人工智能的起源与早期发展

人工智能的萌芽 / Early AI

依据《大英百科全书》的记载，1935年，英国著名逻辑学家与数学家艾伦·麦席森·图灵提出了一种抽象的计算机模型构想：一台具备无限存储空间的机器，配备一个在内存中往复移动的扫描器，能够逐个读取符号并写入新符号。扫描器的动作由内存中存储的一组程序指令来控制。这一数学模型后来被命名为“图灵机”。1950年，图灵研制出了Pilot ACE（自动计算引擎），使其成为世界上首批可编程数字计算机之一。

According to Britannica, in 1935, British logician and mathematician Alan Mathison Turing "described an abstract computing machine consisting of a limitless memory and a scanner that moves back and forth through the memory, symbol by symbol, reading what it finds and writing further symbols. The actions of the scanner are dictated by a program of instructions that also is stored in the memory in the form of symbols." Turing's concept (i.e., mathematical model) eventually became known as the Turing machine. In 1950, Turing completed the Pilot Automatic Computing Engine (ACE), which became one of the first programmable digital computers in the world.

图灵测试 /Turing Tests

1950年，图灵构思了著名的图灵测试（亦称“模仿游戏”），旨在评估人工智能是否展现出智慧并表现出“类人”的行为。测试的关键在于：人类测试者能否分辨书面回答究竟是由人类还是计算机生成的。若要通过测试，人工智能必须具备以下能力：

Alan Turing developed the Turing Test (known as "The Imitation Game") in 1950 to assess if an AI tool exhibits intelligence and acts "humanly." The test boils down to whether a human tester would be able to tell if a person or a computer wrote written responses to a set of questions. To pass the test, a computer would need to exhibit:

知识表示：能够保存已习得的信息以及接收到的资讯；

Knowledge representation, which allows an AI to store what it has learned and what it has heard,

自然语言处理：能够使用英语（或其他语言）与人类顺畅沟通；

Natural language processing, which enables the AI to successfully communicate in English,

自动推理：能够运用既有资讯回答问题并推导出新结论；

Automated reasoning, which allows an AI to use stored information to answer questions and formulate new conclusions, and

机器学习：能够适应新刺激、识别模式并将这些模式应用于未来的情境。

Machine learning, which permits an AI to adapt to new stimuli, detect patterns, and apply those patterns to future situations.

另一位学者Steven Harnad提出了“全面图灵测试”，除了语言能力外，还加入了感知能力（如计算机视觉）和机器人操作能力的测试。机器人能力的测试方式是将物体传送至舱口，若AI能成功操控物体并辨别不同物品，即视为通过测试。

Another test, proposed by Steven Harnad, is the "Total Turing Test." It involves testing a computer's linguistic, perceptual (using computer vision), and robotic abilities. Robotic abilities are tested by passing objects through a hatch. An AI passes the test if it can successfully manipulate objects and discern between different objects presented to it.

人工智能的早期探索 / Early Development of AI

人工智能最基础的形式可以理解为：程序仅对环境做出反应，而不基于过去的记忆或经验来做决策。这类机器会观察环境并据此做出决定，但在与环境的交互方面能力十分有限。

The most basic classification of AI can be thought of as a program only reacting to its environment, without basing decisions on prior memories or prior learning. Machines in this category actively observe their environment to make decisions, yet they are extremely limited in their ability to interact with their environment.

策略博弈与 AI / Strategy Games and AI

国际象棋：1951年图灵发表论文，阐述了能够下棋的算法。1980至1990年代，IBM的Deep Blue与世界冠军加里·卡斯帕罗夫进行对决。在1989年和1996年的交锋中，卡斯帕罗夫均取得胜利。然而在1997年，Deep Blue成为了首个战胜国际象棋特级大师的人工智能。此外，包括AlphaZero和Leela Chess Zero在内的其他AI引擎，也在国际象棋比赛中获胜。如今，众多国际象棋AI引擎采用深度学习技术，通过对过往经验的学习与应用，达到了超越许多顶尖棋手的棋力评级。

In 1951, Alan Turing published a paper describing a program that could play chess. In the 1980s and 1990s, IBM's computer Deep Blue played chess matches against chess world champion Gary Kasparov. In the 1989 and 1996 matches, Kasparov successfully defeated IBM's Deep Blue. But in 1997, IBM's Deep Blue became the first chess AI to win against a grandmaster of chess.Other AI engines, including AlphaZero and Leela Chess Zero, have also won chess matches. Today, many chess AI engines use deep learning, applying learning based on past lessons to achieve chess ratings above many top chess players.

跳棋：1962年，Arthur Samuel开发的跳棋程序击败了当时的冠军Robert Nealy。该程序并非预设所有可能情况，而是引入了机器学习，通过大量对局积累经验并在实战中加以运用。

In 1962, Arthur Samuel's checker-playing computer defeated reigning checkers champion Robert Nealy. Rather than being programmed with potential scenarios, the computer program incorporated machine learning. By playing many games, Samuel's computer acquired knowledge based on past lessons and applied that knowledge throughout the game.

围棋：围棋是源自中国的古老策略游戏，其可能局面数（10¹⁷⁰）远超宇宙中的原子总数（10⁸⁰）。Google研发的AlphaGo在2016年击败了世界冠军李世石，展示了AI在超复杂决策领域的重大突破。

The board game Go is an ancient strategy game invented in China. The possible combinations available in the game of Go (10170) far exceed the number of atoms in the universe (1080), contributing to Go's sophistication and difficulty. The computer AlphaGo was designed by Google to anticipate all possibilities. In 2016, AlphaGo defeated world champion Lee Sedol at Go.

思考题 / To Consider

回想一下你多次经历过的某种情境：

第一次遇到时，你是如何处理的？

再次遇到时，你的反应有何变化？

你获得了哪些经验教训？

如果再次遇到类似情况，你会采取什么不同的做法？

Think of situation that you encountered several times.

How did you respond the first time you encountered the situation?

How did you respond when you encountered that situation again?

What are some of the lessons you learned?

How might you handle the situation differently if you encountered it again?