作　　者： ;

机构地区： 中山大学

出　　处： 《上海师范大学学报（哲学社会科学版）》 2022年第4期88-97,共10页

摘　　要： 制造人类水平的机器人是人工智能的最高目标。要实现这个目标首先就得解决如何衡量机器达到人类水平的问题。其次是路径问题,即什么样的路径最有希望让机器达到人类水平。对于第一个问题,图灵所给出的答案是“图灵测试”。但这个答案并不充分,应该扩充为“全总图灵测试”,要求机器人在真实世界中以一种与人类行动者无法区分的方式做人类行动者所能做之事。对于第二个问题,图灵同样给出了答案,即他所说的“儿童机器”。尽管以图灵的“儿童机器”概念为基础的发展型机器人学是建构人类水平机器人的有效路径,但发展型机器人要发展到人类水平机器人还需解决整合问题,即如何将机器中的不同认知构架整合起来,以实现流畅的信息交流。而要解决整合问题,就要赋予发展型机器人与人脑相当的认知模块和意识水平。 An ambitious goal of artificial intelligence is to build human-level robots. To achieve this goal,two problems need to be cleared up. The first one is how to assess whether the robots have reached human level. The second one is the approach problem,that is,which approach is the most likely one to make robots reach human level. As to the first problem,Alan Turing has proposed a famous solution,that is,the Turing Test. This paper holds that the Turing Test is not sufficient and should be strengthened as the Total Turing Test,which demands the robot must be able to do everything that real people can do,in a way that is indistinguishable to a person from the way real people do it. In terms of the second problem,Turing has also proposed a solution,that is,“child machines”as he calls. Although developmental robotics based on Turing’s concept of“children’s machines”is a feasible approach to constructing human-level robots,the growth of developmental robots to human-level robots still needs to solve the integration problem,that is,how to integrate different cognitive architectures in machines to achieve smooth information exchange. To solve the integration problem,it is necessary to endow developmental robots with cognitive modules and levels of consciousness equivalent to those of the human brain.

关 键 词： 人类水平机器人 全总图灵测试 布洛克脑 中文屋 发展型机器人 儿童机器 认知架构整合

领　　域： [自动化与计算机技术—控制理论与控制工程] [自动化与计算机技术—控制科学与工程]