人工智能的局限性(二) | 经济学人精讲第673期

文章导读

本文选自《经济学人》6月13日刊文章。人工智能在最近几年大火,但本期的《经济学人》杂志却从各个方面讲述了人工智能的局限性。人工智能的发展并非一帆风顺,它70多年的发展历史交织着“寒冬期”和“火热期”,随着最近几年人工智能的突飞猛进,人们越来越意识到它的局限性,它的不可靠、强化偏见、传播错误信息、不安全,让人类对它既充满期待,又包含警惕。

选文精讲

The future 未来
Humans will add to AI’s limitations
人类将增加人工智能的局限性
It will slow progress even more, but another AI winter is unlikely
它将进一步减缓进程,但另一个人工智能冬天不太可能
Jun 13th 2020 |
A sure way to fail is to set expectations too high. In 1969 a widely read book by Marvin Minsky and Seymour Papert, two AI researchers, described how, in contrast to the grand promises, the Perceptron was constitutionally incapable of performing certain elementary tasks. Research into neural networks languished for a decade.
  • incapable of: 没有能力的
  • languish: 停滞、未能取得进展
一个肯定会失败的方法是设定过高的期望。1969年,两位人工智能研究人员马文·明斯基和西莫尔·佩普特写了一本广为阅读的书,书中描述了感知器在本质上是如何无法执行某些基本任务的,这与人们的宏大承诺形成了鲜明对比。对神经网络的研究停滞了十年。
The history of AI is of periodic bouts of overexcitementinterspersed with “AI winters”, in which limits become apparent, enthusiasmdrains away and funding is slashed. In 1973 the British government abandoned nearly all AI research, citing the field’s failure to meet its “grandiose objectives”. After Minsky and Papert’s book, the focus of AI research turned to “symbolic” approaches that focused on formal logic and rigid, top-down reasoning. Once again, early progress caused much excitement before the unmet promises piled high enough to cause another backlash in the 1980s. Funding cuts closed research projects. American officials dismissed efforts at AI as little more than “clever programming”.
  • bouts of: 一阵、几回
  • overexcitement: 过分激动
  • intersperse with: 用…点缀着
  • enthusiasm: 热情
  • drain away: 逐渐枯竭
  • grandiose: 宏大的、巨大的
  • rigid: 严格的
  • unmet: 未满足的
  • little more than: 仅仅是
人工智能的历史是间歇性的过度兴奋,其间点缀着“人工智能冬季”,在冬季,(人工智能)限制变得明显,热情消退,资金大幅削减。1973年,英国政府放弃了几乎所有的人工智能研究,理由是该领域未能实现其“宏伟目标”。在明斯基和佩特的书出版后,人工智能研究的焦点转向了专注于严谨逻辑和严格的自上而下推理的“符号语言”方法。在20世纪80年代,(人工智能)早期的进展再一次引起了人们的兴奋,但随着越来越多的承诺没有兑现,引发了另一场反冲。经费削减导致研究项目关闭,美国官员认为人工智能不过是“聪明的编程”。

感谢阅读

(0)

相关推荐