论文部分内容阅读
本文是 John McCarthy 在1984年 AAAI 会议上的主题发言。McCarthy 提出需要确立 AI 研究的衡量准则,以推动 AI 基础研究。这些准则主要包括三个方面:普遍性,容忍性(如框架问题)与认识合适性。它们不仅适合于理论 AI 研究,而且也是实验 AI 研究的准则。从更高标准看,目前 AI 许多领域缺乏实质性进展。值得一提的是,McCarthy 批评了图灵测验作为 AI 的科学准则,他指出,衡量 AI 程序是否达到人的智能程度的一般准则是看它能否体现人的常识行为,这正是当前 AI 研究的关键问题。我们相信 McCarthy 的意见对国内 AI 研究者有所裨益。
This article is the keynote address by John McCarthy at the AAAI conference in 1984. McCarthy proposed the need to establish a yardstick for AI research to advance AI basic research. These guidelines mainly include three aspects: universality, tolerance (such as the framework of the problem) and awareness of appropriateness. They are not only suitable for theoretical AI research, but also for experimental AI research. Viewed from a higher standard, there is a lack of substantive progress in many areas of AI. It is worth mentioning that McCarthy criticized the Turing Test as the scientific guideline of AI. He pointed out that the general criterion for judging whether an AI program reaches the level of human intelligence is whether it reflects human’s common sense behavior, which is exactly what the current AI study The key issue. We believe McCarthy’s advice will benefit domestic AI researchers.