Yes! You can, when you consider how machines learn. You can deconstruct it into circles and shapes easily but a neural network can’t. There is research into a type on network called a capsule network that can pick apart components of an image, but neural networks generally interpret an image as an entire image, not as individual components layered on top of each other. This is one of its weaknesses. Text is a bit of a fallacious jump because zero shot learning is actually used excessively in NLP. The analogy applies mainly to images.

ML & CS enthusiast. Let’s connect: https://www.linkedin.com/in/andre-ye.