The Tech Future Of Target
Ya know, I'm not even gonna try on this one, this has nothing to do with Target.
There are many topics that I would like to report on. There is so much happening in the technological space, its hard to choose. There are a few runner-ups. These include Cloud Computing, Quantum Computing, and Artificial Intelligence. However, the topic I settled on is Neural Networks. The main reason is that they absolutely fascinate me! They are a masterpiece of programming. They are something that I have wanted to get into for a long time. Back when I did game development, I tried creating a game with procedurally generated levels. At this point, I had only been programming for about a year. I had no idea algorithms existed that would allow you to do this. My approach was to build about 10 templates for levels and have them randomly appear when you advanced levels. It sucked.
This is where something like a neural network would have been amazing. Yes, the algorithms would be more efficient and make the levels actually playable. But watching the neural network, generation after generation, building levels that it thought would be best and most playable would be worth wasting time and a lot of it.
Neural networks are like a black box. You set up parameters, give it a goal, and let it go. Yes, it is significantly more complicated than that. But you get the point. The amazing thing is that large-scale networks are a complete mystery. Not even those who develop them have any idea what is going on. They can see little chunks, but the whole is a mystery.
The problem is, if we build a network too advanced, it may or may not decide that the best way to reach its goal, is by removing humans. But meh, we've been around long enough as is.
There are many topics that I would like to report on. There is so much happening in the technological space, its hard to choose. There are a few runner-ups. These include Cloud Computing, Quantum Computing, and Artificial Intelligence. However, the topic I settled on is Neural Networks. The main reason is that they absolutely fascinate me! They are a masterpiece of programming. They are something that I have wanted to get into for a long time. Back when I did game development, I tried creating a game with procedurally generated levels. At this point, I had only been programming for about a year. I had no idea algorithms existed that would allow you to do this. My approach was to build about 10 templates for levels and have them randomly appear when you advanced levels. It sucked.
This is where something like a neural network would have been amazing. Yes, the algorithms would be more efficient and make the levels actually playable. But watching the neural network, generation after generation, building levels that it thought would be best and most playable would be worth wasting time and a lot of it.
Neural networks are like a black box. You set up parameters, give it a goal, and let it go. Yes, it is significantly more complicated than that. But you get the point. The amazing thing is that large-scale networks are a complete mystery. Not even those who develop them have any idea what is going on. They can see little chunks, but the whole is a mystery.
The problem is, if we build a network too advanced, it may or may not decide that the best way to reach its goal, is by removing humans. But meh, we've been around long enough as is.
Comments
Post a Comment