In my opinion p1 is true.
If you learn more about how neural networks work in AI you'll find that the method to construct a machine-learning algorithm takes inspiration from the idea of neurons firing.
That being said, the concept of "experience" is an entirely different issue.
I can make a simple, one-way web of artificial neurons (Multi-layer perceptron) and teach them to recognize letters in an image, but I can't I can't ask it to draw those letters. It only recognizes but it doesn't understand.
You can make a multi-way web of neurons all firing at each other to remember certain patterns in images (Hopfield neural networks), but you can't "teach" it, only make it remember.
You can make have a self-teaching one-way web of artificial neurons that starts off doing something random, but who's success given a certain criteria (kills, score, items obtained) serves as a validation method to slowly change its neuron configuration to whatever maximizes that score. You can even teach it to draw something.
But none of these things have anything you could consider a conscious. You could try to combine them to mimic a human, which has already been done, but we know from the exact values of every neuron exactly what they'll do given any situation. Their responses are all mathematically calculated, and it's impossible to tell if they're really just experiencing something or if its just a machine following the precise instructions we gave it.