An AI Protocol

All the recent talk about the dangers of AI makes it clear, we need a protocol for when to pull the plug. It may be a simple idea and common sense, but the reality shows us how difficult this idea would be to implement. First there is the greed to make money out of AI. Anthropic began as a nonprofit, but then that did not seem rational to its owners, and even less so now. That is true for Open AI, and for all the others. Then there is competition, not only among each other but also with China, as if to compete and be number one even if it kills us all in the process would be acceptable. I understand the benefits of AI. Every day I enjoy them when I use Google and its Gemini engine gives me quick answers and allows me to pose questions it might have taken hours of research for me to formulate. And that is a very small one. Yet endeavors such as these must balance costs versus benefits. According to whistle blowers and AI experts willing to be open about what they see, the costs may well end up too high. While our administration does not believe in regulating profits, the dangers grow, and if those who sound the alarm are right, they grow exponentially. Among computer engineers there has long been something they call Pdoom. P is for probability, the probability that something would happen to doom us. As I understand it, there is no program where Pdoom is 0. I deduce therefore that it too would grow exponentially. So may I repeat, we need a protocol for when to pull the AI plug.