Accurate, although in most cases, the art is not in choosing the algorithm (often a regular feed-forward, covolutional or recurrent neural network) but the hype

Accurate, although in most cases, the art is not in choosing the algorithm (often a regular feed-forward, covolutional or recurrent neural network) but the hyperparameters (regularization, number of layers, units, etc.) and sometimes the way the data is fed (augment the dataset with distortions? sample negative examples randomly? etc.)