Recommended Reading: Why AI That Teaches Itself to Achieve a Goal Is the Next Big Thing

Amii Fellow-in-Residence and Canada CIFAR AI Chair, Matthew E. Taylor (Associate Professor, University of Alberta) co-authored an article for the Harvard Business Review on Why AI That Teaches Itself to Achieve a Goal Is the Next Big Thing. The article is co-written with Kathryn Hume, interim Head of Borealis AI, the machine learning research lab for the Royal Bank of Canada.

Amii researchers are pioneers and leaders in the field of reinforcement learning (RL), a branch of machine learning that enables AI systems to learn through experience. RL systems interact with their environments, often through trial and error, earning positive or negative rewards based on their actions. Humans define the overall task and relevant rewards that the system uses to discover the best action to take in a given situation.

Instead of being told what actions to take to achieve a goal, the system must learn which actions yield the most reward by trying them. Over time, the system develops a policy (or way of acting) that lets it select the action that will best achieve the goal, which can help us discover the optimal actions to take in a given scenario.

Reinforcement learning can be used for process optimization and improvement, as part of a recommender or intelligent tutoring system, and for adaptive control and decision making in autonomous systems.

With this focus on experience-based learning for decision-making, reinforcement learning differs from other types of machine learning like supervised learning, which requires labelled data as an input, or unsupervised learning, which focuses on finding similarities and differences in data points.

On reinforcement learning, the authors of the article explain: “Companies such as Netflix, Spotify, and Google have started using it, but most businesses lag behind. Yet opportunities are everywhere. In fact, any time you have to make decisions in sequence — what AI practitioners call sequential decision tasks — there is a chance to deploy reinforcement learning.”

We sat down with Taylor, who directs the Intelligent Robot Learning Lab at the University of Alberta where his current research focuses on fundamental improvements to reinforcement learning, the application of reinforcement learning to real-world problems and human-AI interaction.

“Most people recognize that AI and machine learning is changing the world in so many ways, and there is this underused technique of reinforcement learning that should be used in all different ways,” says Taylor. “When more people understand reinforcement learning, they welcome more opportunities to use this technology to do something new and useful in their companies.”

The article continues: “while reinforcement learning is a mature technology, it’s only now starting to be applied in business settings. The technology shines when used to automate or optimize business processes that generate dense data, and where there could be unanticipated changes you couldn’t capture with formulas or rules. If you can spot an opportunity, and either lean on an in-house technical team or partner with experts in the space, there’s a window to apply this technology to outpace your competition.”

The article provides a digestible overview of reinforcement learning, as well as examples of how companies are currently using the technique. The article also teaches business people how to spot an opportunity for reinforcement learning in five steps:

1. Make an inventory list of what you are trying to achieve
2. Consider other options and other techniques, if need be
3. Be careful what you wish for and think carefully about your desired outcomes
4. Determine whether the use of RL is worth it
5. Prepare to be patient

“We are educating people who might have never heard of this technology to learn more about it and see that it can be a game-changer for them when solving a wide range of problems,” says Taylor.

Read the full Harvard Business Review article here.

Learn more about reinforcement learning in the Reinforcement Learning Specialization offered by the University of Alberta and Amii on Coursera, developed and delivered by Martha White and Adam White – Fellows and Canada CIFAR AI Chairs with Amii.