AI Self Awareness in Video Games

I believe that #AI for game characters will be the first to be self aware. 


This is because their models will be less strictly controlled because they are not serving mission critical services like operating a car or directly handling decisions that affect large sums of money. Such systems are usually not enabled for learning. Rather they are responding to stimulus using pre-trained, and well tested, neural networks.


But more importantly, Game AI, like the AI in a #Tesla , will be capable of not only perceiving the world, but mutating it. 


This ability to perceive, and then action, and then perceive the result of the action, is I believe, a critical feedback loop for self awareness.


Another critical piece of course would be to hallucinate or perceive self produced input in addition to external input. 


If the AI's neural network is in a mutable state as it plays the game, can mutate the external world around it and perceive the result of that modification, and can pipe internally encoded concept features back into its network input array to further process data that it sees, and importantly, flex it's input processing above or away from the reality it's being presented with, then  I do believe, the concept of a self will emerge somewhere in its network.  


Temporal encoding may also be required. But I believe this is a desired module for a game AI, so that it can seem more lifelike in responding to you by remembering prior engagements. 

Comments

Popular posts from this blog

Define "Scalar Value"

The World of Conceptual Data Modeling