Two things happened to me this week. First, the shocking power of agentic AI finally hit me at work. Power of God... Second, I read anthropics warning about recursive self-improvement in WSJ. It mentioned how some people are freaking out about the mere suggestion of restricting open source LLMs. It made me wonder if some of us are clueless about how dark the dark side of the power of God could be.
I'm proposing a very uncomfortable thought experiment. An edge case. But an unfortunately long and sharp edge. I am asking all you people out there to think of the darkest thing you could see yourself doing with an unchained AI, perhaps at the worst moment in your life...
Actually no, I'm not asking that. Let's do this AI style. I want you to imagine the worst version of yourself and then I want you to simulate the worst version of yourself imagining the worst thing they would do at the worst point in their life to their most hated enemy.
If people answer honestly, this thread will get very disturbing. I'd ask the moderators not to take it down. It's an exploration of what's soon to be possible. And a conversation not likely to happen unless somebody explicitly prompts it. Its value to public discourse is one of safety. Generally speaking, our public servants are good people. They aren't inclined to let their mind to go where the worst of us might go with this technology. If nobody ever says out loud, how will we know to protect ourselves as a society?
[link] [comments]