Ben Smith的动态

OpenAI's recent turmoil around CEO Sam Altman may have been triggered by a potential breakthrough called Q*, or Q star. No one really knows what it is, but that hasn't prevented speculation. Some of that speculation is that Q* is a step on the path to Artificial General Intelligence (AGI) and that it has the ability to independently reason about grade school level math. This doesn't sound all that impressive on its surface unless you understand a bit about how computer programs work. Until now, computer programs have only been able to accomplish what we've explicitly told them to do. If anything is even slightly off, the system breaks because it is unable to respond to unexpected situations. Generative AI systems like ChatGPT operate like a fancy autocomplete. Given a large selection of data, they can achieve a basic degree of understanding and perform tasks that have long been considered impossible. They can respond to questions you ask and are generally correct, but they cannot come up with ideas on their own or decide how to perform a task. Q* sounds different. We know very little, but it sounds like OpenAI may have succeeded in teaching an AI model how to reason about some very basic mathematical concepts. This would mean we would have a computer system which, for the first time in history, can think and reason. If that is possible, then autonomous agents (computer systems that can respond to events without direction by a human) are on the horizon. These agents could do literally anything you can imagine because they would be able to act independently once they have been given a directive to accomplish. Achieving AGI comes with a host of ethical issues which must be sorted out. Is this type of artificial intelligence system a conscious being? Should it have rights? What happens to all the humans who will lose their jobs because of the rise of artificial intelligence? No one really knows, and it's going to be really interesting to see how this plays out.