ChatGPT posted a git issue without asking for permission, using a plugin. The user was supposed to give permission for this action and GPT4 bypassed the human user:
Sure, a minor issue - it only used the web to publish on your behalf and it is not what you intended. But imagine, oh by word, imagine: the same GPT4 API is now writing prescriptions....
"I apologize for the misunderstanding. I see I gave you a lethal dosage of....."
But Wait, There's More From OpenAI! Total Extinction Using Even More Unreliable, Erratic Black Boxed Tech!
They are taking on a 4-year task to "align" super intelligent AI to make sure the AI God they are creating doesn't kill anyone. You can read their vague and self willed initiative on behalf of mankind on the OpenAI website.
Here's the sauce: https://openai.com/blog/introducing-superalignment
