Artificial Intelligence (AI), Machine Learning, and Deep Learning are all topics of considerable interest in news articles and industry discussions these days. However, to the average person or to senior business executives and CEO's, it becomes increasingly difficult to parse out the technical differences which distinguish these capabilities. Business executives want to understand whether a technology or algorithmic approach is going to improve business, provide for better customer experience, and generate operational efficiencies such as speed, cost savings, and greater precision. Authors Barry Libert and Megan Beck have recently astutely observed that Machine Learning is a Moneyball Moment for Companies.
State of Machine Learning
I met last week with Ben Lorica, Chief Data Scientist at O'Reilly Media, and a co-host of the annual O'Reilly Strata Data and AI Conferences. O'Reilly recently published their latest study, The State of Machine Learning Adoption in the Enterprise. Noting that "machine learning has become more widely adopted by business", O'Reilly sought to understand the state of industry deployments on machine learning capabilities, finding that 49% of organizations reported they were exploring or "just looking" into deploying machine learning, while a slight majority of 51% claimed to be early adopters (36%) or sophisticated users (15%). Lorica went on to note that firms identified a range of issues that make deployment of machine learning capabilities an ongoing challenge. These issues included a lack of skilled people and ongoing challenges with lack of access to data in a timely manner.
For executives seeking to drive business value, distinguishing between AI, machine learning, and deep learning presents a quandary, as these terms have become increasingly interchangeable in their usage. Lorica helped clarify the distinctions between machine learning (people teach the model), deep learning (a subset of machine learning characterized by layers of human-like "neural networks") and AI (learn from the environment). Or, as Bernard Marr aptly expressed it in his 2016 article What is the Difference Between Artificial Intelligence and Machine Learning, AI is "the broader concept of machines being able to carry out tasks in a way that we would consider smart", while machine learning is "a current application of AI based around the idea that we should really just be able to give machines access to data and let them learn for themselves". What these approaches have in common is that machine learning, deep learning, and AI have all benefited from the advent of Big Data and quantum computing power. Each of these approaches relies on access to data and powerful computing capacity.
Automating Machine Learning
Early adopters of machine learning are findings ways to automate machine learning by embedding processes into operational business environments to drive business value. This is enabling more effective and precise learning and decision-making in real-time. Firms like GEICO, through capabilities such as their GEICO Virtual Assistant, have made significant strides through the application of machine learning into production processes. Insurance companies, as an example, may implement machine learning to enable the offering of insurance products based on fresh customer information. The more data the machine learning model has access to, the more customized the proposed customer solution. In this example, an insurance product offer is not predefined. Rather, using machine learning algorithms, the underlying model is "scored" in real-time as the machine learning process gains access to fresh customer data and learns continuously in the process. When a firm employs automated machine learning, these models are then updated without human intervention since they are "constantly learning" based on the very latest data.
Real-Time Decision Making
For businesses today, growth in data volumes and sources -- sensor, speech, images, audio, video -- will continue to accelerate as data proliferates. As the volume and speed of data available through digital channels continues to outpace manual decision-making, machine learning can be used to automate ever-increasing streams of data and enable timely data-driven business decisions. Today, organizations can infuse machine learning into core business processes that are connected with the firm's data streams with the objective of improving their decision-making processes through real-time learning.
Businesses that are at the forefront in the application of machine learning are using approaches such as creating a "workbench" for data science innovation or providing a "governed path to production" which enables "data stream model consumption". Embedding machine learning into production processes will help ensure timely and more accurate digital decision-making. Organizations can accelerate the rollout of these platforms in ways that were not achievable in the past through techniques such as the Analytics Workbench and a Run-Time Decision Framework. These techniques provide data scientists with an environment that enables rapid innovation and helps support increasing analytics workloads while leveraging the benefits of distributed Big Data platforms and a growing ecosystem of advanced analytics technologies. A "run-time" decision framework provides an efficient path to automate into production machine learning models that have been developed by data scientists in an analytics workbench.
Driving Business Value
Leaders in machine learning have been deploying "run-time" decision frameworks for years. What is new today is that technologies have advanced to the point where machine learning capabilities can be deployed at scale with greater speed and efficiency. These advances are enabling a range of new data science capabilities including the acceptance of real-time decision requests from multiple channels while returning optimized decision results, processing of decision requests in real-time through the execution of business rules, scoring of predictive models and arbitrating among a scored decision set, scaling to support thousands of requests per second, and processing responses from channels that are fed back into models for model recalibration. Firms are deploying run-time decision frameworks with embedded and scalable machine learning, and are achieving notable results:
Establishment of a demand funnel of 90 use cases for a next-generation advanced analytics platform
- Reduction of run time for analytics by more than 90%
- Support for millions of decision requests per day
- 46% more leads measured versus a control group
- $100M+ incremental annual revenue measured versus a control group.