人工智慧：一邊革新，一邊剝奪？Artificially Creative

By Anantika Mehra

人工智慧的概念十分寬泛，對不同的人意義不同。對於一些印度人來說，它是印度超級巨星拉吉尼坎塔在《寶萊塢機器人之戀》中扮演的機器人Chitti的形象；對於那些美劇粉來說，《黑鏡》則預示了當電腦足夠了解人類後可能帶來的後果，令人不寒而慄；對於更具技術頭腦的人們來說，人工智慧可能只是一行行的代碼，能夠實現面部識別、自然語言處理等功能。

Artificial Intelligence (AI) is an extremely wideranging term that can mean different things to different people. For some of us in India, it conjures an image of Chitti, the humanoid robot played by the Indian superstar Rajnikanth in the movie Robot. The 「Playtest」 episode of the popular TV show Black Mirror is another chilling example of what is possible if computers learn enough about us. For the more technical, however, it may inspire an image of lines upon lines of code, and functions such as facial recognition or natural language processing.

臉譜網人工智慧演算法可以收集你的數據，來推想你最可能買的東西；你玩的視頻遊戲，不少公司正在測試的無人駕駛汽車，你所使用的谷歌和亞馬遜推出的各種虛擬設備，都應用了人工智慧。如今，人工智慧已經無處不在。

Whether in Facebook algorithms serving you up the most targeted advertising, video games, self-driving cars or Google and Amazon』s virtual assistant devices, AI can be found everywhere these days.

人類對人工智慧未來的猜想經常見諸於媒體報道和文學作品中：我們總是在懼怕有朝一日機器會佔領地球。無論是《機械公敵》、《終結者》還是《機器管家》中，機器人一開始僅僅是機械，但最終都獲得了人性。在電影《黑客帝國》中，機器更是成了主宰者。但目前我們真正需要擔心的是，儘管人工智慧使我們的工作變得更加簡單，它們也在剝奪我們的工作。喬治·奧威爾在小說《一九八四》中描述了機器開始製造廉價的色情作品並為人們寫故事的情景。在過去的幾年裡，日本開發的一款人工智慧程序曾寫出了差點贏得文學獎的小說，美國開發的一款人工智慧演算法正在續寫《權力的遊戲》。至少在一定程度上，奧威爾在其反烏托邦作品中預言的情景已經化為現實。

In our collective imagination, as manifested in media and literature, there has always been some fear of machines taking over the planet. From 2001: A Space Odyssey to iRobot, Terminator and Bicentennial Man, fiction has constantly depicted robots achieving some sort of humanity. In The Matrix, the machines literally take over. But a real concern is that despite AI making our jobs far simpler, it is also taking them away. In novels as early as George Orwell』s 1984, machines were producing cheap pornography and writing stories for the public. In recent years, a Japanese AI program wrote a story that almost won a literary award, and an American AI algorithm is writing the next Game of Thrones book. This, along with many other parts of Orwell』s dystopia, is now our reality.

2017 年 12 月3 日，第四屆世界互聯網大會「世界互聯網領先科技成果發布」在烏鎮互聯網國際會展中心舉行，發布了世界互聯網領域領先科技成果。圖為微軟小冰——情感計算人工智慧。

《麻省理工科技評論》雜誌總有令人著迷的見解。例如，今年的EmTech全球新興技術峰會展示了人工智慧在圖像處理方面的最新進展。從替換背景、調整亮度、到替換背景中的特定對象等操作，幾乎沒有人工智慧做不到的事情。一方面，這讓攝影師和圖像編輯從艱苦的工作中解脫出來。另一方面，人類的創造力也會被人工智慧「外包」。這類人工智慧將完全改變我們對藝術和美的感知，也許它將超越技術層面上的美，使美學變得更加概念化。

MIT Technology Review always has fascinating insights to offer. For example, the EmTech Digital Conference this year had a display demonstrating just how much AI can do with an image. From replacing backgrounds so that improper lighting is corrected to replacing specific objects in the background, it seems AI can do anything. Such technology could save photographers and image editors hours of painstaking work, but is also sort of outsources human creativity. AI like this can and will completely change our perception of art and beauty and perhaps take it beyond the realm of technical beauty to a more conceptual plane.

另一個超級酷的事情是，人工智慧可以學會武術，再將自己的戰鬥技巧編排成電影鏡頭或視頻遊戲。這種技術幾乎完全還原了人類在同樣環境下做出的反應，可以取代人類精細的舞蹈和動畫設計。

Another super-cool thing to do with AI is to teach it martial arts and then let it choreograph its own fight in a movie sequence or a video game. It can precisely mimic human reactions in such a situation and completely eliminate the need for painstaking choreography and animation.

此外，人工智慧在醫學領域的應用也正在逐漸顯現。從美國食品藥品管理局最近批准的用於檢測糖尿病性視網膜病變的裝置，到谷歌雲科學家在簡化放射學中應用的圖像處理、升級識別癌組織的演算法、增強圖像處理以分析體液、以及通過面部識別診斷罕見疾病等領域所取得的成果，人工智慧的前景不可限量。

And the entire AI diagnostics market is emerging gradually. From the recent device that the United States』 FDA approved to detect diabetic retinopathy to Google cloud scientist Jia Li』s work on simplifying the image processing used in radiology, training algorithms to recognize cancerous tissue, analyzing body fluids with enhanced image processing and facial recognition for identifying rare diseases, the possibilities are limitless.

然而，人工智慧在醫學的應用也引起了人們對下列兩個問題的擔憂。首先，人工智慧最終會讓醫生變得多餘嗎？其次，如何使演算法遵循道德倫理準則？

Two concerns arise from the emergence of this field though. First, will AI ultimately make doctors redundant? And secondly, how does one make an algorithm follow a code of ethics?

我看過一部名為《人工智慧時代》的紀錄片，描述的正是人類對於人工智慧的恐懼。人類擔心人工智慧會使人類自身成為多餘的勞動力。自從進入現代文明社會，人們便一直擔心科技會奪走人類的工作。最初，我們建造了機械工具和機器，使我們的工作變得更加輕鬆。隨後，我們開發了計算機，代替很多人類大腦可以完成的機械工作，大大節約了時間。隨著人工智慧的發展，我們已經進入下一階段：機器終於有可能取代我們大腦中所做的創造性工作。機械化的機器人需要人類運行它們，但是人工智慧可以自主觀察和學習，具備自我決策的能力。它可以消化巨大的數據集合，並且比人類大腦更易識別與計算能力相關的規律模式。人工智慧工作的成本也比人類僱員小得多。目前，人工智慧正在取代技術含量較低的人類工作。隨著機器變得越來越強大，未來我們不知道它將對其他工種帶來什麼威脅。

I watched a documentary called Humans Need Not Apply that addresses the fear of AI making humans redundant as workers. The fear of technology taking away human jobs has persisted as long as modern civilization has. First we built mechanical tools and machines to make our work easier. Then we moved on to computers, which did a lot of mechanical work such as calculations that our brains could, but in much less time. With AI, we have moved on to the next step, because, for the first time, machines can take over the creative work that we do with our brains. Mechanized robots require humans to run them, but AI is autonomous, in that it can see and learn. It has the ability to make decisions. It can digest huge data sets and identify patterns that require computational power much more than the human brain. And all of this is at a cost much lesser than that of a human worker. For now, AI is taking away low-skilled jobs. But as machines become more and more powerful, we don』t know what other jobs could be in jeopardy.

第二個問題，即倫理問題，也是極其重要的。人工智慧演算法的價值體系或倫理觀念很大程度上取決於其創造者的倫理觀念。尤其是在需要人類共情的領域，一些科學家認為，用機器替代人類是對人類尊嚴的踐踏。

The second point, of ethics, is also extremely important. The value system or ethics that an algorithm has depends a lot on the ethics of the humans who create it. Some scientists have opined that especially in fields that require empathy, replacing humans with machines is a threat to human dignity.

2017 年1 月13 日，「書法家」阿里雲人工智慧ET 正在中國杭州現場書寫春聯。

即使並非出於惡意，人類也可能做出具有破壞性的人工智慧。批判這項技術的女權主義者認為，人工智慧系統多由男性設計和運行，必然會帶有性別歧視。這確實是一個現實的問題。最近，科學家測試了Siri和微軟小娜這樣女聲機器人程序對性騷擾的反應，結果證實了這個問題的存在。它們並沒有對性騷擾做出「反抗」，有時它們的回應幾乎是在鼓勵對方。雖然這些人工智慧的演算法並不是針對性騷擾而設計的，但這個實例表明，技術的創造者需要具備道德義務，以規範用戶，確保使用者不對這種女聲語音助手進行「調戲」。人工智慧需要有更多不同領域的人參與它們的創造過程，以加強對不同社會身份人們需求的了解。

Another thing humans may do is to build AI that is destructive, even if it is not malicious. The feminist critique of technology, which says that a system designed and run mostly by men is bound to reproduce their biases, rings even truer here. Recently, bots like Siri and Cortana, which have female voice representations, were tested for their responses to sexual harassment, and the results were telling. They did not stand up to the sexual harassment, and sometimes their responses were almost encouraging. Of course, these algorithms were not designed for the purpose of being responsive to sexual harassment, but the argument here is that the creators of these technologies have an ethical responsibility to ensure that users do not normalize this sort of harassment of female voices. AI needs to have more diverse creators to ensure greater sensitivity to the needs of people with different social identities.

以當下的發展速度，一方面，人工智慧將繼續革新我們的生活；另一方面，它們又會剝奪我們的工作。我們所能做的就是等待和觀察。

At the rate of its advancement today, AI promises to revolutionize our lives and steal our jobs. All we can do is wait and watch.

「

作者Anantika Mehra為孟買塔塔社會科學研究院碩士研究生。

The author is a master』s student at Tata Institute of Social Sciences, Mumbai

本文為《中印對話》獨家稿件，歡迎分享，媒體轉載請聯繫我們。

This article is exclusive to China-India Dialogue. Feel free to share this article. To reprint this article, please contact us for permission.

喜歡這篇文章嗎？立刻分享出去讓更多人知道吧！