Source – https://www.forbes.com/

For a moment, let’s set aside the abject tragedy of the Covid-19 pandemic and the demoralizing conditions through which the world continues to persevere. Instead, let’s examine the state of affairs from a dispassionate and scientific position. Seismic changes in behavior are erupting as the burden of the pandemic forces transformation. Crippling inefficiencies in industry and volatile projections of markets have led to unprecedented uncertainty.

To fully address any one of the challenges the world now faces would exhaust this medium. However, the role machine learning (ML) will play in people’s lives is fascinating and worthy of discussion, as the field will undoubtedly contribute to the impending metamorphosis. There are three periods of time to consider: the time before the pandemic (BP), the time during the pandemic (DP) and the time after the pandemic (AP). Central to these three epochs are the two approaches to model human behavior through machine learning.

Canonical ML (CML), the first class of interest, represents traditional approaches in pattern recognition, derived from highly structured and labeled data through computational statistics. This approach is used to explain the state of a system or to predict behaviors. Its success often depends on engaged scientists to explain and interpret the model’s ascribed result. CML could be used to predict the weather in a particular region by analyzing the features of the region over time, where each of those features in isolation would fail to fully predict a future state. When CML models are contextualized, you’re able to explain how predictions are made. For example, you can predict rain in Austin tomorrow given the regression and ensemble models you’ve developed for various weather features in aggregate.

I’ll call the second class of techniques reinforcement ML (RML) as it deploys a fundamentally different modeling paradigm: In RML, models self-adjust their individual actions to optimize a collective outcome. These models operate much more autonomously than CML and fully embrace early failures in favor of long-term gains through repeated environment exploration and self-learning. Examples of RML include applications in autonomous driving, gaming, computer vision and even natural language processing. Only recently has RML become tenable for industrial deployments and is still met with much trepidation because of its unexplainable methods and unclear accountability. In other words, when RML models are correct, it is hard to trace what led to that specific output. CML outputs, on the other hand, often are easier to explain. Throughout my career, the applications of CML have represented the overwhelming majority of successful solutions, while RML, in its fledgling state, has only begun to transform the industrial world.

Qualcomm Highlights Mobile Audio With Snapdragon Sound

In 2019 (BP), CML reached its zenith. Nearly every industry was disrupted to some degree by machine learning. From financial services to healthcare to defense, leaders embraced the capabilities of a robust data science solution capitalizing on CML, often materialized into what is commonly known as “deep learning.” Scores of historical data and behavioral modeling contributed to measurable ROI on data science initiatives and applications. Many companies had deployed CML, and some had begun to experiment with RML. Companies around the world were transforming their industries through CML on their own terms. On the surface, the union between science and industry was thriving.

Then the pandemic enveloped the planet. Overnight, the pandemic obliterated the utility of millions of models. Every sector that had benefited from CML was in a difficult position: Companies had to either trust that the ML models their businesses depended on would correct over time or reverse course and manually drive mission-critical insights for their business. I believe CML has failed many businesses across many industries, and the business world has yet to realize the full effects of these failures.

The companies that adopted RML before the pandemic, however, may have an advantage over their peers, as RML models are not as dependent on finely tuned conditions from a scientist, but rather seek to optimize for success as defined by scientists. While RML requires exorbitant amounts of data for training, the increase in the frequency of data collection has eased that challenge in some cases.

Topically, the post-pandemic era will likely resemble the pre-pandemic era but with a heavier slant to digital behavior, as well as customer behavior based on new habits and efficiencies identified during the pandemic. Once industry realizes advantages gained by the firms that adopted RML before the pandemic, I expect there will be an algorithmic arms race. For example, an RML approach for product recommendations for an online retailer will likely adapt to the wildly new engagement model of the post-pandemic epoch. The advantages over the retailer’s CML pre-pandemic competitors will be decisive. Simply put, I believe RML techniques are far more robust at predicting behavior post-pandemic than techniques using CML. Those that are successful in the adoption will have a higher chance to survive and differentiate from their competition.

But what of the explainability of RML? The pressures of the pandemic will greatly shift industry’s willingness to deploy unexplainable or opaque models, or “black box models” as they’re often referred to. As the advantage for the few through RML becomes clear, many firms will likely forgo the accountability of CML in favor of RML’s adaptability. It is the AI equivalent of the adoption of telehealth or remote work during the pandemic, and is arguably much more impactful. Scientists must now work to ensure RML techniques that are deployed can be responsible and accountable or they might compromise the integrity of their operations.

There are many reasons to be excited for the next frontier of commerce. Industries have evolved their priorities, shifted relationships and in many ways removed tedious operations, such as pattern recognition based on outdated labeled data sets that are relics of former industrial epochs. ML will continue to play an integral role in industrial transformation, and as it adapts to the changes people have made in their own lives, I trust my colleagues and peers across industry to ensure we develop this capability in a way that is inspirational, dynamic and responsible.