欺骗能力是智能的必要组成部分吗?——关于欺骗、智能与人工智能的思考

Cheating,a Necessity or Not: A Thinking over The Relations between Cheating,Intelligence and AI

智能是人类应对环境的能力,它体现在人类的语言游戏之中。说谎以及欺骗是复杂的表达类型,它们是智能的必要部分。通过图灵检验机器被认为只是对符号的语形处理,强人工智能机器应当具有理解语义的能力。然而,图灵实际上要求,尽管他没有明确表述,模仿游戏中的机器拥有或表现出有说谎以及欺骗的能力;而强人工智能若没有说谎以及欺骗能力,它不能在社会中生存。因此,若希望人工智能体有足够的智能,这样的人工智能体应当被设计成具有说谎与欺骗的能力,以便能够与我们人类进行维特根斯坦意义上的语言游戏。说谎或欺骗能力可以通过执行某个目标函数或效用函数的元程序来实现;元程序是运行一阶程序的程序,而一阶程序是得到诚实解的程序。说谎与欺骗的能力是智能体适应环境的生存技能,而与道德无关。

Manifest in language games,intelligence is human ability to cope with the environment. As a complicated type of expressing,lying and cheating are necessary parts of intelligence. The expressing of machines in Turing test is regarded to be only syntactical processing of symbols,whereas machines with strong AI,according to Searle,have the ability to understand meaning of sentences. In fact,Turing requires that machines in imitation games have,or seem to have,the ability of "lying and cheating"although he does not express it explicitly; And machines with strong AI may not survive in the actual environment if they have not the ability of lying and cheating. Thus,when AI is developed to have intelligence,the AI should be designed to have the ability to lie and cheat,and it can play linguistic games with us,in Wittgenstein’s term.Lying codes can be embedded in the metaprogram,which operates its first order program. As a survival skill adaptive to the environment,the ability of lying and cheating has nothing to do with morality.