I really don’t like to talk about AI in public, but I do want to. I’m very excited about what OpenAI is doing, and I don’t just mean in that it’s playing chess in the next few years. In fact, they’re really starting to push the boundaries of what AI can do. I’m just telling you about them now because I want to get excited about them.
I like to call it a “fooled by AI” because I’ve always felt that we humans are not smart enough to do things that are so complex that they don’t involve a bunch of other people making them.
In my opinion that’s why we think AI is a joke. While I think theres a lot of progress we can make by thinking about the human brain in a more intelligent way, how complex we make our creations is still way above the level that most people think they are, and I think thats the biggest problem with AI. We don’t have the imagination or the drive to do things that are really complex.
And that brings us to the third point. Most AI systems are incredibly simple. They just take in input (text, images, or whatever) and do some very simple math. It’s like someone with a calculator in their hand just doing addition and multiplication and addition and multiplication and addition and then doing division and division and division and division and division. Of course that’s boring and doesn’t really improve upon the fact that we are still stuck with our brains.
The reality is that we will always have to improve upon our brain function to keep it functional for the long term. That is why we are called humans. The current trends towards AI in games are not that much different. It’s just that it does not take any time to program a game that can beat people at the game.
AI has been around for a long time. Back in the early 1980’s, we were programming machines that would do the same thing that we do, but we were not quite able to beat them. The best AI we could get was a program that would learn the game from scratch and beat the human player at the end. If you want to see how much better an AI could get, check the AI at the end of Deus Ex: Human Revolution.
I don’t think these AI programs actually did anything for us, but I do feel that at least the AI programs that we are currently scripting do have some sort of quality that could be used to teach us.
The thing is, as I’ve mentioned it’s not a “game” that we’re making. The AI is actually using our own game engine (openai stateoftheart machine ai) to learn. It’s a very simple game where the AI learns only from what we tell it, and then it can learn from our own game and beat us at the end of the game. In the end, that is what we want to do.
The thing is, at least for the AI programs that we are currently scripting, we have a lot of work ahead of us. For example, we want our AI programs to be able to not only mimic the actions of the human player, but to also mimic the actions of the human players AI as they play the game. We also want to implement a lot of what we learned from the game into our AI programs.
As I mentioned before, our AI programs are a lot more intelligent than the human players, and we are more human in many ways. As such, we want to make sure they are not just making up their actions. It is pretty important to us that we give them the capacity to think, plan, and reason. Unfortunately, this isn’t always possible because the human players are also much more intelligent than we are.