A primer for understanding reinforcement learning
Reinforcement learning, a type of artificial intelligence (AI), is fast becoming a mainstream business tool, helping companies improve business processes and deliver better customer experiences (CX). For example, some telecom companies are improving CX with reinforcement learning algorithms that automatically reduce network congestion and optimize cell phone connectivity, and many eCommerce platforms are using it to customize recommendations for users.
In this article, we’ll introduce the concept of reinforcement learning and provide some practical applications for business process improvement and CX design.
What is reinforcement learning?
Reinforcement learning, one of three primary machine learning paradigms, enables the creation of intelligent agents, which are computer programs capable of making decisions. Reinforcement learning differs from supervised learning and unsupervised learning algorithms primarily in how they use datasets, the former of which doesn’t require as much input data.
It’s also different from deep learning, in which neural networks — which mirror the way the human brain works — are used to process extensive amounts of data.
Unlike other forms of machine learning, reinforcement learning agents are dynamic and learn from trial-and-error and continuous feedback from their environment, rather than learning from patterns in datasets. They are typically based on the Markov decision process, which provides a mathematical framework for modeling sequential decision-making. Operating much like humans do, reinforcement learning is particularly useful in uncertain and complex environments.
Comparitively, supervised machine learning models are trained with labeled datasets, enabling the models to learn and improve over time, while unsupervised machine learning algorithms look for patterns in unlabeled data. This type of program might, for instance, parse online purchasing data in order to segment different types of customers. By contrast, reinforcement learning models might learn from customer feedback in order to personalize recommendations.
Practical applications of reinforcement learning
Reinforcement learning has numerous practical applications, ranging from chatbots and medical diagnostics, to autonomous vehicles and robotics.
In fact, many personalized or transactional chatbots that are used by brands to deliver customer support are made possible thanks to reinforcement learning algorithms. Customers interact with a conversational bot that uses machine learning and natural language processing to understand requests and respond accordingly. The evolution of this application is being used for conversational AI, like Siri and Alexa. For those looking to get started, choosing a suitable approach can seem tricky. Here’s how to select the right machine learning algorithm.
Reinforcement learning is also helping doctors analyze massive medical imaging datasets in order to improve diagnostics. GE Healthcare, for example, has optimized machine learning algorithms to analyze X-ray imagery, making it possible to automatically detect a pneumothorax (collapsed lung) within seconds.
Moreover, reinforcement learning is helpful for training mobile robots to learn new behaviors. For example, think about how a robot vacuum cleaner learns how to avoid obstacles. Based on feedback from its environment, the robot learns new behaviors to keep it from colliding with walls and furniture.
Despite impressive applications, reinforcement learning is not without its challenges. Given that it is designed to work with limited data, reinforcement learning may not deliver accurate data logging, choose the best reward structures or make the models reproducible.
The essential guide to AI training data
Discover best practices for the sourcing, labeling and analyzing of training data from TELUS International, a leading provider of AI data solutions.
The future for reinforcement learning
Reinforcement learning is already mainstream, and the future promises to make it ubiquitous. That potential is helping to drive investment in AI technology overall. According to Stanford University’s Artificial Intelligence Report 2021, the total global investment in the technology increased by 40% in 2020 relative to 2019, for a total of $67.9 billion.
Gaming is one of the many industries that reinforcement learning is poised to transform. New features are already starting to roll out, such as text-to-speech and synthetic voices. AI also promises to improve processing and security, and to help companies scale games as user engagement grows.
Another use of AI that is gaining traction is the use of reinforcement learning to help shape and implement public policy. In fact, reinforcement learning is being used to help public health officials in the battle against COVID-19. As decribed by one of the authors of a report in the PLOS One journal, “We used deep reinforcement learning, and the algorithm was trained to enable agents to try to find optimal public health strategies that maximized total reward on controlling the spread of COVID-19.”
Reinforcement learning is also finding its way into other business decision-making processes. “We believe a reinforcement learning framework is well-suited for uncovering insights on how the behavior of economic agents could be influenced by pulling different policy ‘levers,’” said Richard Socher, chief scientist at Salesforce, in a recent Forbes report.
Other ways in which brands can use reinforcement learning include creating personalized recommendations, optimizing advertising budgets, choosing advertising content, increasing customer lifetime value and predicting customer responses to actions such as price increases, among others.
And as reinforcement learning improves, it will only increase its ability to help optimize business decision-making while having a positive impact on the world’s most pressing challenges.
Check out our AI Data solutions to learn more about how TELUS International can help you put reinforcement learning to work.