Skip to content

📖 Artificial Intelligence: A Very Short Introduction (2018)

💡 About

This is my first attempt to use LLMs for book summarization. The book was chosen as I want to dig deeper into the history of AI, and the VSI series is a good place to go through deep knowledge in short time. After the summarization process, I found that it could also be meaningful to use LLMs to translate the summaries!

📝 Prompt Used

I want you to act as a professional text summariser. I will send you some texts and you shall summarize it with minimal technical jargons and make sure adults from a different field would understand. Do not write explanations. Forget everything after you have done the summarization. Do not summarize the text with the memory of the previous prompts or answers. Do not start with "In summary" or "In conclusion". Only return the summarized text and do nothing else.

  • Paragraphs existed the maximum length of prompt are manually segmented; texts that violate the TOS of ChatGPT were manually removed.
  • Depth of Summarization: 3

Corpus

  • The corpus used is Artificial Intelligence: A Very Short Introduction (Very Short Introductions), which can purchased here.

Disclaimer

The raw content of the book is the property of Oxford University Press and shall not be redistributed.

Abstract

Artificial Intelligence (AI) is the use of computers to perform tasks that would normally require human intelligence such as reasoning, perception, prediction, and planning. The ultimate goal of AI research is to create systems with human-like general intelligence, which remains a major challenge in the field. Researchers use a variety of methodologies and techniques such as heuristics, planning, mathematical simplification, and knowledge representation to achieve this goal. The current research focuses on using AI to recognize and respond to human emotions, rather than on creating AI systems that can experience emotions themselves. The backpropagation algorithm is a popular method used to train Artificial Neural Networks (ANNs) which is based on the principle of "backpropagation" of errors and it is found to be effective in training deep neural networks.

人工智能（AI）是使用計算機執行通常需要人類智能的任務，如推理，感知，預測和規劃。 AI研究的最終目標是創建具有人類般通用智能的系統，這仍然是該領域的一個主要挑戰。研究人員使用各種方法和技術，如啟發式，計劃，數學簡化和知識表示來實現這一目標。目前的研究重點是使用AI來識別和回應人類的情感，而不是創建能夠經歷情感的AI系統。反向傳播算法是一種常用於訓練人工神經網絡（ANNs）的方法，基於“錯誤的反向傳播”原理，並且被發現在訓練深度神經網絡中是有效的。

Content

1. What is artificial intelligence? (人工智能是什麼？)

Artificial Intelligence (AI) is the use of computers to perform tasks that would normally require human intelligence such as reasoning, perception, prediction, and planning. It involves the use of virtual machines, which are information-processing systems that the programmer has in mind when writing a program. There are five major types of virtual machines used in AI: classical or symbolic AI, artificial neural networks or connectionism, evolutionary programming, cellular automata, and dynamical systems. AI pioneer Ada Lovelace foresaw part of AI in the 1840s, but how to achieve it was still a mystery. Alan Turing clarified the mystery of AI in the 1940s by showing that every possible computation can in principle be performed by a mathematical system called a universal Turing machine. Early AI research was heavily influenced by the cybernetics movement of the 1940s and 1950s, which focused on biological self-organization and the concept of feedback.

人工智能（AI）是指使用計算機進行需要人類智能的任務，如推理，感知，預測和計劃。它涉及使用虛擬機器，這是程序員在程序寫作時考慮的信息處理系統。人工智能中使用了五種主要類型的虛擬機器：經典或符號性人工智能，人工神經網絡或連接主義，演化程序，元胞自動機和動態系統。人工智能先驅阿達·洛維蕾斯在19世紀40年代預見了人工智能的一部分，但如何實現它仍然是個謎。艾倫·圖靈在20世紀40年代澄清了人工智能的謎，他表明，原理上可以通過一種名為通用圖靈機的數學系統實現每一種可能的計算。早期的人工智能研究受到20世紀40年代和50年代生物自組織和反饋概念的網絡運動的嚴重影響。

2. General intelligence as the Holy Grail (通用智能是聖杯)

General intelligence as the Holy Grail refers to the ultimate goal of AI research, which is to create systems with human-like general intelligence. The field of AI has made significant progress in achieving specialized tasks, but creating systems with general intelligence remains a major challenge and the ultimate goal of the field.

To achieve this goal, AI researchers use a variety of methodologies and techniques such as heuristics, planning, mathematical simplification, and knowledge representation. Heuristics are problem-solving strategies that involve using shortcuts or educated guesses to find a solution. Planning involves breaking down a problem into smaller, more manageable sub-problems and solving them in a specific order. Mathematical simplification involves using mathematical techniques to reduce the complexity of a problem, making it more manageable to solve. Knowledge representation involves organizing information in a way that makes it more easily accessible and usable for problem-solving.

While these strategies can greatly improve the efficiency of solving problems, they also have limitations and assumptions that can limit the scope of the AI system and may not be applicable in all situations. It is important for AI practitioners to be aware of these limitations and to consider their implications when designing and using AI systems.

通用智能被稱為人工智慧研究的終極目標，目的是創造具有人類般通用智能的系統。人工智慧領域已經在特定任務上取得重大進展，但是創造具有通用智能的系統仍是面臨的主要挑戰和領域的終極目標。

為了達到這個目標，人工智慧研究者使用各種方法和技術，如啟發式演算法，規劃，數學簡化和知識表示。啟發式演算法是指使用捷徑或受過教育的猜測來尋找解決方案的問題解決策略。規劃涉及將問題分解為較小、更易於管理的子問題，並按特定順序解決它們。數學簡化涉及使用數學技巧來簡化問題的複雜性，使其更易於解決。知識表示涉及以易於訪問和使用的方式組織資訊，以便解決問題。

雖然這些策略可以大大提高解決問題的效率，但它們也有侷限性和假設，可能會限制人工智慧系統的範圍，並且可能不適用於所有情況。人工智慧從業者應該瞭解這些侷限性並在設計和使用人工智慧系統時考慮它們的影響。

通用智能是一個很大的挑戰，研究者們仍在不斷地研究與發展，希望能夠創造出能夠獨立思考，學習和進行決策的系統。同時，研究者們也需要考慮人工智慧系統對社會和環境的影響，確保他們的研究和開發是可持續和適當的。

3. Language, creativity, emotion (語言，創造力，情感)

The current research focuses on using AI to recognize and respond to human emotions, rather than on creating AI systems that can experience emotions themselves. This is a challenging task as emotions are complex and multi-faceted, and there is still much debate among researchers about how they should be represented in AI systems and how they should be used to influence the behavior of the AI. However, some researchers argue that incorporating emotions into AI systems could lead to more natural and human-like interactions, which would be beneficial for a wide range of applications such as healthcare, education, and customer service. Overall, while progress has been made in the field, much more research is needed to fully understand and model emotions in AI systems.

目前的研究集中在使用人工智慧來識別和應對人類情緒，而不是在創造能夠自己體驗情緒的人工智慧系統。這是一項具有挑戰性的任務，因為情緒是複雜的和多方面的，而且研究人員之間仍有很多爭論，關於他們應該如何在人工智慧系統中表示和如何用來影響人工智慧的行為。然而，一些研究人員認為將情緒納入人工智慧系統可能會導致更自然和更像人類的互動，這對於諸如醫療保健、教育和客戶服務等廣泛的應用是有益的。總的來說，儘管該領域取得了進展，但還需要更多的研究來充分理解和模擬人工智慧系統中的情緒。

4. Artificial neural networks (人工神經網絡)

The backpropagation algorithm is a popular method used to train ANNs. It uses the gradient descent method to adjust the weights of the connections between units in the network, in order to minimize the error between the network's output and the desired output. This algorithm is based on the principle of "backpropagation" of errors, where the error is propagated backwards through the network, starting from the output units, and adjustments are made to the weights of the connections accordingly. This method has been found to be effective in training deep neural networks, which have many layers of interconnected units and are capable of learning complex patterns and features.

Deep learning is a subfield of machine learning that is focused on the development of deep neural networks and the application of these networks to various tasks such as image and speech recognition, natural language processing, and decision making. The main advantage of deep learning is its ability to automatically learn features and representations from large amounts of data, without the need for explicit feature engineering. This has led to breakthroughs in many areas of AI, such as computer vision and natural language processing.

Overall, ANNs and deep learning are important areas of AI research that have led to many practical applications and have also had theoretical implications for the understanding of the brain and intelligence. The field is constantly evolving, with new techniques and architectures being developed to improve the performance of these networks.

反向傳播演算法是一種常用於訓練人工神經網路的方法。它使用梯度下降法來調整網路中單元之間連接的權值，以最小化網路輸出與期望輸出之間的誤差。該演算法基於“誤差反向傳播”的原理，其中誤差從輸出單元開始向後傳播，並相應地調整連接的權值。這種方法在訓練深度神經網路時被發現是有效的，這些網路具有許多互相連接的單元層，能夠學習複雜的模式和特徵。

深度學習是機器學習的一個子領域，專注於深度神經網路的開發和這些網路在各種任務中的應用，如圖像和語音識別，自然語言處理和決策。深度學習的主要優勢在於其能夠從大量資料中自動學習特徵和表示，而無需顯式特徵工程。這在許多人工智慧領域都取得了突破，如電腦視覺和自然語言處理。

總的來說，人工神經網路和深度學習是人工智慧研究的重要領域，已經導致了許多實際應用，也對理解大腦和智能有理論意義。該領域不斷發展，不斷推出新的技術和體系結構來提高這些網路的性能。

5. Robots and artificial life (機器人和人工生命)

Artificial Life (A-Life) is a branch of AI that focuses on modeling and simulating the characteristics and behaviors of living organisms. It combines ideas from computer science, biology, and other fields to create artificial systems that exhibit characteristics of life, such as self-organization, adaptation, and evolution. Robotics is one of the key applications of A-Life in AI, as robots are quintessential examples of AI. Recent advances in robotics have come from a shift from psychology to biology, leading to the development of "situated" or "behavior-based" robots that can navigate rough terrain and adapt to unexpected situations. Evolutionary AI is another subfield of AI that uses techniques from evolutionary biology to generate solutions to problems. Self-organization is a key feature of biological systems and is the spontaneous emergence of order from a less ordered state, and it is an important concept in the field of A-Life and AI in general.

人工生命(A-Life)是人工智慧的一個分支，主要關注對生物的特徵和行為進行建模和模擬。它結合電腦科學，生物學和其他領域的思想，創造出具有生命特徵(如自我組織，適應和進化)的人工系統。機器人是A-Life在人工智慧中的關鍵應用之一，因為機器人是人工智慧的典型例子。機器人的最新進展來自從心理學轉向生物學，導致了“位置”或“行為為基礎”機器人的發展，這些機器人能夠在崎嶇的地形上行走並適應意外情況。進化人工智慧是另一個人工智慧的子領域，使用進化生物學的技術來生成解決問題的方案。自我組織是生物系統的關鍵特徵，即從一個不太有序的狀態自發形成秩序，是A-Life和人工智慧領域中一個重要的概念。

6. But is it intelligence, really? (但這真的是智能嗎？)

The question of whether AGI systems can truly possess intelligence, understanding, creativity, selves, moral standing, free choice, and consciousness is a philosophical one that is highly controversial and has no unchallengeable answers. However, careful arguments and use of AI concepts by philosophers can help to shed light on these questions and provide a more nuanced understanding of the nature of real minds. The Turing Test, proposed by Alan Turing in 1950, is a test to determine whether a machine can be considered truly intelligent by being able to hold a conversation with a human and not be distinguished from a human. However, it has been criticized by philosophers as it does not prove that a machine is truly intelligent or conscious. The problem of consciousness, also known as the "hard problem," refers to the question of how and why we have subjective experiences or qualia. The field of machine consciousness (MC) aims to build computer models of consciousness, focusing on functional consciousness rather than phenomenal consciousness. However, despite some progress, the question of consciousness remains a philosophical morass with many unresolved issues.

AGI系統是否真正具有智能、理解、創造力、自我、道德地位、自由選擇和意識是一個哲學問題，非常有爭議，沒有無可爭議的答案。然而，哲學家們細心的論證和使用AI概念可以幫助闡明這些問題，並提供對真正頭腦性質的更細膩的理解。1956年由艾倫·圖靈提出的圖靈測試是一種測試，旨在確定一台機器是否可以被認為是真正的智能，能夠與人類進行對話，且與人類沒有區別。然而，它被哲學家們批評，因為它並不能證明機器是真正的智能或有意識。意識問題，也稱為“硬問題”，指的是我們如何以及為什麼有主觀經驗或感性的問題。機器意識領域(MC)的目標是建立電腦意識模型，專注於功能意識而不是現象意識。然而，儘管取得了一些進展，意識問題仍然是一個充滿未解決問題的哲學困境。

7. The Singularity (科技奇點)

The concept of whole-brain emulation, also known as mind uploading or mind transfer, is related to the Singularity hypothesis. It refers to the idea of transferring the information in a biological brain to a computer or artificial substrate, resulting in a form of artificial intelligence. The idea is that this could be a way to achieve true AGI, and some proponents argue that it could be a way to achieve the Singularity. However, many experts in the field are skeptical of this idea, as it raises numerous technical, ethical, and philosophical challenges. There is no consensus on whether whole-brain emulation is possible or desirable, and it remains a topic of ongoing debate and research.

為了達到真正的AGI，有人提出整顆腦部模擬的想法，也就是將生物學大腦的資訊轉移到電腦或人工基底上，產生一種人工智能。這個想法是這可能是達成科技奇點假說的方法之一，一些支持者認為這可能是達成科技奇點假說的方法。然而，該領域的許多專家對這個想法持懷疑態度，因為它引起了許多技術，倫理和哲學挑戰。尚未就整顆腦部模擬是否可能或可取達成共識，並且它仍然是一個正在爭論和研究的話題。