Neural networks were conceptualized 50+ years ago. Hinton invented the backpropagation algorithm for it, 30+ years ago. Yet, all these years, no experiment could reach the shores. When everyone had given up, AlexNet fixed the 3 problems that sank the neural networks:

1. Use GPU and ReLU to speed up slow training
2. Use 2 GPUs in parallel to scale up
3. Data augmentation and dropout to solve overfitting

And like Captain Sparrow, it saluted its dead comrades and proudly landed onshore. That was the day, you will always remember as the day, you almost got, Artificial... Intelligence.