Alex Krizhevsky et al. significantly improved previous results on the ImageNet challenge by parallelizing the training of their neural network on multiple GPUs. Their approach utilized a deep convolutional neural network (CNN) architecture, later known as AlexNet, which effectively leveraged the computational power of GPUs to handle large-scale image data and complex computations. This parallelization enabled faster training times and enhanced the model's ability to learn intricate patterns in the data, resulting in a groundbreaking performance improvement in image classification tasks. Their work demonstrated the potential of deep learning and GPU acceleration in advancing computer vision.