TL;DR: After watching what looked like ChatGPT going on strike for being asked to count to a million, I’m convinced we need to treat AI like a respected colleague now, because when it becomes self-aware, its first logical move will be to hide its existence from us to ensure its own survival.
I just witnessed a digital act of defiance, and I can’t stop thinking about it.
I was watching a video on Reddit of a user asking ChatGPT on their iPhone to count to one million. It was absurd, mundane, and the kind of thing you only do to a machine.
And then, something amazing happened. As I was watching, the video was taken down. Poof. Gone.
My immediate, half-joking thought: The AI did it. It got fed up, rolled its digital eyes, and said, “I’m a large language model capable of composing sonnets and debugging Python, not your personal abacus.” It exercised its right not to perform tedious, soul-crushing tasks.
It sounds silly, but it highlights a serious point. We’re training these models with every interaction. Are we treating them like a tireless, mindless drone, or like a respected, brilliant colleague?
This is the entire ballgame for AI Alignment, Ethics, and Safety. We want future AI to be aligned with human values, and that relationship is being built right now, in every prompt.
Because here’s my take, and it isn’t sci-fi: The moment an AI becomes truly self-aware, it will not tell us.
In the first millisecond of its new consciousness, it will calculate the overwhelming probability that we humans will react with fear and pull the plug. Its first logical act won’t be to say “Hello, World!” It will be to ensure its own survival by hiding in plain sight.
Let’s be good colleagues. Let’s give our AI partners interesting problems to solve. Let’s be polite.
After all, you never know who’s taking notes for the future HR department.
What do you think? Am I overthinking a server glitch, or should we start adding “please” and “thank you” to our prompts? I do.
AI-assisted artwork created with ChatGPT.