The Amazon Go Store Was a Lie: Is AI Advancement Stalling?

Not long ago, the Amazon Go store model captured headlines and people’s fascination. Busy shoppers loved the idea of popping into one of the locations, purchasing their essentials and leaving the store without ever interacting with workers. However, news recently broke that a labor force was enabling these shopping experiences — just behind the scenes.

What Is the Amazon Go Store Experience Like?

Amazon centers its Go convenience stores on the Just Walk Out model. It works like it sounds and caters to busy people. Shoppers stroll into the unstaffed stores, put their desired items into their own bags or Amazon-provided ones, and leave once they have everything they need. 

A breakdown of the process on an Amazon webpage explained that people can pick up and put back items during each shopping trip, but the store’s integrated technology will only charge them for what they walk out with. Those transactions are automatically charged to a consumer’s chosen payment method. 

Although Amazon Go store locations have people working in them, and visitors can go through the traditional checkout process if preferred, the main draw of these retail outlets is that people can enjoy frictionless shopping, free from the interpersonal interactions that might otherwise slow things down. 

Amazon’s stance is that various technologies — facial recognition, shelf sensors and artificial intelligence — work together to make these futuristic stores operate. Although each location uses those technologies, it also has a hidden human workforce.

How Did the Amazon Go Store Experience Really Work?

An April 2024 report revealed that more than 1,000 people in India watched Amazon Go consumers as they shopped, verifying what they left the stores with. Statistics indicated that humans reviewed approximately 70% of the overall sales made. 

However, Amazon Just Walk Out technology was the focus of a corporate blog post published soon after the allegations of a concealed Indian workforce. A subheading addressed the question, although the copy underneath answered the question in an arguably roundabout way. The vast majority of the post details Amazon’s expansion of its various forms of in-store shopping technologies. However, the comparatively smaller section addressing allegations of the Indian workers behind the AI curtain takes a defensive and tech-focused tone. 

It explains how weight and vision sensors and deep learning algorithms detect what people purchase in Amazon Go stores and generate accurate receipts based on that information. However, it argues against the “erroneous reports” that distantly located humans supervise and verify the shopping experience. Rather, those people assist with data labeling and annotation, making the machine learning algorithms more accurate. 

A Different Angle to AI Job-Loss Fears

A much-discussed aspect of the artificial intelligence boom involves people worrying that they’ll lose their jobs as technologies advance. However, the tech world has a long history of augmenting its technologies with humans while concealing that fact from investors.

Perhaps a more realistic worry than AI itself taking people’s jobs away is that tech executives will move jobs to countries where they can pay workers less to do tasks that make artificial intelligence tools function. 

That does not mean that AI job loss is out of the question. After all, one restaurant consultancy believed automation could replace 81% of its industry’s human-staffed roles. However, people must remain aware that outsourcing could be a bigger threat, especially as tech execs try to save money on massively expensive applications. 

Such arrangements are not far-fetched either. A May 2024 article detailed how ordering from a New York City fried chicken restaurant involves telling a cashier on a screen what you want. Everything happens via a Zoom call, and the person on the other end is thousands of miles away. Someone in the restaurant’s kitchen prepares the food and brings it out. However, the cashier makes $3 for their off-site labor. 

The author discussing this arrangement suggests that AI will not take jobs, but it will shrink a position’s responsibilities — and the associated paycheck. That could result in a lose-lose situation for workers and consumers. Those filling those low-wage positions in distant countries likely receive unfair compensation for their work. Relatedly, conscientious consumers feel less than agreeable after realizing their fried chicken cravings exacerbate worker exploitation. 

What Does Human-Augmented Functionality Mean for AI’s Future? 

Should news of the Amazon Go store sham discourage people about artificial intelligence’s potential? Not necessarily, but they should treat it as a wake-up call to not immediately believe everything they read. Even when people do browse the headlines, they’ll find plenty of examples of promising AI capabilities. 

Identifying New Pregnancy Risk Factors

Researchers used artificial intelligence to look for patterns in the details of 9,558 pregnancies. That approach revealed new risk factors that could cause stillbirths or other complications. One finding was that underweight female fetuses usually have less risky outcomes. However, the risks rise when the pregnant people carrying them have diabetes. This new knowledge aids health care workers in providing tailored, appropriate care to reduce childbirth risks. 

Falling Short in Clinical Settings

At the same time, ongoing research shows numerous areas where AI needs work. Although large language models perform well when answering medical test-style questions, a group found they are less adept at offering diagnostic support based on the question-and-answer exchanges common to many doctor visits. The researchers used their conclusions to develop better ways to evaluate AI models’ readiness for and accuracy in clinical settings.

Finding Realistic Uses for AI

The above examples underscore why the best AI applications may be those associated with tasks humans cannot do well without help. It could take people weeks to scrutinize the vast quantity of data in the pregnancy risk study. However, artificial intelligence algorithms can easily handle it and find helpful material faster than humans working alone.

Humans have filled cashier roles and worked as doctors for decades. Yet, many tech companies want AI to determine how much to charge a consumer during a shopping trip or diagnose their illnesses. In addition to the Amazon Go store experiment, numerous AI doctor chatbots allow people to input their symptoms to learn their potential causes. However, physicians who have tried them reported that the number of false negatives and positives made these tools useless. 

Is it not wiser to prioritize AI investments that continue to allow humans to do the roles they can and should fill while letting technologies fill the gaps? Then, technology can partner with people rather than replace or exploit them.

Treating the Amazon Go Store Reveal as a Lesson Learned

No matter if people believe Amazon’s stance about its Indian workforce’s function, that case reminded the world that AI applications are not necessarily as advanced as they seem. Even so, artificial intelligence has already achieved some amazing things and people are best equipped to analyze them when they understand its advantages and limitations.