人工智能与机器学习:算法基础和哲学观点

Artificial Intelligence and Machine Learning:Algorithmic Foundations and Philosophical Perspectives

图灵将人工智能简化为具有形式逻辑、自动证明和计算能力的符号处理系统。但是人类智能与语言理解有关。文章提出一个可以在自然界和技术上实现的关于智能度的工作定义来替代图灵测试关于机器智能的定义。自然智能在具有不同程度复杂性的神经系统和大脑的自然进化中出现。人工智能是在技术发明中发展起来的,依赖于传统图灵机计算机能力的指数增长。大脑、自动机和机器似乎是完全不同的,但它们在语言识别方面是数学等价的。具有不同复杂程度的自动机和机器的层次结构是可以区分的,因为它们可以通过适当的神经网络识别相同类型的语言。根据这样的工作定义就得到了自然和技术上的智能度。特别值得注意的是模拟神经网络,它能够像人类大脑那样具有自然语言能力,这超出了图灵可计算性。因此,需强调可通过神经形态计算体系结构实现的模拟和数字智能。但是,智能决不简单等同于大脑和计算机。模拟和数字元素也被整合到全球物联网中,以解决不同程度的智能问题。

Since Alan M. Turing,Artificial Intelligence（ AI） was reduced to symbolic AI with formal logic,automatic proving and computing. But,human intelligence has also been associated with language understanding. Instead of the Turing test,I suggest a working definition of intelligence degrees which can be realized in nature and technology. Natural intelligence emerged during the natural evolution of nervous systems and brains with different degrees of complexity. Artificial intelligence was developed in technical inventions depending on an exponential growth of computer power in tradition of the Turing machine. Brains,automata,and machines seem to be completely different,but they are mathematically equivalent with respect to language recognition. A hierarchy of automata and machines with different degrees of complexity can be distinguished. They recognize the same kinds of languages which are recognized by appropriate neural networks（ and with that by corresponding biological brains with this degree of complexity）. According to our working definition,we get degrees of intelligence in nature and technology. It is remarkable that analog neural networks can realize natural languages beyond Turing computability like human brains. Therefore,this paper argues for analog and digital intelligence which can be realized by neuromorphic computational architecture. But intelligence is by no means reduced to single brains and computers. Analog and digital aspects are also integrated in the global Internet of Things（ Io T） to solve problems with different degrees of intelligence.