Large-scale image recognition faced a fundamental bottleneck before 2012. Researchers Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton