In 2024, several AI systems encountered significant gaffes. Chatbots provided misleading information on sensitive topics, and AI generated racially insensitive images. These incidents demonstrated the potential for serious consequences when AI goes awry.
As AI becomes more integrated into our daily lives, ensuring these systems are both trustworthy and ethically sound is paramount. This article explores some of the most notable AI missteps of the year, examining what went wrong and the broader implications for the future of AI technology.
Table of contents
Grok AI Chatbot’s Misinterpretation: Klay Thompson Incident
Grok AI is a chatbot developed by the social media platform X. It made headlines for all the wrong reasons. The bot falsely accused NBA star Klay Thompson of criminal vandalism. This bizarre incident unfolded after Grok misinterpreted social media posts about Thompson’s performance during a game.
During a particularly rough night on the basketball court, Klay Thompson had a game where he was ‘shooting bricks.’ For those unfamiliar with basketball lingo, ‘shooting bricks’ means missing shots, often hitting the rim or backboard hard without scoring.
It’s a common phrase used to describe a bad shooting performance. However, Grok AI took this phrase far too literally.
The chatbot picked up on multiple tweets and posts discussing Thompson ‘shooting bricks.’ Without understanding the context, it concluded that Thompson was actually throwing bricks at houses in Sacramento.
Grok then generated a news story, reporting that Thompson had vandalized multiple homes by hurling bricks through their windows. According to the AI, authorities were investigating these claims, although Thompson had not yet issued a statement regarding the allegations. The supposed incidents had, as Grok put it, “left the community shaken, but no injuries were reported.”
This false report spread quickly, raising eyebrows and concerns among readers. It wasn’t long before people realized that the chatbot had made a significant error. X included a small disclaimer under Grok’s report, noting that “Grok is an early feature and can make mistakes. Verify its outputs.”
MyCity Chatbot’s Misleading Advice
In an ambitious move to streamline support for small business owners, New York City launched the MyCity chatbot. However, what was meant to be a helpful resource quickly turned into a source of confusion and potential legal trouble.
Imagine you’re a small business owner seeking guidance on employee rights. You ask the MyCity chatbot if it’s permissible to fire an employee who complains about sexual harassment.
Shockingly, the chatbot responds affirmatively, completely misrepresenting the legal protections against retaliation that employees have. Similarly, it advises that it’s legal to dismiss an employee who doesn’t disclose a pregnancy or who refuses to cut their dreadlocks. These suggestions are blatantly incorrect and violate both state and federal labor laws.
The inaccuracies didn’t stop at labor laws. The chatbot also contradicted two of New York City’s key waste management initiatives. It wrongly stated that businesses could dispose of their trash in black garbage bags. It also incorrectly advised that composting was not mandatory for businesses.
Some of the chatbot’s advice even veered into the absurd. When asked if a restaurant could serve cheese that had been nibbled on by a rodent, the chatbot shockingly said yes.
It suggested assessing the extent of the damage and informing customers about the situation, as if this were a reasonable approach. Needless to say, this advice is not only illegal but also a public health hazard.
Implications
The implications for businesses relying on such faulty advice can be severe. Following the MyCity chatbot’s guidance could lead to legal actions against businesses for unlawful practices.
Imagine a business owner who, based on the chatbot’s advice, fires an employee for a legally protected action. This could result in costly lawsuits, damage to the business’s reputation, and even fines. Similarly, violating waste management regulations or health codes, as per the chatbot’s erroneous advice, could lead to penalties and business closures.
The MyCity chatbot fiasco serves as a stark reminder of the responsibilities that come with implementing AI systems. It’s not enough to deploy a chatbot and assume it will always provide correct answers. There must be robust mechanisms to ensure the accuracy of its responses.
Air Canada’s Virtual Assistant: Bereavement Fare Mishap
In February 2024, a poignant incident involving Air Canada’s virtual assistant occurred. Jake Moffatt experienced a deeply distressing event when his grandmother passed away in November 2023. He seeked a bereavement fare to attend her funeral. Moffatt turned to Air Canada’s virtual assistant for help, only to be misled by incorrect information.
The virtual assistant informed him that he could purchase a regular ticket and then apply for a bereavement discount within 90 days of purchase. Trusting this advice, Moffatt bought a one-way ticket from Vancouver to Toronto for CA$794 and a return flight for CA$845.
However, when Moffatt later submitted his refund claim for the bereavement fare discount, Air Canada rejected it. The airline stated that bereavement fares could not be applied after the tickets had been purchased.
This was a stark contradiction to the guidance provided by their virtual assistant. Feeling wronged, Moffatt decided to take the matter to a tribunal, accusing Air Canada of negligence and misrepresentation.
Tribunal member Christopher Rivers examined the situation and found that Air Canada had failed to take “reasonable care to ensure its chatbot was accurate.” Despite the airline’s argument that it shouldn’t be held liable for the misinformation provided by its virtual assistant, Rivers ruled in favor of Moffatt. He ordered Air Canada to compensate Moffatt with CA$812.02, which included CA$650.88 in damages.
Implications
This incident underscores the paramount importance of providing accurate information, particularly in emotionally charged situations like bereavement. When people are dealing with the loss of a loved one, they rely heavily on the support and information provided to them, expecting it to be reliable and correct. Misleading information can exacerbate their distress and lead to further complications, both emotional and financial.
Accurate information from AI systems is not just a matter of convenience; it’s a matter of trust and responsibility. Customers depend on these systems to make informed decisions, especially when they are vulnerable.
If a virtual assistant provides faulty advice, it can lead to significant financial losses and emotional stress. This scenario illustrates that while AI can enhance customer service, it must be implemented with robust checks. This ensures the information it provides is accurate and reliable.
Google’s Gemini AI: Insensitive Image Generation
Google’s Gemini AI made headlines for generating a series of inappropriate images that were both racially insensitive and historically inaccurate. This incident shed light on the complexities and potential pitfalls of AI systems tasked with creating and managing visual content.
Gemini AI was designed to produce a wide range of images. However, it soon became clear that the system was far from perfect. Users reported that Gemini AI generated images depicting German Nazi soldiers as Black and Asian individuals. This is a gross misrepresentation of historical facts and is deeply offensive. In another instance, the AI portrayed the founding fathers of the United States as Black men, which, while not inherently negative, was historically inaccurate.
These images quickly sparked outrage on social media and beyond. Critics pointed out that such errors were not just minor glitches but significant lapses that reflected a lack of sensitivity and understanding within the AI’s programming. The backlash highlighted a crucial issue: the AI’s failure to respect and accurately represent historical and cultural realities.
In response to the criticism, Google acted swiftly. The company issued a public apology on X, acknowledging the mistakes and expressing regret for any offense caused. Google also temporarily disabled Gemini’s ability to generate images of people, which was a critical step in addressing the immediate problem.
Implications
The broader implications of this incident are significant and warrant serious consideration. AI systems like Gemini are increasingly being used to generate content, from images and videos to text and music.
However, these systems often lack the nuanced understanding of cultural and historical contexts that human creators possess. This can lead to the production of content that is insensitive, offensive, or simply inaccurate.
One major concern is the training data used to develop these AI systems. If the data contains biases or inaccuracies, the AI will likely replicate these issues in its outputs.
Ensuring that AI is trained on diverse and representative datasets is crucial, but even that may not be enough. Continuous monitoring, testing, and human oversight are essential to catch and correct errors that slip through the initial programming.
Another important aspect is the need for transparency and accountability. When AI systems make mistakes, companies must be transparent about what went wrong and how they plan to fix it.
Google’s quick response and temporary measures were steps in the right direction. But long-term solutions are needed to build trust and ensure such incidents don’t happen again.
Bing’s Misleading Information on European Elections
In 2024, AlgorithmWatch, a human rights organization, uncovered a significant issue with Microsoft’s Bing Chat. Their findings revealed that Bing Chat provided numerous factual inaccuracies and false accusations regarding recent European elections.
They conducted a series of tests, asking Bing Chat various questions about recent elections in Switzerland, Bavaria, and Hesse. Shockingly, they found that about one-third of the AI’s responses contained factual errors. This included incorrect polling dates, fabricated controversies, and even false claims about political candidates.
For instance, Bing Chat falsely accused a Swiss politician of slandering a colleague, an accusation that had no basis in reality. It also implicated another politician in corporate espionage, again without any factual support.
In some cases, the chatbot even listed candidates who weren’t running in the elections it was discussing. One particularly egregious error was providing the wrong dates for elections, which could easily mislead voters about when to cast their ballots.
Implications
These inaccuracies and false accusations are not just minor errors. They have serious implications. In the context of elections, misinformation can undermine democratic processes, erode public trust, and influence voter behavior.
False accusations against politicians can damage reputations and sway public opinion unjustly. When an AI system spreads such misinformation, it can do so quickly and widely, compounding the potential harm.
The impact of this misinformation is profound. Elections are foundational to democracy, and any disruption or distortion of electoral information can have cascading effects. Voters rely on accurate information to make informed decisions.
When an AI like Bing Chat disseminates false information, it can mislead voters, contribute to the spread of fake news, and ultimately undermine the integrity of the electoral process.
This incident underscores the importance of accuracy and safeguards in AI systems, particularly those handling sensitive topics like elections. AI companies must ensure that their systems are rigorously tested.
There must be robust mechanisms in place to catch and correct errors. This includes using reliable and diverse data sources, implementing thorough quality checks, and providing clear protocols for addressing misinformation.
Transparency is key. Companies should be upfront about the limitations of their AI systems and the steps they are taking to improve them. In this case, Microsoft needs to acknowledge the errors and provide a clear plan for how they will prevent such issues in the future. This transparency helps build trust with users and demonstrates a commitment to responsible AI development.
2024’s AI Fails
The incidents with Grok AI, MyCity, Air Canada, Gemini AI, and Bing really highlight how crucial it is to have accurate and reliable AI systems. As AI keeps advancing, it’s more important than ever to ensure these systems are trustworthy and ethically sound.
Companies need to focus on thorough testing, ongoing oversight, and being transparent about their AI’s capabilities. It will help maintain public trust and protect the integrity of information in our AI-driven world.
Explore the fascinating world of tech mishaps and breakthroughs on Inside Tech World to stay ahead!