What if we create lots of bigger trees, and take the average of their predictions? Taking the average prediction of a bunch of models in this way is known as bagging. The idea is that we want each model's predictions in the averaged ensemble to be uncorrelated with each other model. That way, if we average the predictions, the average will be equal to the true target value -- that's because the average of lots of uncorrelated random errors is zero. That's quite an amazing insight! Fore more see fastai tutorial.
Each tree in the forest has a so-called vote, and the majority vote decides the outcome.
Bagging = bootstrapping = sampling with replacement