On Tuesday night, Snapchat’s new My AI feature experienced a brief but disturbing glitch that is raising larger questions about the reliability and safety of AI chatbots.
Around 9:30pm on August 15th, My AI appeared to gain a mind of its own when it posted a cryptic one-second video to its Snapchat story showing a wall and ceiling. The video was quickly deleted, but not before users captured screenshots and sounded alarms on social media.
During this incident, My AI also ignored or gave strange responses to users who tried chatting with it, typing out replies only to delete them before sending.
Snapchat confirmed the next day that the video and chat oddities were caused by a temporary outage that disrupted My AI. The company stated that the AI was not actually acting on its own or purposefully ignoring users.
However, this glitch follows previous controversies surrounding My AI. In February, underage users reported receiving inappropriate advice about sex and weed from the bot, prompting Snap to implement stricter filters.
While Snapchat insists My AI’s latest aberrant behavior was an anomaly, AI experts and ethicists are expressing renewed concerns about chatbot technology, which remains imperfect.
“When you have a complex AI system like this, glitches are inevitable,” said Dr. Rebecca Green, professor of computer science at Georgetown University. “The question is how the company handles them and whether adequate precautions are in place to prevent harm.”
Chatbots like My AI that are powered by natural language processing can mimic human conversation remarkably well most of the time. But they don’t have human judgment and can make unpredictable mistakes, especially in new situations.
This leaves room for potentially dangerous errors. If an AI chatbot gave unsafe advice to vulnerable teens, for example, the consequences could be tragic.
Experts say chatbot creators have a responsibility to implement rigorous testing and safeguards, and be transparent with users about capabilities and limitations. Parental controls and human content moderation are also critical.
While AI chatbots create exciting possibilities for education, convenience and entertainment, Snapchat’s glitch demonstrates they remain a work in progress. As these bots become more prevalent, we must continue addressing the complex ethical challenges involved in deploying powerful but imperfect AI. With vigilance and care, we can harness the benefits of this emerging technology while minimizing risks.
