Neural Network Risks Suggested by Optical Illusions

Abstract

Although there are various theories about the principle of “optical illusion”, many can be explained relatively simply by assuming laws of perspective as a basic theory. Considering the perspective as a brain function cultivated through experience of estimating distance to the target, AI can also see the illusion through learning. Therefore, it is important for safety measures to improve the social environment to eliminate risk factors that are misidentified by AI. Developing the cognitive environment to reduce misperception is not just an issue for AI. Similar problems occur in the elderly. Recognizing that there are signs, fonts, and graphics that the elderly misidentifies, it is time to need the social investment to realize a cognitive barrier-free society that is friendly to the elderly and to AI. It is an urgent issue to build a cognitive barrier-free society for the elderly, just as we realized a barrier-free society for physical ability.