So basically, I copy the main AI brain and made 20 of them that are slightly different. And then when you tell whether the AI is right or wrong, you are actually judging the first of the 20 brain's work.

If correct, it will kill delete those that are not the same with the first Brain's answer because they are all retarded like me.

If wrong, then any Brain with the same answer will die be deleted. Obviously, there is a flaw in this bit beause other brains might also get it wrong, but I on't have enough UI space to put in the "So what's the answer" question, plus, I'd be really lazy as the user to always press one key.

After le brains are murdered peacefully dispatched, the first surviving brain will be saved as the main AI and the brains will be regenerated from the surviving brains to fill up the 20 brains as before.

This means that the SAVED AI Brain ironically does nothing, instead, it's job is to exist as the foundation of the first generation and become the model of the next generations when the AI Brain gets SAVED. I could and probably should optimize this because each brain has like 210 neurons and over dozens of thousands of connections and deleting this useless idling brain would help a lot with memory space

Why talk about training rather than the inner workings of the AI?

If you really want to, watch YouTube, they have visuals and I only have anger.

I rather spend my time explain the training because, unlike walking or jumping robots, there are rarely any articles or meta questions on robots trained by humans. And because it is on a one on one training with the user, it also means that I can't have a fitness level, there is no "You ran faster than average" or "My cat can program a better algorithm". There is only YES or NO, plus you can't automatically know Brain is doing well. There used to be other problems like having a limited imagination but it seems a little off after knowing the bad solution.

Sadly this project is closed

I have several excuses reasons why:

• I didn't go to any "AI theory" class or some sort of guidline like Khan Academy, so this AI is either somehow impossible/incredibly hard to train
• I am unmotivated by my obstacles and I met one I somehow immediately gave up on unlike my other obstacles
• I think this is too hard for me, I should work on something easier, like more games and testing which are more fun
• I have other ideas like:
  • A story/game
  • Particle simulator or something
  • Playing around the canvas tag and the JavaScript functions
  • Learn react.js because yes
  • Try to remake the match game (Although I am trying to be more independent from p5.js)

But I might come back to this project, probably in a form of a new file, but it was sorta fun while I was working on this, similarly how I had two attempts on the same idea for a square game and crushed my problems in the second attempt. Same thing again for an old textbox project I had in Scratch.

(12 of Sept, 2022) Also my StackOverflow account is being warned for bad questions, I'm sad and confused.