Artificial Intelligence Redefined

人藝智能：重新定義AI的本質與未來

錢 宏 Archer Hong Qian

“啊，人類，只有你才有藝術。”（席勒）

AI的未來不僅依賴於技術創新，更需要一種新的哲學視角來引導其發展。Amorsophia為我們提供了這一視角，它融合了智慧與愛的結合理念，強調AI必須在情感、倫理和智慧的平衡中發展。通過心（孞）聯網（MindsWeb 或 MindsNetworking）的技術實現，AI將突破能耗、系統思維和情感倫理的瓶頸，進入一個真正與人類交互主體共生的時代。

為了實現這一願景，全球亟需召集一次“新達特茅斯人工智慧模因會議”（NEW Dartmouth Research Project on AI-Amorsophia），集思廣益制定出與AI哲學相匹配的技術框架和倫理標準。通過這一會議，制定出《AI-Amorsophia憲章》，並將其作為超主權、超地緣、超文明的《全球共生公約》的重要組成部分，引領人類進入一個“讓任何統治全世界的帝國政治企圖成為不可能”的共生新時代。 最終，通過“愛之智慧”，我們將不再僅僅面對AI作為工具或威脅的現實，而是打開一扇通往共生未來的大門，實現人類與AI真正的共生與繁榮。

——摘自《Amorsophia-MindsWeb：突破AI三大“瓶頸”之道》http://symbiosism.com.cn/8878.html

1956年，達特茅斯會議首次提出並確立了“Artificial Intelligence”這一概念與學科術語。這一選擇是基於“Artificial”一詞能夠涵蓋非機械式的智能表現，並突顯其在虛擬性和創造性方面的潛力。相比之下，其他候選術語如“Simulated Intelligence”或“Anthropomorphic Intelligence”更多局限於模仿或擬人化的層面，而無法完全體現AI的技術廣度與哲學深度。

“Artificial Intelligence”強調了人工構建的智能系統如何超越簡單仿真，成為具備創造性、虛擬化屬性的全新範疇。在達特茅斯討論過程中，與會者們沒有選擇“Simulated Intelligence”（仿真智能）、“Counterfeit Intelligence”（仿造智能）、“Twin Intelligence”（孿生智能）、“Imitative Intelligence”（偽造智能）、“Model Intelligence”（模態智能）或“Anthropomorphic Intelligence”（擬人化智能），而是一致接受了麥卡錫提出的“Artificial Intelligence”。或許他們意識到了，這裡的“Artificial”不僅意味着人工或人為，還蘊含着詞頭“Art”的深層意涵，即藝術性、虛構性與虛擬性。

因此，將“Artificial Intelligence”翻譯為“人藝智能”，而非“人工智能”，可能更能體現AI的本質特性。這一創新翻譯不僅在語言上更加貼切，還能夠帶來思維方式的深刻轉變，為AI研發指引新的方向。

儘管將“Artificial Intelligence”重新翻譯為“人藝智能”，可能引起爭議與現實權衡，比如第一，傳統認知慣性：目前“人工智能”已成為固定術語，改變翻譯可能需要較大的推廣和教育成本。第二，藝術的狹義理解：有些人可能將“人藝智能”誤解為僅與藝術領域相關，而忽視其廣義的技術與社會應用。

但是，我相孞，將“Artificial Intelligence”翻譯為“人藝智能”，不僅更富有創意，而且也有“發生學”意義的根據。Artificial一詞的詞頭“Art”，最早的意義與技術、技藝、技巧、技能相近，在希臘語中寫作“τεχνη(techne)”，在拉丁語中則寫作Ars。那個時候的藝術尚未分化，它與美、道德和實用都有關係。在相當長的歷史時期里，在人們的觀念中，藝術與技術幾乎沒有什麼區別，藝術家與工匠（中文稱“手藝人”）的社會地位，也大致差不多。18世紀法國思想家狄德羅(D. Diderot, 1713～1784)主編的《百科全書或科學、藝術、手工業詳解辭典》中，“技術”條目就是用的“Art”一詞，他把技術解釋為“為了完成特定目標而協調動作的方法、手段和規則相結合的體系”，這是已知最早的技術的定義。

與“人工智能”相比，“人藝智能”在詞義上更貼近“Artificial”一詞的多重含義，同時也強調了AI技藝的本質特性。這種翻譯的優勝性如下：

“人藝智能”的優越性

1. “Artificial”的多重含義

“Artificial”一詞不僅指“人工的”或“人為的”，還包含“藝術性”（artistic）、“虛擬性”（virtual）、“創造性”（creative）等隱含意義。

• 詞頭“Art”來源於拉丁詞根“ars”，意為“技巧、藝術或手法”。它並不限於機械性或勞動力驅動，而更多指向創造性和構建性。

• “人藝智能”捕捉了“人工”背後的藝術化和虛構性，精準傳達了AI本質上的虛擬構建屬性。具體而言，藝術化體現在生成式AI在文學、音樂、繪畫等領域的原創性創作中，例如AI能夠基於輸入生成具有藝術審美的詩歌或繪畫作品。虛構性則表現為AI通過算法模型構建虛擬環境或角色，例如虛擬助手和元宇宙中虛擬角色的交互，這些虛擬構建並非真實存在，但通過技術手段實現了逼真的模擬與功能化。

• 
  

1. 更符合AI的核心特性

• 
  

• 創造性（Creativity）：AI不僅是機械模擬，更以數據、算法和模型為基礎構建新知識、新行為和新系統，這種“創造”的過程與藝術創作本質上具有相似性。

• 虛擬性（Virtuality）：AI的智能多表現在“虛擬環境”或“虛擬交互”中，其推理和行為通常通過模型的虛擬運算完成。

• 工具性（Instrumentality）：AI既是人類設計的工具，也是擴展人類智能的藝術性工具。

• 
  

1. 突破“人工”一詞的局限性

• 
  

• “人工”更偏向勞動力替代，側重機械性和物理操作，容易忽略AI在設計、創新和智能表達中的藝術化特徵。

• 長期使用“人工智能”可能導致公眾誤解，認為AI僅是機械的延伸，而非創造性賦能的體現。例如，有些人將AI視為“工具型智能”，僅僅用於執行重複性任務，而忽略其在生成藝術、寫作和創造新知識領域的潛力。例如，早期對ChatGPT的認知僅局限於問答工具，直到它在生成文學作品和創意內容上的表現才逐漸改變這一認知，這說明長期固化的術語可能限制公眾對AI創造性特性的全面理解。

• 
  

1. 哲學視角下的“人藝智能”

• 
  

• “人藝智能”強調AI作為人類藝術與智能結合的產物，既體現了人類的創造力，又展示了AI在擴展這種創造力方面的潛力。

• 這一概念符合“交互主體共生”（Intersubjective Symbiosism）的理念，超越單純對抗或替代關係，倡導共生合作的哲學願景。

• 
  

1. 更好的文化適應性與未來思考

• 
  

• 從語言角度：中文中的“人藝”不僅涵蓋技術層面，還包含精神和文化層面，具有廣泛包容性。

• 從技術發展角度：AI未來發展方向將偏向藝術化與創造性應用（如生成藝術、情感計算等），“人藝智能”這一翻譯更能體現其文化與技術演變。

看清與突破AI研發中的三大瓶頸

將“Artificial Intelligence”重新定義為“人藝智能”，不僅僅是語言上的調整，更能幫助看清、克服、突破目前AI研發中面臨的三大瓶頸。

1. 高能耗與低能效的不匹配

當前主流AI模型（如深度學習）在訓練和推理過程中需要消耗巨量能源，導致應用成本陡增，成為大規模普及的障礙。例如，GPT-3的訓練據估算需要耗費超過1287兆瓦時的電力，相當於一輛普通燃油汽車連續行駛140萬公里的碳排放量。這種高能耗直接限制了AI技術的普及與應用，尤其是在能源資源有限的情況下。

• 
  

• 現狀：大型語言模型的訓練需要數萬千瓦時的電力，其能耗與智能性形成反差。

• 突破路徑：“人藝智能”強調系統設計的精巧性與效率，從自然和藝術中汲取靈感，開發低能耗、高能效的智能系統。

• 低功耗神經網絡：借鑑自然界高效的能量使用機制，開發輕量級、分布式神經網絡，減少計算冗餘。

• 藝術化分布式架構：通過模擬藝術創作的分步構建方式，設計模塊化AI系統，降低單點能耗。

• 
  

1. 系統思維的局限性

現階段的AI設計以系統性思維為核心，將信源（數據）、信道（算法）、信果（結果）分離處理。這種線性設計難以應對複雜動態系統，尤其在多元交互與不可預知環境中表現不足。

• 現狀：系統模塊割裂設計缺乏彈性與適應性。

• 突破路徑：“人藝智能”倡導動態共生與藝術化思維，為複雜環境中的AI設計提供新的方向。具體操作方式包括：

• 多模態協作：通過結合語言、視覺、聲音等多模態信息，實現AI系統在複雜交互場景中的更高適應性，例如智慧城市中同時處理交通、氣象和能源管理。

• 動態反饋機制：藉助實時數據輸入和動態調整算法，使AI系統能夠根據外部環境的變化即時優化自身行為，如智能製造中根據生產需求動態分配資源。

• 藝術啟發設計：借鑑藝術創作中的非線性思維，開發具有靈活性和創造力的算法框架，例如通過音樂和繪畫的靈感優化神經網絡的結構與功能。

• 共生式設計：不再割裂信源、信道、信果，通過動態交互實現自適應優化。

• 非線性競合模型：借鑑藝術創作中的旋律、節奏、和聲關係，為AI設計提供多層次動態調節的啟示。

• 
  

1. 數據+算法+算力+神經網絡 ≠ 智慧

儘管AI在數據處理和預測能力上表現卓越，但真正的智慧包含創造力、倫理價值和情感理解，當前模型難以達到這一高度。

• 現狀：現有AI模型以算力驅動，忽略智慧的多維度。

• 突破路徑：“人藝智能”強調藝術性與哲學性的融入，推動AI從模擬智能向規範智慧邁進。

• 藝術啟發智慧：通過融入藝術創作的審美與創造力，開發原創性表達的AI系統。

• 規範智慧框架：從哲學與倫理角度定義智慧內涵，使AI行為符合人類價值觀。

未來展望：人藝智能的可能性

以“人藝智能”替代“人工智能”，不僅是術語上的更新，更是思維範式的革命性轉變。這種新視角有助於：

1. 突破高能耗瓶頸：從自然與藝術中獲取啟發，構建低能耗高效能的智能系統。

2. 超越線性思維：通過藝術化的共生設計，增強AI在複雜環境中的適應力。

3. 規範智慧內涵：推動AI從模擬智能向真正智慧進化，同時融入人類價值觀。

結論

將“Artificial Intelligence”翻譯為“人藝智能”，是一種更貼合AI本質特性的創新嘗試。相比於“人工智能”，“人藝智能”更能體現AI的虛擬性、藝術性和創造性，同時突出了其作為人類藝術與技術結合產物的本質。如果在學術、哲學和技術研發中逐步推廣，這一翻譯可能成為更加精準且富有文化意涵的表達，為AI的未來發展注入新的活力，消除人們（如馬斯克、伊里亞、辛頓、赫拉利）對AI未來不確定性的疑慮，鼓勵和規範人們（如奧特曼）對Ai的樂觀精神！

2024年12月18日於溫哥華

ChatGPT4o翻譯如下：

Human-Artificial Intelligence: Redefining the Essence and Future of AI

In 1956, the Dartmouth Conference first proposed and established the concept and terminology of "Artificial Intelligence." This choice was based on the term "Artificial," which encompassed non-mechanical intelligent expressions and highlighted its potential for virtuality and creativity. In contrast, other proposed terms such as "Simulated Intelligence" or "Anthropomorphic Intelligence" were more confined to imitation or anthropomorphism, failing to fully capture the technological breadth and philosophical depth of AI. "Artificial Intelligence" emphasizes how artificially constructed intelligent systems transcend simple simulation to embody creativity and virtualized attributes. During the discussions, the conference did not choose "Simulated Intelligence," "Counterfeit Intelligence," "Twin Intelligence," "Imitative Intelligence," "Model Intelligence," or "Anthropomorphic Intelligence," but instead unanimously accepted McCarthy's proposed term, "Artificial Intelligence." Here, "Artificial" not only signified man-made or human-made but also carried the profound connotation of "Art," implying artistry, fabrication, and virtuality.

Therefore, translating "Artificial Intelligence" as "Human-Artificial Intelligence" rather than "Man-Made Intelligence" may better reflect the essence of AI. This innovative translation not only aligns more closely with the language but also brings a profound shift in thought, guiding new directions for AI development.

The Superiority of "Human-Artificial Intelligence"

1. The Multifaceted Meaning of "Artificial"

The term "Artificial" refers not only to "man-made" or "human-made" but also encompasses "artistic," "virtual," and "creative" dimensions.

• The prefix "Art" is derived from the Latin root "ars," meaning "skill, art, or craft." It is not limited to mechanical or labor-driven connotations but instead points toward creativity and construction.

• "Human-Artificial Intelligence" captures the artistic and fabricated aspects of "man-made," accurately conveying the virtual constructive nature of AI. Specifically, artistry is reflected in generative AI's original creations in literature, music, and painting, such as AI-generated poems or artwork based on inputs. Fabrication is manifested in AI's construction of virtual environments or characters through algorithmic models, such as virtual assistants and metaverse avatars. These virtual constructs are not real but achieve realistic simulation and functionality through technological means.

1. Better Aligning with the Core Features of AI

• Creativity: AI is not merely mechanical simulation but builds new knowledge, behaviors, and systems based on data, algorithms, and models, akin to the process of artistic creation.

• Virtuality: AI intelligence is often expressed in "virtual environments" or "virtual interactions," with reasoning and behavior accomplished through model-based virtual operations.

• Instrumentality: AI serves as a tool designed by humans and an artistic tool that extends human intelligence.

1. Overcoming the Limitations of "Man-Made"

• The term "man-made" leans towards labor substitution, focusing on mechanical and physical operations while overlooking AI's artistic attributes in design, innovation, and intelligence expression.

• Prolonged use of "man-made intelligence" may lead to public misconceptions, viewing AI merely as an extension of machinery rather than as a tool for creative empowerment. For example, some perceive AI as "tool-based intelligence" solely for repetitive tasks, neglecting its potential in generative art, writing, and new knowledge creation.

1. Philosophical Perspective on "Human-Artificial Intelligence"

• "Human-Artificial Intelligence" emphasizes AI as a product of the combination of human art and intelligence, showcasing human creativity and AI's potential to extend it.

• This concept aligns with the philosophy of "Intersubjective Symbiosism," transcending simplistic confrontational or substitutionary relationships and advocating for cooperative coexistence.

1. Better Cultural Adaptation and Future Considerations

• From a linguistic perspective: The term "human-artificial" not only covers technical aspects but also includes spiritual and cultural dimensions, offering broad inclusivity.

• From a technological perspective: The future direction of AI development leans towards artistic and creative applications (e.g., generative art, emotional computing), and the translation "human-artificial intelligence" better reflects its cultural and technological evolution.

Understanding and Overcoming the Three Major Bottlenecks in AI Development

Redefining "Artificial Intelligence" as "Human-Artificial Intelligence" is not merely a linguistic adjustment; it helps identify and overcome the current three major bottlenecks in AI development.

1. Mismatch Between High Energy Consumption and Low Efficiency

Current mainstream AI models (e.g., deep learning) consume massive amounts of energy during training and inference, leading to soaring application costs and hindering widespread adoption. For example, GPT-3 training is estimated to consume over 1,287 megawatt-hours of electricity, equivalent to the carbon emissions of a conventional gasoline car driving 1.4 million kilometers. This high energy consumption directly limits the widespread adoption of AI technologies.

1. Limitations of Systemic Thinking

Present AI design relies on systemic thinking, separating signal sources (data), channels (algorithms), and results (outputs). This linear approach struggles to handle complex dynamic systems, particularly in multi-variable interactions and unpredictable environments.

1. Data + Algorithm + Computing Power + Neural Networks ≠ Intelligence

Although AI excels in data processing and prediction, true intelligence encompasses creativity, ethical values, and emotional understanding, dimensions that current models fail to achieve.

Future Prospects: The Possibility of "Human-Artificial Intelligence"

Replacing "man-made intelligence" with "human-artificial intelligence" is not only a terminological update but also a revolutionary paradigm shift. This new perspective can contribute to:

1. Overcoming High Energy Consumption Bottlenecks: Drawing inspiration from nature and art to design low-energy, high-efficiency intelligent systems.

2. Surpassing Linear Thinking: Enhancing AI adaptability in complex environments through artistic and symbiotic design.

3. Defining the Essence of Wisdom: Promoting AI evolution from simulated intelligence to genuine wisdom while integrating human values.

Conclusion

Translating "Artificial Intelligence" as "Human-Artificial Intelligence" is an innovative attempt that more accurately reflects the essence of AI. Compared to "man-made intelligence," "human-artificial intelligence" better captures AI's virtuality, artistry, and creativity while emphasizing its nature as a product of the fusion of human art and technology. Gradually introducing this term in academic, philosophical, and technological discussions could establish a more precise and culturally meaningful expression, providing new vitality to AI's future development and addressing concerns about its uncertainties.