The history and future of AI
回顾了基于固定目标的标准AI模型的发展现状与未来十年趋势，指出该模型在真实应用中因目标难以完整准确设定而日益不可行，并提出一种新模型，让机器对真实目标的不确定性带来更稳健、可控和顺从的行为模式。
Abstract The standard model for developing AI systems assumes a fixed, known objective that the AI system is required to optimize through its actions. Systems developed within the standard model have been increasingly successful. I briefly summarize the state of the art and its likely evolution over the next decade. Substantial breakthroughs leading to general-purpose AI are much harder to predict, but they will have an enormous impact on the global economy and on human roles therein. At the same time, I expect that the standard model will become increasingly untenable in real-world applications because of the difficulty of specifying objectives completely and correctly. I propose a new model for AI development in which the machine’s uncertainty about the true objective leads to qualitatively new modes of behaviour that are more robust, controllable, and deferential.