Machine Learning Goes Mainstream
What if a small business could suddenly unleash powerful search tools on its previously pedestrian website so that browsing shoppers become paying customers at the blink of an eye? Or how about being able to harness highly intuitive algorithms that stop cybercriminals snooping around a company’s network?
These kinds of previously out-of-reach services are now accessible through advanced machine learning models that have gone mainstream.
Leaders in cloud-computing have invested heavily in developing machine learning models for their own business needs, but in the process have created new services that have wide-ranging applications for companies of all sizes. Amazon, Google, and Microsoft are competing to provide cloud-based machine learning services to organizations that need robust search capability, image recognition, video analysis, turning text into speech (and vice versa), translation, language analysis, chatbots, and more.
“Over time these technology giants have greatly refined their machine learning models for applications like product recommendations, even understanding language semantics when people ask questions during online searches,” said Mark Henman, TDK Technologies Chief Technology Officer. “Now the companies are offering their models as a service, which removes a significant barrier for businesses that could greatly benefit from improved product searches on their individual websites.”
The Possibilities of Mainstream Machine Learning
Machine learning involves using algorithms and statistical models to analyze data and teach computer systems how to answer specific questions or manage certain tasks. (By contrast, Artificial Intelligence is a broader concept that involves making a system seem more lifelike or human). Machine learning systems learn each time they deliver an answer, increasing the probabilities that the results are accurate and trustworthy.
The concept of machine learning has been around for decades, well before high powered computers and technical expertise existed to make this level of scalability and availability possible. That all changed thanks to the cloud. Large providers are leveraging the results of their own systems – not to mention the huge computers, large memory, and lots of storage available in cloud computing – to make machine learning an affordable commodity.
“These machine learning systems are pre-trained and proven for a number of specific use cases. That means it’s easy to get one, train it, feed it data, do some fine tuning and have it help a client’s business by performing whatever action it was trained to do,” Henman said. “You don’t need Amazon's entire product list or Google's massive search index. But you can point the Amazon Web Services (AWS), Google or Microsoft engine at your own website and use that same technology to help people find relevant information right there. That way you’re not sending customers to Google and hope they come back.”
Machine Learning and Revenue
Take an example where a developer is building a website with a lot of products and a reasonable amount of existing sales data. They could choose to create a product search function from scratch, which would require lots of time and expense. Or the developer could choose to utilize a service like AWS to create a robust search engine and product recommendation system tailored to the website. All that is required is to utilize the AWS machine learning engine, algorithm and training data - but replace Amazon’s production information with the business’ own information - and it's ready to go. The same concept could be used to add functions such as image classification or text searching.
“It removes a whole lot of barriers for entry. Monthly expense would be a miniscule fraction of what it would cost to develop from scratch,” Henman said. “Developers can then focus the organization’s precious resources elsewhere to solve other business problems for them.”
Machine Learning and Cost Saving
Another area where machine learning really comes in handy is looking for anomalies, which is helpful in analysis of sales trends. The systems can learn all types of tendencies including website traffic patterns, how many visitors put items into their carts, the length of time involved in the check-out process, which products are selling compared to those which are not, and much more. The system can learn to identify changes in the tendencies, which may prompt managers to take appropriate action more quickly.
Machine learning can also help identify vulnerabilities in systems, by providing notifications of unusual activity that may indicate security risks. This functionality can identify the way both humans and other machines are interacting with systems and analyze whether irregular behavior may pose a problem.
“Machine learning provides a lot of resources to identify behavior that converts shoppers to customers for increased revenue,” Henman said. But it can also identify behavior that could lead to loss. So, machine learning helps answer the question, “How do I maximize my revenue and minimize my loss?”
When organizations tap the potential of machine learning, which is now well within reach for organizations of all sizes, the possibilities are nearly endless. Machine learning is more common, available and accessible to developers everywhere thanks to cloud computing and the willingness of major vendors to offer the services on the open market. The concept of mainstream machine learning is a cost-effective tool that allows software developers to reallocate resources in ways that make systems more robust, more secure, more predictive, and enhance user experience.