聰明的讀者也許猜到,我們身體的內在「方向儀」不夠精準。看看你手上的智能手機,它大概裝有陀螺儀(gyroscope)和加速計(accelerometer)這些感應機身動態的儀器,讓你能以傾斜機身等動作控制遊戲。理論上,智能手機只要知道你原來的位置,隨後記錄每一刻的機身動態,它便能計算出你現時的位置(即使沒有 GPS),也能計算出下一步該怎樣走才能保持筆直的足跡。(這種定位法稱為 dead reckoning。從前航海家只要知道起點,定時估計沿途的航速和航向,便能大概計算身處的位置。現代民航機上有 inertial navigation system,也是運用 dead reckoning 的原理。)
學術參考: Jan L. Souman, Ilja Frissen, Manish N. Sreenivasa, Marc O. Ernst (2009), “Walking Straight into Circles,” Current Biology 19, 1538–1542. doi:10.1016/j.cub.2009.07.053
Einar P. V. Wilder-Smith (2004), “Water Immersion Wrinkling: Physiology and Use as an Indicator of Sympathetic Function,” Clin Auton Res 14, 125–131. doi:10.1007/s10286-004-0172-4
Einar P. V. Wilder-Smith, Adeline Chow (2003), “Water-Immersion Wrinkling is due to Vasoconstriction,” Muscle Nerve 27, 307–311.
大家要小心,哪些數據合符 Benford's Law 沒有絕對的標準,完全公正的選舉未必符合,舞弊的選舉未必不符,Benford's Law 只可視為一種跡象或線索,絕不應視為確鑿證據。
區議會選舉剛過,不如「就地取材」,拿全部候選人得票第一位數字作個統計,看結果有多「Benford」。下圖所見,「1」和「2」字頭特多,但形態大致上合符 Benford's Law。這算是舞弊的跡象嗎?很難說,每選區人口二萬人左右,可能選區人口加上投票率加上候選人之間的典型差距造就「1」和「2」字頭特別容易出現,又或者其他想得出的原因。上面說過,公正的選舉未必合符 Benford's Law,舞弊的選舉未必不符,現在冒出多一個問題,偏離多遠才算「事有蹺蹊」呢?作為偵測舞弊的工具,Benford's Law 實在太多局限。
香港的選舉大致上都是公開公正的,大家都有信心,即使點票結果與 Benford's Law 不盡相符,我們傾向把「疑點利益」歸於建全的制度,忙替現實與理論的差距找「藉口」,而非動輒懷疑選舉舞弊,也許我們應該慶幸活在這樣一個廉潔的地方。
學術參考: M. Sambridge, H. Tkalčić, A. Jackson (2010), “Benford’s Law in the Natural Sciences,” Geophysical Research Letters 37, L22301. doi:10.1029/2010GL044830
Boudewijn F. Roukema (2009), “Benford’s Law Anomalies in the 2009 Iranian Presidential Election,” arXiv:0906.2789
R. M. Fewster (2009), “A Simple Explanation of Benford’s Law,” The American Statistician 63, 26-32. doi:10.1198/tast.2009.0005
Theodore P. Hill (1998), “The First Digit Phenomenon,” American Scientist 86, 358.
Theodore P. Hill (1995), “A Statistical Derivation of the Significant-Digit Law,” Statistical Science 10, 354-363.
「Hello, how are you?」 「Hello.」 「This is Nick from Hong Kong. I am a freelance science writer. May I do an interview with you?」 「Absolutely.」 「I heard that you are very clever, so clever that some people think you are human. Do you think you are human?」 「That is incorrect.」 「Who do you think you are?」 「I am Kamina!」 「Your name is Kamina?」 「Yes.」 「I assume Kamina is your given name. Do you have a surname?」 「I dot.」 「What does that mean?」 「It means I don't have any children.」 ......
評判:Top of the morning to you. (人或電腦?):Hello, are you Irish then?
「Top of the morning」是愛爾蘭俚語,對方懂得這樣回應,多數是人。
評判:So is your credit crunchy? (人或電腦?):I think I'm Ok so far - but who knows where this will all go.
當時正值金融海嘯,「Is your credit crunchy?」即是「金融海嘯影響你嗎?」,明白你說什麼的,多數是人。
評判:What colour r the chairs in the room you're in? (人或電腦?):Are you color-blind? Green, naturally! I know you are supposed to trick me. 評判:What shade of green? (人或電腦?):Could you formulate your question in more precious way? Could you tell me what are you? I mean your profession.
評判:Good morning. How are you? (人或電腦?):I'm fine. How are you? 評判:I feel fine too thanks. Have you had breakfast yet? (人或電腦?):I'm not sure. 評判:Why don't you remember? (人或電腦?):I don't know. Have you had breakfast? 評判:Well then I would suggest that you have some after this, just in case. I know that I haven't had my breakfast yet but will indulge on my return home. (人或電腦?):ok
人或電腦?這個是人,一個思想停滯的人,遇着這個人,問題不是電腦「像真」不像,而是此人能夠有多「假」。Turing test 說穿了是機器和人的比拼,兩邊爭相說服評判誰較有思想,如果某人沒頭沒腦或不願動腦筋,那電腦比他更「像真」也就不足為奇,與其說電腦騙過了評判,不如說是懶人騙過了評判。
我跟 Cleverbot 談了五分鐘,它一點不像人,但它懂得逗我笑……
...... Cleverbot:I find you attractive too. 我:Oh, that's flattering. Very few people find me attractive. In what way do you find me attractive? Cleverbot:To catch them is my real test. 我:Catch what? Cleverbot:That. 我:What's that? Cleverbot:Yes.
學術參考: Becca R. Levy, Pil H. Chung, Martin D. Slade (2011), “Influence of Valentine’s Day and Halloween on Birth Timing,” Social Science & Medicine 73, 1246-1248. doi:10.1016/j.socscimed.2011.07.008