The 6 Biggest AI Fails... So Far

Nobody's perfect - especially when it comes to algorithms.
Wikimedia Commons
Machines always do what you tell them. They never fail, and that’s usually their biggest problem. In a sense, machines only fail by succeeding in doing what we tell them. We gathered for you some of the biggest and most awful AI fails (so far) that made us feel really disappointed by technology.

Imagine your uncle being arrested for wishing "Good Morning" on Facebook

We all know how the elderly just L-O-V-E wishing each other Good Morning, Good Evening, and Happy Birthday publicly on Facebook. This is exactly what a Palestinian man did - he uploaded a photo of himself next to a bulldozer with the caption “‫”یصبحهم‬ meaning “good morning” in Hebrew. The social media automatically translated this to “attack them” and soon, a policeman arrested the innocent person.
A Facebook post saying "good morning" in HebrewA Facebook post saying "good morning" in Hebrew
A Facebook post saying "good morning" in Hebrew

Google's racist algorithm

This is probably one of the most worldwide-known AI scandals of our times, so we must mention it. In 2015 a software engineer mentioned a big problem in the Google Photos image recognition algorithm that mistakes gorillas for black people. What the hell, right? The tech company was quick to apologize and claimed to fix the problem. Regarding some reports, all they did was removing the monkeys from the training data.

Self-Driving Uber killed a 49-year-old woman

AI can be very dangerous too. Last year a test self-driving Uber vehicle hit and killed a woman in Arizona who was walking with her bike east across Mill Avenue. The pedestrian was in dark clothing and had no side reflectors on her bike facing the path of the oncoming vehicle.
Self-driving Uber (green) killed a woman (orange) on northbound Mill Avenue. Right - the vehicle after the incident.Self-driving Uber (green) killed a woman (orange) on northbound Mill Avenue. Right - the vehicle after the incident.
Self-driving Uber (green) killed a woman (orange) on northbound Mill Avenue. Right - the vehicle after the incident.

When Alexa won't assist you at the wrong time

Imagine leading a big lecture in front of thousands of people in Las Vegas when your voice assistant won’t assist you and even has the audacity to claim that you are not right. Really, Alexa? The situation happened at Consumer Electronic Show 2019 when a representative from Qualcomm Technologies talked about voice assistants' life-changing influence on next-generation vehicles. Talk about bad timing!

And when Alexa is too helpful at the wrong time

Everybody wants to sell something to the kids because who can say “No” to screaming and crying child who desperately wants a new toy! And when you thought that you went away with another pointless purchase, Alexa orders an expensive dollhouse for your kid without your knowledge.
A reporter said the sentence “Alexa ordered me a dollhouse” live on TV, which activated a helpful Echo in one household. The assistant decided to follow the order without asking. Imagine a terrifying future with hundreds of hidden messages in kids' commercials!

Taylor Swift wanted to sue Microsoft's AI

"TayTweets" once was an AI Twitter chatbot created by Microsoft - and by "once," we mean for not-so-good 16 hours. Very soon, the bot account started tweeting offensive tweets and even claimed that the genocide was fake. Тhis affected not only a large young audience but also Taylor Swift herself, known by the nickname “Tay.” The signer’s crew tried to sue Microsoft for the association between the popular singer and the chatbot and the caused damages to her public image.