efficiencies to better customer experiences. But misconceptions remain about these emerging technologies, which can prevent other organizations from exploring the advantages that AI and ML offer across different businesses and industries.

Here are three of the more persistent myths around machine learning, and how senior business leaders can overcome them to help drive development and deployment.

Myth #1: Only large enterprises can afford to invest in machine learning and AI.

When business leaders consider the concepts of machine learning and advanced AI, many assume they are accessible only to large global players with lots of resources. Not true, says Sri Elaprolu, senior leader, Amazon Machine Learning Solutions Lab, a team of data scientists and machine learning experts that helps Amazon Web Services (AWS) customers successfully adopt ML.

“I work with customers of all types and sizes including midsize enterprises, startups, public sector organizations, as well as not-for-profit entities,” Elaprolu says. “The reality is that machine learning can be absolutely beneficial for organizations of every size and every scope.”

For example, a Seattle-based startup, Convoy, has leveraged ML to help disrupt the way trucking brokers were connecting shippers and haulers. Using ML models to automate processes that relied on traditional methods such as email, address books, and phone calls, Convoy now provides better matches for shippers and truckers that let them move freight more efficiently, while reducing costs for both parties.

Companies of all sizes can take the first step on their ML journey by realizing that the technology is not just available for those with deep pockets.

Myth #2: We have to do everything ourselves to build and deploy ML across our business.

In the early days of ML and AI, it was true that companies needed to develop and hire a large data science team to develop the technologies. But in recent years, thanks to innovations in cloud computing and processing power, many third-party companies have developed platforms and tools that can help companies in their ML journey.

“Right now there are a number of services, capabilities, platforms, and tools available,” says Elaprolu. “Companies are able to move much faster by leveraging ML capabilities and not have to go back and reinvent the wheel or hire a large data science team to get started.”

For example, AWS AI Services enables developers to add capabilities like image and video analysis, natural language, personalized recommendations, virtual assistants, and forecasting to applications without deep expertise in machine learning. For ML developers and data scientists, Amazon SageMaker  provides a platform to build, train, and deploy ML models, removing the complexity that can sometimes derail a successful implementation. AWS also has learning tools such as AWS DeepRacer, AWS DeepLens, and AWS DeepComposer that help organizations improve their ML knowledge. And finally, AWS provides programs such as the ML Solutions Lab and AWS Machine Learning Embark to provide hands-on onboarding, training, and implementation support.

Myth #3: Data scientists will solve all of our problems.

Data science is a critical skill set in modern organizations. But successful ML deployments extend well beyond a small team of dedicated data scientists.

“Just bringing data science to a problem doesn’t automatically solve it,” says Elaprolu. “You need to think about which use cases are addressable by machine learning versus thinking machine learning is a secret weapon that you can use to solve all things.”

Busting this myth requires involvement from key business stakeholders to work with the data teams to prioritize the business problems that ML models can help to solve.

“Every part of the organization should be thinking about machine learning,” says Elaprolu.

Learn more about machine learning and AI services for your business. 

For more machine learning insights from Sri Elaprolu, check out the new Ahead of the Pack podcast.

 

Copyright © 2020 IDG Communications, Inc.