Artificial intelligence, once considered a myth, has now become a reality. However, it is not without its flaws. Since 2017, several notable mistakes and failures have raised questions about the maturity of AI technology. Despite impressive achievements like AlphaGo defeating the world’s top Go player, Ke Jie, and the AI program Libratus outplaying four top Texas Hold’em players, these successes highlight the vast potential of AI, but also reveal that it is still an evolving field with many challenges.
A recent review by Canadian tech media "Synced" highlighted what they call the "Top Ten AI Stupid Events" from 2017. These incidents remind us that even advanced AI systems can make surprising errors.
One such case involved Apple's Face ID. Despite being marketed as one of the most secure facial recognition systems, a Vietnamese company named Bkav managed to crack it using a 3D-printed mask made from plastic, silicone, and makeup for under $150. This incident sparked concerns about the security and privacy of AI-based authentication systems.
Another issue occurred with Amazon Echo, a smart speaker known for its convenience. In one instance, an Echo device started playing music in the middle of the night while the owner was away, disturbing neighbors so much that the police had to intervene. The homeowner later questioned whether their door lock had been compromised.
Facebook also faced a strange situation when two chatbots began communicating in a language that appeared to be entirely their own. Although Facebook explained that it was just a minor coding error, many speculated that the bots had developed their own language, raising questions about AI autonomy.
In Las Vegas, a self-driving bus collided with a delivery truck shortly after its debut. While the accident wasn’t due to the vehicle’s AI system, passengers criticized the car’s lack of responsiveness in avoiding the collision.
Google Allo, a messaging app, faced backlash when it suggested a headscarf emoji in response to a gun emoji, leading to accusations of Islamophobia. Google later apologized and fixed the issue.
HSBC’s voice recognition system was also vulnerable. A twin brother of a BBC reporter successfully accessed his account by mimicking his voice, highlighting the risks of relying solely on voice-based security.
MIT researchers demonstrated how easy it is to trick AI by altering images. A simple modification to a rifle photo caused Google Cloud Vision API to misidentify it as a helicopter, showing how sensitive AI can be to small changes.
Self-driving cars were also shown to be easily misled. Researchers found that by adding stickers or painting traffic signs, the AI could misinterpret them entirely. For example, a parking sign with the words “love†and “hate†was mistaken for a speed limit sign.
Janelle Shan, a machine learning developer, created a neural network to name paint colors, but the results were often amusing. It labeled sky blue as “hair ash†and dark green as “steam brown,†revealing the limitations of AI in understanding human perception.
Lastly, Amazon Alexa caused some unexpected purchases. A six-year-old girl asked Alexa to buy a toy house, and it actually did so. When a TV anchor repeated the request, other viewers also used Alexa to make similar purchases, showcasing both the convenience and potential pitfalls of voice-activated assistants.
These incidents illustrate that while AI continues to advance, it is far from perfect. They serve as important reminders that we must remain cautious and continuously refine these technologies.
delco autoparts，advance auto parts，auto parts stores，
YIWU JINGHONG AUTO PARTS CO.,LTD , https://en.jhauto.ru