The crisis has caused massive fluctuations in consumer purchasing habits as people opt out of big-ticket items in favor of household consumer goods. The week of April 12-18, 2020 saw items such as toilet paper, face masks, hand sanitizer, and paper towels rise to the top of Amazon search results. Nobody could have predicted changes of this magnitude to occur - especially machines.
Typically, AI algorithms are responsible for managing the nuances of online marketplaces such as inventory management, fraud detection, and more. Machine-learning models are trained to analyze human behavior and correspondingly adjust for marketplace fluctuations. However, consumer spending habits have drastically shifted since the pandemic began, causing humans to reassess current AI algorithms and machine-learning models.
Machine learning models typically are used to track and analyze human behaviors within online marketplaces. The problem is they're no longer functioning correctly due to unpredictable consumer behaviors and habits. As the definition of normal continues to change, these automated systems require fine-tuning and adjustments.
Many have viewed AI as independent systems that operate with little to no human intervention. The problem, as one AI consultant stated, is "AI is a living, breathing engine." As a result, AI must maintain human interaction to ensure continued functionality. During the pandemic, AI mishaps have created issues in supply chains that didn't previously exist. Predictive algorithms forecasts' are no longer valid due to widespread consumer changes. Companies have been ill-equipped to handle such massive fluctuations, resulting in depleted inventories and supply-chain management issues.
Other aspects of our lives have also been impacted by AI system deficiencies, such as Netflix accounts. Many large streaming providers have experienced influxes of content-hungry subscribers forced to stay home due to widespread social distancing guidance. Although these models typically help users select future content, insufficient data has created inaccurate representations of unique user behavior.
Many businesses purchase machine learning systems without the intention of installing future updates. These organizations may lack the resources or knowledge necessary for proper maintenance, thereby creating system deficiencies. Properly retraining models require human assistance and regular system updates.
The pandemic has shown that AI must be continually trained on the nuances of daily life, including worst-case scenarios. No one predicted the impact of COVID-19, but creating efficient machine-learning models is essential for assessing human behavior adequately. As the world continues to change, systems must be able to adapt to current circumstances continually.
Amazon and other online retailers have struggled to adapt to COVID-19 during times of crisis. A volatile marketplace has caused fluctuations to occur in consumer behavior from week to week. COVID-19 has seen unfamiliar goods such as toilet paper, gym equipment, and puzzles all become top-selling items. Although consumer behavior may be predicted based on past purchases, the same cannot be said during COVID-19. Instead, delivery time is playing a key role in consumer behavior habits. Marketplaces that are currently offering expedited delivery times may experience improved sales.
During COVID-19, humans have become more keenly aware of potential algorithmic issues. To sufficiently manage these AI algorithms, dedicated teams must be able to keep tabs on current marketplace situations. Automated systems are not necessarily autonomous, as they still require human interaction and adjustments. People who trust machines must be able to provide adequate oversight to ensure proper functionality.
Read the white paper below to learn how Bitvore Cellenus can help alleviate manual burdens and drive better risk mitigation tactics in a time of uncertainty.