Humans Confuse AI with Unpredictable Pandemic Purchases

Our online searches and shopping habits have been unpredictable, so machine learning systems don't know what to predict next.
Fabienne Lang

A surprising character has fallen into the grips of the pandemic: Artificial Intelligence (AI). The typically next-to-flawless machine learning systems are suffering a bout of confusion because our online habits have drastically changed during these times. 

With so much chaos and uncertainty floating around the world right now, it's no wonder the usually predictable AI algorithms are struggling to keep pace with our human volatility. 

MIT Technology Review reported on the matter, pointing out that algorithms for big companies such as Amazon are struggling to keep up with these swift changes.


AI still needs human interaction

AI algorithms are built to recommend products on sites, taking in new search data, and adapting as appropriate. But right now, as MIT Tech's report pointed out, people's online search and purchasing habits have taken a 180 degree turn. Now, these algorithms are finding themselves stumped. 

For instance, Amazon's top searches during normal times fall under the categories of phone cases, phone chargers, Lego, and more, but as the pandemic started to spread across the world these searches changed to toilet paper, hand sanitizer, N95 masks, Clorox wipes, and such. These are items that people don't usually buy, let alone in bulk and so widespread across the world. 

London-based consultancy firm, Nozzle that specializes in algorithmic advertising, put together a simple graph that shows the timeline of when certain countries' searches switched to COVID-19 items. It's a fast turnaround time for algorithms to keep up, too fast it seems. 

Humans Confuse AI with Unpredictable Pandemic Purchases
Nozzle's graph shows the timeline of countries' surge in COVID-19 supply searches, Source: Nozzle

Humans are now having to keep a closer eye on their AI algorithms and step in to redirect them in an appropriate direction. For example, MIT Tech reported that a company that detects credit card fraud had to step in and change its algorithm to take into account a jump in interest in gardening and power tools. 

Most Popular

Then, an online retailer found its algorithm was ordering the wrong stock that no longer matched what customers were searching for, so had to step in as well. 

"The situation is so volatile," Rael Cline, CEO of Nozzle, told MIT Tech. "You’re trying to optimize for toilet paper last week, and this week everyone wants to buy puzzles or gym equipment."

Many companies are seeing this as an opportunity to improve their AI algorithms, though, teaching them to predict more volatile climates. And even though machine learning systems are extremely useful, they still need a physical helping hand now and then.