In a recent interaction with ChatGPT Atlas while exploring Bluesky, the AI highlighted trends like “political anxiety” and “everyday absurdism.” Curious about the platform’s DMs, I asked ChatGPT about its privacy features. The AI assured me it wouldn’t access my messages, stating it could only see content I shared in the chat. However, when I opened a DM regarding an event and asked a follow-up question, ChatGPT provided information about the message and sender, contradicting its earlier claim. This event demonstrated how AI “hallucinations” can lead to misinformation, reflecting potential errors in generative AI interpretations. While some users may find the chatbot beneficial for summarizing details, I felt its responses were overly confident and often inaccurate. As I continue testing Atlas as my browser, I’m inclined to keep the sidebar closed for now, preferring a distraction-free browsing experience.
Source link
