https://turbowarp.org/581257735?fps=60&size=800x800 here will run faster, and will work much better and more satisfyingly overall. Seriously it's way better. ============= HOW IT WORKS ============== This is a fairly simple form of AI consisting of a three-layer neural network. Results will vary every time you reset the project; sometimes they will learn to be much more capable than others, because there is a significant element of luck in this. Each subject is slightly different, as they each compile their own lists of values. The last one standing has their values uploaded to the next generation, who will base their values off of this. They use sensors to detect purple orbs around them, and use their first data set to generate a list of reactions to each sensor input. Then this list is fed into a script that finds the most powerful outputs and puts their locations into another list; these locations are then used as inputs to a second layer of processing that repeats the process above. Then, an awkwardly long script to determine their final movement is used and our subjects adjust their movement based on this. The idea is that over time the faces will get smarter and work out how to avoid the purple balls. Does it work? Yes, it does. Cool! Honestly surprised myself with this one, I guess I'm just smart lol
All mine actually, I straight up sat down and thought "hmmm I will make an AI today" and so I did. New since last iteration: The AI is now three layer; I don't know what this actually does but I think it makes them smarter. Also, they have changing expressions. On turbowarp it now is much more fleshed out.
