Cédric Villani:人類與機器的錯亂關係還會維持【AI大神說】

撰文:
最後更新日期:

曾幾何時,01哲學小編認為,哲學家是這個世界上最性感的職業,因為哲學家看起來總是知道那些你還不知道、卻又決定你命運的真相。其中,催人渴求的未知與性的魅惑有時候根本無法區分,更何況遇上特別帥氣的男哲學家,例如【AI大神說】上一集介紹的馬庫斯・加布理爾(Markus Gabriel),或者你還可以想想沙特、卡繆或者維根斯坦⋯⋯

 

 

Markus Gabriel:不要神化人工智能【AI大神說】

 

怎麼,原來你們都不同意嗎?認為這都是哲學小編自己的 FF 嗎?好吧!如果連擅長營造語言暈眩的哲學家,都沒辦法跟性感相關聯,那麼離群索居、枯坐冷板凳,終日操作公式記號與邏輯演算的數學家,想必就更困難了吧?

然而令人難以置信的是,當代法國竟然有一位「數學界的 Lady Gaga」!他是媒體的寵兒──鏡頭總是聚焦他的中分長髮、他那十九世紀花花公子般的衣著與一枚個人標誌性的蜘蛛飾物。不過最重要的是,他真的是一位徹頭徹尾的數學家,研究的課題諸如「regularizing effects of grazing collisions in kinetic equations」和 「Space-inhomogeneous convergence to equilibrium」(小編就不獻醜把它們強翻成中文了)。2010年,他還獲得了數學界的奧斯卡獎──菲爾茲獎(Fields Medal),以表彰他在代數和幾何領域的卓越成就。他為自己設定的重要任務,是將數學學科的樂趣與迷人之處藉助媒體向大眾宣揚。

 

「數學界的 Lady Gaga」賽德里克・維拉尼(Cédric Villani)教授

 

他就是賽德里克・維拉尼(Cédric Villani)教授,索邦大學亨利龐加萊研究所(the Institut Henri Poincaré)總監,更是法國總統馬克龍(他最近陷入麻煩,必須面對半世紀以來當地最大規模暴動)派往世界各地宣傳法國人工智能政策的特使。可見,他還是一位政治家。就在剛剛過去的十一月,維拉尼教授到訪香港;藉此機會,01哲學編輯與維拉尼教授在香港科技大學商學院碰了面,並向他提出了兩個關於AI的哲學問題。讓我們一起看看這位數學大神對於 AI 懷有怎樣性感的觀點。

【以下附上英文原文對話,以供讀者參考】

 

維拉尼教授曾在 TED 講座開講,標題為「數學有甚麼好性感的?」 (What’s so sexy about math?),影片連結:http://bit.ly/2E0Mms5

 

01哲學:第一個問題關於「愚蠢」。電影、娛樂與流行科幻小說中的 AI 形象一定程度上具有誤導性。我自己更喜好「機器智能」這個說法而不是「人工智能」。機器智能是基於概率來發揮功能的。為了找到問題的最恰當的解決方案,機器愚蠢地需要執行大量運算。而我們人類並不總是先進行大量運算才把問題解決掉。我們先理解問題,然後便迅速將問題解決了。關於機器的愚蠢而不是智能,您有什麼看法?

01 Philosophy: The first question is on stupidity. The images of AI in films, entertainment contents and popular science fiction are misleading in some sense. I myself prefer the term machine intelligence to artificial intelligence. The way the machine functions is based on probability. In order to figure out the most proper way to solve the problem, the machine, stupidly, needs to do a huge number of calculations. We humans are not always solving problems by firstly performing a huge number of calculations. We understand the problem and then solve it immediately. What do you think of the stupidity of the machine, instead of its intelligence?

賽德里克・維拉尼:是啊,這是一個很好的問題。在現今關於人工智能的算法中並沒有智能可言。它確實是一部愚蠢的機器,圍繞著大量概率、統計、關聯性等問題工作。一些科學家甚至會說那根本就不是科學,因為機器都在努力複製任務而並不理解它們究竟在做什麼。在醫療領域,機器能夠定位癌症卻並不知道癌症意味著什麼。而這卻是可行的。一部自動翻譯機器恰恰在變得愚蠢、以自動複製策略取代智能語法的、語義學的、本體論的方法時才更為有效。這對人類智能來說是相當煩惱的。

Cédric Villani: Yes, it's a very good question. In current algorithm about artificial intelligence, there is no intelligence. It's indeed rather a stupid machine, working on a lot of probability, statistics, correlation etc. Some scientists will even say that's not science, because they strive to reproduce the tasks without understanding what they are doing. In health, it would be about identifying cancer without knowing anything about cancer. And it works. An automatic translation became much better when it turn to a stupid, automatic reproduction strategy, rather than intelligent syntactic, semantic, ontological approach. Very vexing for human intelligence. 

然而,將會有一個時刻,正如我在講座中所堅持的那樣,在那時概率方法將與更為本體論的方法結合起來。這是一條路徑,在其中機器的強力破解方法已不再足夠。某些智能,即原創思想,將必然會發生。而且,關涉到對於人類認知過程的理解,以及我們可以從自動化智能與人類智能的來來回回當中學到的東西──如今它們還大相徑庭,將會有極多的邊緣地帶被發現。

However, there would come a time, I insist this in my talk, in which the probabilistic approach has to be combined with a more ontological approach. And it is a way in which brute force would not be sufficient. Some intelligence — original thinking ── would have to take place. And also, there are enormous margins to be found in respect to the understanding of our cognitive processes, and what we can learn back and forth from automatic intelligence and the human intelligence - so far they are extremely different. 

還有我們必須承認某些案例中的算法非常聰明從而改變了我們的知識。或許最矚目的就是圍棋博弈。在由 DeepMind 開發的機器中,機器做出的一些舉動是完全不可預測的,並不僅僅是複製人類知識。它發現了某些新的舉動是人類從來沒有思考過的。所以,人工智能的愚蠢會挑戰人類智能並且幫助我們尋覓出新的領域──智能的領域。這並不是非黑即白。
 
Also we have to acknowledge that these algorithms in some cases have found very clever and thus they change our knowledge. Maybe the most spectacular is the emblematic Go game-playing. In the Machine built by DeepMind, some of the moves that it has made were totally unexpected — not just reproducing the human-based knowledge. It was finding some new moves that no human has ever thought of. So, this artificial stupidity can challenge our intelligence and help us in finding new territories — intellectual territories. It is not that black and white.

 

訪談現場的維拉尼教授

 

01哲學:第二個問題關於「熵」。機器與人類之間存在差異。在人類生命中,總有一些東西不可預測、不確定,甚至是錯的,即總有一種不透明性。但是機器可以是透明的:一切都早已包含在機器運作範圍之內。考慮到熵,即創生的終結,概率的終結,可能性的終結,會存在一種情況,人工智能將把人類帶到一切都早已被人工智能計算殆盡,從而再也沒有新的事物的地步?

01 Philosophy: The second question is about entropy. I think there are some differences between machine and human. In human lives, there are something unpredictable, uncertain and even wrong — non-transparency. But in machine, it can be transparent — everything already contained in its operation. To consider the question of entropy, namely, the end of production, the end of probabilities, the end of possibilities, can it be a situation that AI will bring us into a situation that everything have already been calculated by AI - there is nothing new anymore?

賽德里克・維拉尼:首先,我們是基於功能失調與功能相結合的奇特存在,其中一些不運作的東西一直面臨著猶如處於演化與改良的遊戲的狀態。你或許知道,一位美國生物學家史提芬(Stephen Jay)會稱之為熊貓的拇指,顯示出熊貓儘管沒有拇指,但它需要拿住竹子。從演化上說,為熊貓摘竹子提供的一個人工拇指,就像在某樣不完美的事物之上重構、優化某樣事物,卻保留不完美,並且一直與它共同生活下去。我們一直必須處理這種在我們身上隨處可見的不完美。

Cédric Villani: First, we are strange beings based on the combination of dysfunctionality and functionality, in which some not-working things have been faced as if they were in the game of evolution and improvement. You may know this, an American biologist Stephen Jay could call it The Panda's Thumb — showing pandas do not have the thumb, but it needs to take the bamboo. Evolutionarily speaking, an artificial thumb for panda's bamboo-picking, it is like reconstructing, perfecting something on top of something imperfect but keep the imperfection and live with it always. We always have to handle with this imperfection that we have everywhere. 

 

點擊查看更多維拉尼教授圖片:

 

我曾為兩部漫畫寫作劇本,其中一部是個科幻故事:在某個遙遠的未來,人類創造的機器人還在,人類卻不在了。機器人要重新發現人類是什麼。漫畫中的關鍵字句由我節選自一首法語歌曲:「我們是方程式的一個錯誤結果」。我對這種包含在我們當中的不完美情有獨鐘。它是我們自我身份的組成部分。

I wrote scripts for two comics, one of them is a science fiction story, which in a long-distant future, there are robots created by mankind, but mankind is not there. There are robots rediscover what was mankind. The key sentence in the comic which I took from a French song is "we are the result of a wrong equation". I like this imperfection so much in us. It is part of our identity. 

現在你或許要問「人工智能將會帶來更為完美同時也是令人恐懼的東西嗎?」在這一刻,這毋庸置疑是一項理論爭辯,因為人工智能在如此實用主義的路徑上得到發展。我們也需要如此之多的工作和努力來獲取與人工智能的良好合作。我的意思是,人類與機器,事情還會持續誤入歧途一段相當長的日子。我們還將持續擁有一個非常不完美的世界,而如今甚至變得更不完美而不是完美一段相當長的日子。

Now we may ask "will the AI bring in more perfection which might be frightening in some way?" For the moment, it really is a theoretical debate, because AI has developed in such a pragmatic way. And it too needs so much work and effort for us to have a good cooperation with the artificial intelligence. I mean, the human and the machine, things will continue to go wrong for a long time. We'll continue to have a world that is very imperfect and currently going rather more imperfectly than perfectly for a long time. 

關於熵以及這將把我們引向何方的詮釋,非常不清晰。就熵或複雜性而言,將我們的大腦與人工智能作比較嗎?我們的大腦描述起來要更為複雜,更為有效率,特別在以下意義上,即我們的大腦於如此緊密的部位耗費如此少的能量,而機器卻如此巨大並且耗費巨大能量。我們的大腦就印象和感覺而言更為不完美。我們僅僅探索了可能事物的一小部分,但是直覺卻誘導我們認為把握了更高機會去給出解決方案,而機器則是野蠻地探測:它非常暴力。這是完全不同的路徑。

In terms of the interpretation about entropy and where will this lead, it is very not clear. Is it comparing our brain to the artificial intelligence, in terms of entropy, in terms of complexity? Ours is much more complicated to describe, much more efficient also in a sense we use so little energy in such a compact place when we have this big machine which use huge energy whatever. Ours is much more imperfect in regard with impression and feeling. We are exploring just a small portion of possible things, but intuition leads us to think there are higher chance to give a solution, whereas machine is brutally exploring: it is very violent. It is a very different kind of approach.

我們人類在更為脆弱、更為有彈性的建構這一邊。我們的係統內有更多冗餘,而且非常穩固。我們知道,你大腦中有一些枝幹可能遭受打擊,而我們還可以繼續思想。然而計算機倘若有一部分芯片損毀了,它就不能運作,就必須拿些別的芯片來替換。所以在結構上真是大相徑庭的。由生命演化出來並在結構上邊執行一些結構原本不該執行的運算。我覺得人類智能比人工智能更性感,同時也更為神秘。

We remain on the side that is more fragile, more resilient construction. There is more redundancy in our system, very robust. We know you may have some trunks of your brains hit by stroke or something, and still we can continue to think. While in the computer if there are some parts of chips that are damaged, it just doesn’t work, it needs to just get other ones to replace them. So it is really a different construction: evolved from life and performing some computation on top of construction that is not supposed to do it. I think this is much more fascinating than the artificial intelligence; but also much more mysterious. 

 

 

技術人文實驗室【AI大神說】欄目將第一身與思想大神、技術大神接觸,為讀者帶來大神們對於人工智能和當代技術的新鮮思考與犀利觀點。

 

 

關於當代技術對人類社會的重塑,我們應如何看待?哲學可以對此給出甚麼分析、探索與回應?

01哲學「技術人文實驗室」於11月正式成立,作為實驗室第一彈專題研習活動,我們即將舉辦【哲學1001工作坊第七期】,以「機器智能時代的形上學」為主題,歡迎報名參加!

工作坊詳情及報名連結

 

 

《01哲學》,哲學入門,深入淺出,更好地理解,更好的邏輯。
立即下載《香港01》App:https://hk01.onelink.me/FraY/hk01app



X
請使用下列任何一種瀏覽器瀏覽以達至最佳的用戶體驗:Google Chrome、Mozilla Firefox、Internet Explorer、Microsoft Edge 或Safari。為避免使用網頁時發生問題,請確保你的網頁瀏覽器已更新至最新版本。