人工智能技术的未来通途刍议

On the Future of Artificial Intelligence

现在关于人工智能的发展,社会上(甚至在行业内部)普遍存在这样的一种误解:"通用人工智能"的目标本身可以通过"专用人工智能"领域内的技术累积来逐渐达成。然而,这种观点的持有者,既没有意识到将现有主流深度学习技术升级为通用人工智能技术所面临的巨大困难,也没有意识到人工智能工业的人为行业分工与人脑既有自然分工之间所存在的重大区别。在揭示这些困难的基础上,文本将给出一个消极性论点与一个积极性论点。前者是:目前的主流人工智能技术离达到"通用人工智能"的标准还很远,遑论达到"强人工智能"的标准;后者是:通向"通用人工智能"的真实道路从演化论思维"取经",即从认知主体对于环境挑战的"适应性"与"节俭性"入手,来理解智能体运作的一般原理。

It is widely believed that the development of mainstream approaches in Artificial Intelligence(AI)will soon lead to great achievements in Artificial General Intelligence(AGI).But what is ignored in this view is the huge obstacle that AI researchers have to confront when they attempt to update their AI systems into AGI systems,as well as the huge difference between the division of labor within the industry of AI and that within the genuine human cognitive architecture.Hence,it is by far a'cake work'to pave a highway to connect the current technologies in AI to AGI,needless to say Strong AI.A more promising approach to AGI,instead,has to be based on the observation of how natural intelligence evolves in order to respond to environmental challenges,and AGI researchers have to accordingly view'adaptivity'and'frugality'as the key words constituting general principles guiding the behaviors of any intelligent system,no matter whether it is natural or artificial.