Although the news of Indian workers supporting Amazon’s automated stores came a few weeks ago, anecdotes like this have been happening for decades. There are several occasions in which artificial intelligence turned out not to be so.
Silicon Valley’s hasty race towards the development of AI has made companies want to get on the bandwagon as soon as possible, and to seek to generate hype about new products based on this technology to dazzle the market.
However, in recent weeks we have seen some disappointment that adds to a list of processes supposedly based on artificial intelligence that, in the end, turned out to be needing human help to function.
Below we list some of the technologies that, according to their creators, worked completely automatically and yet needed a push that no one could find out about.
An AI that creates comic scripts
Last January, the heirs of American comedian George Carlin sued the hosts of a podcast for allegedly using an AI model to impersonate the comedian in a special episode. The program, called “George Carlin: I’m Glad I’m Dead” could be made, in theory, by an artificial intelligence fed with five decades of Carlin’s work.
However, when faced with the complaint, the creators of the podcast had no choice but to admit the truth. The script had actually been written by a person named Chad. This episode of the show is no longer available on YouTube.
Automatic attendant in automatic food restaurants
Several American fast food chains have tested the effectiveness of some artificial intelligence tools, such as voice recognition. However, the result is ultimately not as surprising as it was thought to be. Specifically, the Del Taco company began using a voice assistant developed by the company Presto Automation in 2022. Although the food company assured in 2023 that the new tool was going smoothly, the chain announced in 2024 that it was going to stop using this artificial intelligence assistant.
Del Taco was using this ‘smart’ voice to take down orders from drivers passing through the drive-through. However, at no time was the company able to leave AI alone. Restaurant employees still had to check orders and make sure they were correct. In the end, this assistant was able to take care of only 30% of orders without requiring human help.
The AI that manages emails and meetings
The startup x.ai promised an AI assistant capable of reading emails and taking care of the user’s agenda by annotating and scheduling meetings. This virtual assistant, which was launched in 2014, was named Amy Ingram, and the way she spoke was incredibly similar to that of a human person. The truth is that the company had hired a team of AI trainers who analyzed as much as possible the emails that the assistant captured.
The work of these coaches was long and tedious, dedicating themselves to clicking and highlighting phrases in emails for up to 12 hours a day, as stated in a report by Bloomberg. This company closed in 2021.
The smart errands app
An artificial intelligence application called GoButler claimed to be able to do tasks ranging from booking plane tickets to ordering French fries for delivery.
As in the previous case, in reality a staff of employees remained on the other end of the line, supervising and even acting on behalf of the bot. These people even read sexual messages sent by users, who thought that no human eye could read them.
‘M’, Facebook’s AI
In 2015, Facebook launched an experiment. It was called M, an automatic service embedded in the platform’s messaging application that seemed to be able to answer any question and perform a multitude of tasks. For example, make a drawing. This is much less impressive now, ten years later, but it surprised many users back then.
This chatbot responded to what you said and sent you memes. I could also write the lyrics to a song or tell you a story. The kind of thing ChatGPT does now. However, disappointment. Each of the messages M received were reviewed by people working at the company’s headquarters in Menlo Park.
Amazon’s cashier-less stores
This is the most recent case. In 2021, Amazon introduced a new automatic grocery store model with “Just Walk Out” technology. This technology allowed customers to leave the store without having to check out. As customers placed the products they wanted in the basket, sensors spread throughout the space detected what product it was and added the price to the receipt, which was then charged to the person through their Amazon account.
This invention seemed to glimpse the future without human cashiers in supermarkets. However, it was a few weeks ago, in the middle of the progressive closure of these stores, when it was discovered that some 1,000 workers from India were verifying that the system behind all these stores was working correctly. Finally, Amazon has shelved this physical store project.
stochastic monkeys
@monospodcast
Another “artificial intelligence technology” that was actually a thousand Indians working remotely.
Amazon closes “Just Walk Out”, the supermarket purchasing system without cash registers. They said that with cameras and sensors… but in reality also human eyes https://t.co/tSa9auJnOn
April 12, 2024 • 9:35 p.m.
126
13
