In simple words: AlexNet helped demonstrate the power of deep CNNs on complex images.
A larger image task
AlexNet is a CNN that showed strong performance on the ImageNet image classification challenge. It combined deeper layers, a large labelled dataset, GPU training, and Rectified Linear Unit (ReLU) activations.
Avoid memorising
Dropout randomly disables some units during training. This encourages the network not to rely too heavily on a few units and can reduce overfitting. Dropout is normally turned off for inference.
See the flow
large image set + GPU + deeper CNN→category
Simple example
LeNet reads simple digit images; AlexNet was designed for more varied colour photographs and many categories.
Remember: AlexNet helped demonstrate the power of deep CNNs on complex images.