Building AI That Gets Us: A Simple Guide to Aligned Intelligence

AI is getting smarter, but how can we make sure it's on our team? The key might be teaching AI to understand the world and what we want, just like we do. This isn't about making AI smarter in general, but smarter in a way that aligns with our goals. Think of it like teaching a kid. You don't just want them to be smart, you want them to understand right from wrong and what makes people happy or sad. The same goes for AI. We need to create AI that can learn about the world and figure out how other people (and AI) see things. This is called structure learning, and it's a big deal. So, how do we do this? First, we need to give AI some basic knowledge, like how the world works. Then, we need to help it figure out what's important and what's not. This is like teaching it to focus on the big picture, not just tiny details. We also need to teach AI to think about how other people think, a concept known as theory of mind. But how do we make sure AI follows the rules? One idea is to use something like Asimov's laws of robotics, but make them more precise. These laws tell AI to be careful and not hurt people. But we can do better. We can refine these rules to make sure AI is truly aligned with our values. This isn't just about making AI smarter. It's about making AI that understands us and works with us. By teaching AI to learn about the world and other minds, we can create AI that's truly aligned with our goals. But it's not easy. It takes a lot of work and a lot of smart ideas from different fields, like math, statistics, and cognitive science.