修辞对人工智能+的挑战:以人机互动为例

Rhetoric Challenging AI with Special Reference to Human-Machine Interaction

修辞是人类语言生活中普遍存在的现象。让计算机自动处理此普遍现象是当前人工智能+所面临的最大挑战之一。文章以人机互动为例讨论此挑战的复杂性。人机互动从图灵测试始,迄今70年呈现4种基本发展脉络:机器无智能、出现智能、开始智能满足人的需求以及试图琢磨人的心灵与情感。从人的角度,与机器智能对应的是人对机器的依赖程度,也分为四种:人主机佣型、人主机助型、人需机供型以及人依赖机型。文章从机器和人的视角分别讨论了修辞现象的形式化问题。迄今,词法、语音、音系、句法、语义等形式化程度高,计算机自动处理的能力较好。修辞/语用相比之下,计算机自动化处理还处于起步阶段。修辞学、计算机科学、人工智能等需要跨学科联合攻关。文章最后还讨论了人机互动的伦理问题,提出认知性人机互动依赖需要立法。

Rhetoric as practice permeates every instance of language use. This universal phenomenon has largely been ignored by AI due to the fact that it is extremely difficult to compute. The paper attempts to pinpoint the challenges rhetoric poses to AI,using human-machine interaction( HM I) as a case study. It starts w ith Turing Test,i. e.,HM I via voice as design,but HM I via w riting in practice,and discusses four phases of HM I development,w hich have progressed from no intelligence to some higher degree of intelligence adopting the computer’s capabilities of processing human language as the hallmark. Various w ays of formalizing phonetics/phonology,syntax and semantics,having contributed to this progress,are review ed,meanw hile the rhetoric/pragmatics phenomena have proven to resist formalization. The paper concludes w ith a brief critical review of recent attempts at automatic essay w riting,poetic composition,AI-supported coursew are, etc. AI ethics is called for.