基礎知識指南

什麼是 AI?

Artificial intelligence (AI) is the field of building machines that take in information and produce outputs such as predictions, recommendations, content or decisions that would otherwise need human judgment.

聆聽此指南AI 語音敘述 · ~9 min

本頁閱讀時間 10 分鐘
  1. 概述
  2. 重點摘要
  3. 深入探討
  4. Decide whether a product is really AI
  5. 現實世界的實施
  6. The Future of What Is AI?
  7. 付諸實踐
  8. 資料來源與延伸閱讀
  9. 不斷探索
  10. 常見問題

概述

It is an umbrella term: rule-based programs, machine learning, deep learning and today's chatbots all sit under it, and each has different strengths, costs and failure modes worth knowing before trusting one.

重點摘要

  1. AI is an umbrella term for machines that infer outputs such as predictions, content, recommendations or decisions from their inputs.
  2. Rule-based AI follows knowledge people wrote down, while machine learning extracts patterns from examples; most modern products combine learned models with conventional software.
  3. Alan Turing posed the question of machine thinking in 1950, and the 1955 Dartmouth proposal gave the field its name ahead of the 1956 workshop.
  4. Every AI system in use today is narrow, built and tested for particular tasks; artificial general intelligence remains hypothetical with no agreed test.
  5. AI systems can be confidently wrong, sensitive to wording and biased by their training data, so outputs need checking where mistakes matter.
  6. Running AI has become much cheaper, with GPT-3.5-level inference costs down more than 280-fold from 2022 to 2024 per the Stanford AI Index.

深入探討

Artificial intelligence is the effort to make machines do things that would require judgment if a person did them: recognize a face, translate a sentence, recommend a film, drive a car or answer a question. Because the phrase covers so much, it helps to use a precise definition. The European Union's AI Act, whose definitions have applied since 2 February 2025, describes an AI system as a machine-based system that operates with some autonomy, may adapt after deployment, and infers from its input how to generate outputs such as predictions, content, recommendations or decisions that can affect physical or virtual environments. The key word is infers. A calculator follows fixed steps; an AI system works out an output that nobody spelled out in advance for that exact input.

The field has two broad families of methods, and most real products mix them. The first is symbolic AI, sometimes called rule-based or classical AI. People write down knowledge as rules and the program searches through possibilities. IBM's Deep Blue is the textbook case: it evaluated about 200 million chess positions per second using 32 processors and a hand-tuned scoring function. The second family is machine learning, in which the program is not told the rules but learns them from examples. A spam filter trained on millions of labeled emails, or a model that predicts delivery times from past trips, works this way. Deep learning is the branch of machine learning that uses neural networks with many layers, and since the early 2010s it has driven most of the visible progress in speech, vision and language, as the 2015 Nature review by Yann LeCun, Yoshua Bengio and Geoffrey Hinton describes.

A modern machine learning system is built in a fairly fixed sequence. Someone defines the task and collects data, such as photos with labels or text from the web. They choose a model, a mathematical function with adjustable settings called parameters. Training repeatedly adjusts those parameters so the model's outputs match the examples better, as judged by a score called the objective or loss. The trained model is then tested on data it has never seen, because doing well on the training examples alone proves little. Finally it is deployed inside a product, where it receives real inputs, and its behavior is monitored because the world drifts away from the training data over time.

The history runs back more than seventy years. In 1950 Alan Turing opened his paper Computing Machinery and Intelligence with the question of whether machines can think, and proposed replacing it with a test, the imitation game, in which an interrogator chats with a hidden human and a hidden machine. He guessed that within about fifty years a machine could fool an average interrogator often enough that they would have no better than a 70 percent chance of a correct identification after five minutes. On August 31, 1955, John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon dated their proposal for a summer research project at Dartmouth College on artificial intelligence, and the 1956 workshop that followed gave the field its name. Later milestones include Deep Blue's 1997 chess win, the 2012 to 2015 deep learning breakthroughs, AlphaGo's 2016 Go victories, and ChatGPT's public release on November 30, 2022.

People often sort AI by how broad it is. Narrow AI does one kind of task, such as playing chess or transcribing speech, and every system in commercial use today is built and tested for particular tasks. Artificial general intelligence, or AGI, refers to a hypothetical system that could learn and perform most intellectual tasks a person can; there is no agreed test for it and no agreed date for it. Generative AI is a newer label for systems that produce text, images, audio or code. It is not a separate kind of intelligence but an application of deep learning. Robotics adds sensors and motors so the outputs act on the physical world.

It also helps to compare AI with things it is often confused with. Ordinary software follows explicit instructions and gives the same answer every time; AI systems can handle messy inputs but can also be wrong in ways that are hard to predict. Automation means any machine doing a task without a person, and much automation, such as a timer on a sprinkler, contains no AI at all. Human intelligence is general, grounded in physical experience and able to learn from a handful of examples; most AI systems need large amounts of data and can fail when inputs differ from what they saw in training. AlphaGo could beat a professional at Go but could not play chess without being rebuilt and retrained.

The limits are practical, not abstract. OpenAI's own launch note for ChatGPT warned that it sometimes writes plausible-sounding but incorrect answers, changes its answer when a question is rephrased, guesses at ambiguous requests instead of asking, and can still respond to harmful instructions or show biased behavior. Learned systems absorb the patterns, and the gaps, in their training data, so they can treat groups of people unequally. They can be fooled by inputs designed to trick them. And an AI system has no built-in sense of when it is outside its competence, so the people deploying it must decide where it is safe to rely on.

Using AI is cheap for individuals and expensive to build at the frontier. The Stanford AI Index 2025 reports that the cost of running a model at GPT-3.5's level fell more than 280-fold between November 2022 and October 2024, which is why capable assistants are now free or low cost. At the same time, US private investment in AI reached 109.1 billion dollars in 2024, and nearly 90% of notable new models came from companies rather than universities, because training the largest systems takes specialized chips, large data centers and large datasets. For most organizations the realistic cost is not building a model but integrating an existing one, checking its outputs, and training people to use it well.

04工作範例

Decide whether a product is really AI

  1. Pick three everyday products: a sprinkler that turns on at 6 a.m., an email spam filter, and a chatbot on a bank's website.

  2. For each one, ask what its output is. The sprinkler outputs on or off, the filter outputs spam or not spam, the chatbot outputs a written reply.

  3. Ask whether a person wrote the exact rule for every case. The sprinkler's schedule was written by a person; nobody wrote a rule for every possible email or question.

  4. Ask whether the system infers its output from the input it receives, the test used in the EU AI Act definition. The filter and chatbot do; the timer does not.

  5. Ask whether it learned from data. A modern spam filter and a chatbot are trained on examples, which makes them machine learning systems.

  6. Write down for each AI product what could go wrong: a real email sent to spam, or a confident but wrong answer about fees.

它顯示了什麼

The spam filter and the chatbot count as AI and the timer does not. The check is whether the system works out its output from the input rather than following a fixed rule, and the list of possible errors shows where a person still needs to verify the output.

現實世界的實施

Game playing by search: IBM's Deep Blue examined about 200 million chess positions per second on 32 parallel processors and beat world champion Garry Kasparov 3.5 to 2.5 in May 1997, a year after losing to him 4 to 2.

Learning plus search: DeepMind's AlphaGo combined neural networks trained on expert games and self-play with tree search, and in 2016 became the first program to beat a human professional at full-board Go, winning 5 to 0.

Conversation: ChatGPT, released by OpenAI on November 30, 2022, is a GPT-3.5 series language model fine-tuned with reinforcement learning from human feedback so it can answer follow-up questions in a dialogue.

Medicine and transport: the Stanford AI Index 2025 counts 223 AI-enabled medical devices approved by the US FDA in 2023, up from six in 2015, and reports Waymo providing more than 150,000 driverless rides each week.

Everyday business software: the same report finds that 78% of surveyed organizations used AI in 2024, up from 55% the year before, which shows how quickly AI moved from pilots into routine work software.

The Future of What Is AI?

Several questions about AI remain open in 2026. The first is measurement: Turing's 1950 imitation game is still discussed, but researchers now rely on large benchmark suites, and the Stanford AI Index 2025 reports that scores on demanding benchmarks rose within a single year by 48.9 percentage points on GPQA and 67.3 on SWE-bench. Fast saturation makes it hard to say how capable systems really are outside tests. The second is governance. The EU AI Act's definitions began applying on 2 February 2025, and the AI Index counted 59 AI-related regulations from US federal agencies in 2024, more than double the year before. How these rules work in practice, and whether other countries converge on similar definitions, is still being settled. The third is reliability. Hallucinated facts, sensitivity to wording and biased outputs, which OpenAI listed as known limitations at ChatGPT's 2022 launch, are active research areas rather than solved problems. Whether more data and computing power will keep producing broader abilities, or whether new methods are needed for systems to reason dependably, is debated among researchers rather than settled.

付諸實踐

為什麼這很重要

  • 更明確的決策. 它可以幫助您將清晰的技術聲明與行銷語言分開。
  • 成本與預算. 在花費金錢或時間之前,您可以提出更好的實施問題。
  • 團隊與工作流程. 具有共同理解的團隊可以做出更好的產品、政策和學習決策。

留意

  • 不同的團隊可能會以不同的方式使用相同術語,因此請儘早定義範圍。
  • 基準測試可能看起來很強大,但實際效能卻參差不齊。
  • 忽視數據品質和評估計劃通常會產生脆弱的結果。

後續步驟

  1. 從您需要的結果的簡單語言定義開始。
  2. 在測試之前選擇一種成功指標和一種失敗條件。
  3. 使用代表性資料運行小型試點,而不是完善的演示集。
  4. Document where What Is AI? helps and where simpler methods are better.

資料來源與延伸閱讀

  1. Mind (A. M. Turing, 1950), course reprint at UMBCComputing Machinery and Intelligence
  2. AI Magazine (AAAI)A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence, August 31, 1955
  3. IBM深藍
  4. Nature (Silver et al., 2016)Mastering the game of Go with deep neural networks and tree search
  5. Nature (LeCun, Bengio and Hinton, 2015)深度學習
  6. OpenAIIntroducing ChatGPT
  7. Future of Life Institute (EU AI Act explorer)Article 3: Definitions, EU Artificial Intelligence Act
  8. Stanford Institute for Human-Centered AI2025年人工智慧指數報告

不斷探索

免費電子報

獲取每日 AI 簡報

每個工作日上午提供三則經過驗證的 AI 新聞,以淺顯易懂的英文撰寫。永久免費,無廣告。

每個工作日一封電子郵件。一鍵取消訂閱。我們絕不出售或分享您的地址。

測試一下自己

Take the What Is AI? quiz

每個答案皆有即時回饋,通過課程後可獲得具備可驗證 ID 的可分享證書。

開始測驗

支持免費的 AI 教育。AI Understanding 是一個 501(c)(3) 非營利組織——絕無廣告,絕無付費牆。 捐款

常見問題

什麼是 AI?

Artificial intelligence (AI) is the field of building machines that take in information and produce outputs such as predictions, recommendations, content or decisions that would otherwise need human judgment. It is an umbrella term: rule-based programs, machine learning, deep learning and today's chatbots all sit under it, and each has different strengths, costs and failure modes worth knowing before trusting one.

路線規劃者使用明確的規則來搜尋可能的旅程。這說明了人工智慧的什麼特色?

該指南包括人工智慧內的明確規則和搜尋;機器學習是更廣泛領域內的一種方法。

機器學習與人工智慧的關係如何最好地描述?

機器學習是更廣泛的人工智慧領域中的一種方法。開發人員不再手寫每條規則,而是讓系統從範例中學習。

A delivery service estimates arrival times from route and traffic data. What should a user understand about that output?

The guide’s delivery example is an estimate based on route and traffic data, not a guaranteed arrival time.

A team is choosing between two AI tools for an ambiguous support task. Which comparison follows the guide?

The guide recommends evaluating task performance on unfamiliar and ambiguous cases, with attention to consequential errors.

A photo organizer labels familiar pictures well. Which next check addresses a limit described in the guide?

Performance on familiar examples does not establish reliability on unfamiliar inputs or costly mistakes.