theunitofcaring:

typicalacademic:

[Link to paper]

“Likewise, on CIFAR-10, 49.7% of the convolutional network’s false positives were
classified as frogs, and none were classified as airplanes, automobiles, horses, ships, or trucks.”