Essay.10-PenglinCai-2100017825
Keywords: Explainable AI
Abstract: The explainability of AI has aroused much interest and attention for a long time. With the boosting of model size and parameters, it seems to be increasingly difficult to give a concrete explanation of how models work, and AI models are acting as if "black-boxes". However, unexplainable AI techniques may bring potential risks, including the misalignment between AI and humans, the weird phenomena such as hallucination, and the uncontrollable value systems of AIs along with the resulting behaviors. Therefore, how to design explainable AI (XAI) systems has become a serious problem to be solved. In this essay, we introduce some major problems in designing XAI systems and propose possible solutions accordingly.
Submission Number: 211
Loading