(…) Shortcut learning essentially means that neural networks love to cheat. At first glance, AI often seems to work excellently — for example, it can recognize whether a picture contains animals, e.g., sheep. Only upon closer inspection, it is discovered that the neural network has cheated and just looked at the background.
(…) We noticed the very same pattern in very different settings: ‘shortcut learning,’ or ‘cheating,’ appeared to be a common characteristic across both artificial and biological intelligence.
Take a shortcut to ‘Exploring the notion of shortcut learning in employees and deep neural networks, by tech guru Chriet here: