1
00:00:00,620 --> 00:00:04,460
我細個嗰陣係個書蟲

2
00:00:05,320 --> 00:00:07,496
我估你哋一啲人都係

3
00:00:07,520 --> 00:00:08,736
（笑聲）

4
00:00:08,760 --> 00:00:11,976
仲有你呀，阿生
笑得最大聲嗰個，睇嚟宜家都仲係

5
00:00:12,000 --> 00:00:14,256
（笑聲）

6
00:00:14,280 --> 00:00:17,776
我喺德州北部平原一個鎮仔大嘅

7
00:00:17,800 --> 00:00:21,136
我阿爺係牧師，我老豆係警察

8
00:00:21,160 --> 00:00:23,510
所以自細就係乖乖仔

9
00:00:24,040 --> 00:00:27,296
睇微積分書當娛樂

10
00:00:27,320 --> 00:00:28,856
（笑聲）

11
00:00:28,880 --> 00:00:30,576
你都係咧

12
00:00:30,600 --> 00:00:34,336
於是我自己整
激光啦、電腦啦、火箭模型啦

13
00:00:34,360 --> 00:00:37,360
最後仲喺房裏邊整埋火箭燃料

14
00:00:37,960 --> 00:00:41,616
用我哋宜家科學嘅講法

15
00:00:41,640 --> 00:00:44,896
係搞搞震無幫襯

16
00:00:44,920 --> 00:00:46,136
（笑聲）

17
00:00:46,160 --> 00:00:48,170
就喺嗰陣

18
00:00:48,170 --> 00:00:51,500
Stanley Kubrick 拍嗰部
《2001太空漫遊》上畫

19
00:00:51,500 --> 00:00:53,800
我嘅生活從此改寫

20
00:00:54,280 --> 00:00:56,336
我係部片嘅忠實擁躉

21
00:00:56,360 --> 00:00:58,896
特別中意哈爾 9000

22
00:00:58,920 --> 00:01:00,976
哈爾係一部機械人

23
00:01:01,000 --> 00:01:03,456
專為太空船導航而設計

24
00:01:03,480 --> 00:01:06,016
指引太空船由地球飛往木星

25
00:01:06,040 --> 00:01:08,096
但哈爾都有缺點

26
00:01:08,120 --> 00:01:12,400
佢最後為實現目的而罔顧人類安全

27
00:01:12,840 --> 00:01:14,936
哈爾雖然係虛構

28
00:01:14,960 --> 00:01:17,616
但佢反映咗我哋嘅恐懼

29
00:01:17,640 --> 00:01:24,786
就係怕無感情嘅人工智能
以後會操控人類

30
00:01:25,880 --> 00:01:28,456
我認為呢種恐懼唔成立

31
00:01:28,480 --> 00:01:32,496
講真，人類呢一刻嘅歷史
可以話係最輝煌嘅

32
00:01:32,760 --> 00:01:37,736
我哋冇向身體同大腦限制低頭

33
00:01:37,760 --> 00:01:43,126
反而製造出精良、複雜又美觀嘅機器

34
00:01:43,126 --> 00:01:47,136
協助人類逹至超乎想象嘅領域

35
00:01:47,726 --> 00:01:51,760
我以前喺空軍學校同太空指揮部做嘢嘅

36
00:01:51,760 --> 00:01:53,976
咁而家，我做咗系統工程師

37
00:01:54,000 --> 00:01:59,386
最近我著手於美國太空總署嘅
火星任務嘅一個工程問題

38
00:01:59,386 --> 00:02:01,856
到目前為止，所有上月球嘅太空船

39
00:02:01,880 --> 00:02:07,036
我哋都可以留喺休斯頓控制中心控制

40
00:02:07,040 --> 00:02:10,576
不過，火星比月球遠 200 倍

41
00:02:10,576 --> 00:02:13,992
於是乎由地球傳送到火星嘅訊號

42
00:02:14,000 --> 00:02:17,000
平均要花 13 分鐘先可以送到

43
00:02:17,000 --> 00:02:18,690
所以如果中途出咗咩問題

44
00:02:18,690 --> 00:02:20,840
我哋都唔夠時間解決

45
00:02:20,840 --> 00:02:23,336
所以解決方法係

46
00:02:23,360 --> 00:02:28,986
我哋要將任務控制台
裝喺太空船獵戶號嘅墻板裏邊

47
00:02:29,000 --> 00:02:31,896
另一個方法係

48
00:02:31,920 --> 00:02:36,716
喺人類登陸之前
喺火星表面放置一個機械人

49
00:02:36,720 --> 00:02:38,376
機械人前期負責興建設施

50
00:02:38,400 --> 00:02:41,760
後期就加入科學團隊擔當協助

51
00:02:43,400 --> 00:02:46,136
而當我企喺工程師嘅角度嚟睇

52
00:02:46,160 --> 00:02:51,546
我好清楚我要打造嘅
係一個高智能、富團隊精神

53
00:02:51,566 --> 00:02:53,936
同善於交際嘅人工智能

54
00:02:53,960 --> 00:02:58,256
換句話嚟講，我要整到好似哈爾咁樣

55
00:02:58,280 --> 00:03:00,696
不過係唔會當人類
係冚家鏟要消滅嘅機械人

56
00:03:00,720 --> 00:03:02,080
（笑聲）

57
00:03:02,920 --> 00:03:04,736
等陣先

58
00:03:04,760 --> 00:03:08,656
咁係咪真係有可能整個咁嘅機械人先？

59
00:03:08,680 --> 00:03:10,136
其實，係可以嘅

60
00:03:10,160 --> 00:03:14,926
好多時，設計人工智能元素係好難嘅

61
00:03:14,926 --> 00:03:19,616
咁唔係講緊整機械人頭髮

62
00:03:19,640 --> 00:03:22,296
學 Alan Turning 話齋

63
00:03:22,320 --> 00:03:24,696
我無興趣整隻有感覺嘅機器

64
00:03:24,720 --> 00:03:26,296
亦都唔係生產另一個哈爾

65
00:03:26,320 --> 00:03:28,736
我所追求嘅係一個會簡單思考

66
00:03:28,760 --> 00:03:32,210
同有少少智慧嘅機械人

67
00:03:33,000 --> 00:03:35,970
自從哈爾喺電影出現之後

68
00:03:35,970 --> 00:03:37,656
計算科學已經發展咗好多

69
00:03:37,680 --> 00:03:40,896
我想像得到，如果發明佢嘅
Chandra 博士今日喺度嘅話

70
00:03:40,920 --> 00:03:43,256
佢實有好多問題問我哋

71
00:03:43,280 --> 00:03:49,470
我哋有冇可能
從數以億計嘅設備裡面讀取數據

72
00:03:49,470 --> 00:03:53,176
同時預測機器犯嘅錯誤
同埋及早更正呢？

73
00:03:53,200 --> 00:03:54,416
可以

74
00:03:54,440 --> 00:03:57,616
我哋可唔可以
整隻機械人出嚟識講人話？

75
00:03:57,640 --> 00:03:58,550
可以

76
00:03:58,550 --> 00:04:00,960
我哋可唔可以整隻機械人

77
00:04:00,960 --> 00:04:05,256
識得識別物體、辨別情感
自帶情感、打機，甚至識讀唇語？

78
00:04:05,280 --> 00:04:06,210
可以

79
00:04:06,210 --> 00:04:09,446
我哋可唔可以整隻機械人識訂立目標

80
00:04:09,446 --> 00:04:12,296
實現目標兼且喺過程中自學？

81
00:04:12,320 --> 00:04:13,410
可以

82
00:04:13,410 --> 00:04:16,896
我哋可唔可以整隻
有思維邏輯嘅機械人？

83
00:04:16,920 --> 00:04:18,416
我哋宜家嘗試緊

84
00:04:18,440 --> 00:04:22,470
我哋可唔可以整隻機械人
明白道德觀念同底線？

85
00:04:22,480 --> 00:04:24,870
呢個任務我哋責無旁貸

86
00:04:25,360 --> 00:04:26,736
咁姑且我哋有可能

87
00:04:26,760 --> 00:04:31,836
為呢個任務或者其他任務
整隻咁樣嘅機械人

88
00:04:31,840 --> 00:04:33,230
跟住你實會問

89
00:04:33,230 --> 00:04:35,856
咁嘅人工智能會唔會造成威脅？

90
00:04:35,880 --> 00:04:40,796
時至今日，每項新科技面世
唔多唔少都會帶嚟不安

91
00:04:40,800 --> 00:04:42,496
以前啲人第一次見到汽車時

92
00:04:42,520 --> 00:04:46,536
就驚車禍會造成家破人亡

93
00:04:46,560 --> 00:04:49,256
以前第一次見到電話時

94
00:04:49,280 --> 00:04:52,176
啲人就驚人同人之間嘅交流會受到破壞

95
00:04:52,200 --> 00:04:56,136
曾幾何時啲人見到文字可以傳送

96
00:04:56,160 --> 00:04:58,656
又驚人類嘅記憶力會喪失

97
00:04:58,680 --> 00:05:00,736
呢啲不安喺一定程度上嚟講無錯

98
00:05:00,760 --> 00:05:03,080
但同時呢啲科技

99
00:05:03,080 --> 00:05:08,620
拓闊咗人類嘅體驗

100
00:05:09,840 --> 00:05:12,180
我哋不如再講遠啲

101
00:05:13,120 --> 00:05:17,856
我唔驚呢啲人工智能面世

102
00:05:17,880 --> 00:05:21,696
因為佢最終會接受人類嘅一啲價值

103
00:05:21,720 --> 00:05:23,900
試下咁唸：製造感知機械人

104
00:05:23,900 --> 00:05:27,100
同製造以前傳統嘅軟件密集型機械人

105
00:05:27,100 --> 00:05:28,560
係有根本性分別

106
00:05:28,560 --> 00:05:29,840
我哋而家唔係編程機械人

107
00:05:29,840 --> 00:05:31,300
我哋係教機械人

108
00:05:31,300 --> 00:05:33,696
為咗教機械人識別花

109
00:05:33,720 --> 00:05:36,736
我攞幾千種我鍾意嘅花畀佢睇

110
00:05:36,760 --> 00:05:39,016
為咗教曉機械人點打機——

111
00:05:39,040 --> 00:05:42,600
我真係教㗎。你都會咁做

112
00:05:42,610 --> 00:05:44,640
我真係鍾意花呢——

113
00:05:45,440 --> 00:05:48,296
為咗教識機械人打「殺出重圍」

114
00:05:48,320 --> 00:05:50,376
我畀佢打幾千局遊戲

115
00:05:50,400 --> 00:05:54,536
不過呢個過程，我又會教佢
辨別好局同劣局

116
00:05:54,536 --> 00:05:58,216
如果我想整隻法律助理機械人

117
00:05:58,240 --> 00:06:00,236
我除咗會教佢法律

118
00:06:00,236 --> 00:06:05,896
我仲會教佢法律嘅寛容同公義

119
00:06:06,560 --> 00:06:09,536
套用科學術語
呢啲係我哋所謂嘅參考標準

120
00:06:09,560 --> 00:06:11,576
而且重要嘅係：

121
00:06:11,600 --> 00:06:13,056
製造呢啲機械人時

122
00:06:13,080 --> 00:06:16,496
我哋將自己嘅價值觀灌輸畀佢

123
00:06:16,520 --> 00:06:19,656
最終，我相信人工智能

124
00:06:19,680 --> 00:06:23,520
會同一個受過專業訓練嘅人一樣

125
00:06:24,080 --> 00:06:25,296
不過，你哋可能會問

126
00:06:25,320 --> 00:06:31,336
如果呢啲人工智能被非法分子利用呢？

127
00:06:31,336 --> 00:06:35,136
雖然我哋唔可能杜絕所有暴力事件發生

128
00:06:35,160 --> 00:06:39,696
但我唔擔心人工智能落入一啲壞人手中

129
00:06:39,720 --> 00:06:44,886
因為人工智能需要持續同細微嘅改進

130
00:06:44,886 --> 00:06:47,296
單憑個人資源唔使旨意做到

131
00:06:47,320 --> 00:06:51,866
而且，唔好當成植入網絡病毒咁簡單

132
00:06:51,866 --> 00:06:54,570
唔好以為隨時隨地㩒個掣

133
00:06:54,570 --> 00:06:57,416
全世界嘅電腦瞬間就會爆炸

134
00:06:57,440 --> 00:07:00,256
人工智能係複雜好多嘅嘢

135
00:07:00,280 --> 00:07:02,515
但我相信遲早有一日人工智能會出現

136
00:07:02,520 --> 00:07:08,266
我驚唔驚人工智能會威脅全人類？

137
00:07:08,280 --> 00:07:12,450
電影《黑客帝國》、《大都會》

138
00:07:12,450 --> 00:07:15,856
《終結者》、電視劇《西部世界》

139
00:07:15,880 --> 00:07:18,016
呢類影視作品都係刻畫呢種恐懼

140
00:07:18,040 --> 00:07:23,926
無錯，哲學家 Nick Bostrom 
喺《超人工智能》一書裏邊都認為

141
00:07:23,926 --> 00:07:27,936
超人工智能唔單止危險

142
00:07:27,936 --> 00:07:31,536
而且仲危及全人類

143
00:07:31,840 --> 00:07:34,056
佢嘅基本論點有︰

144
00:07:34,080 --> 00:07:40,096
咁樣嘅機械人
最終唔會滿足於眼前擁有嘅資訊

145
00:07:40,120 --> 00:07:43,016
機械人可能會因而自己鑽研學習方法

146
00:07:43,040 --> 00:07:47,526
以至到最後發現
自己有啲目標同人類需要有矛盾

147
00:07:48,000 --> 00:07:49,856
有人支持博森博士嘅觀點

148
00:07:49,880 --> 00:07:54,200
包括 Elon Musk 同霍金

149
00:07:54,880 --> 00:08:02,460
我想指出其實幾位智者諗錯咗

150
00:08:02,480 --> 00:08:05,656
Nick Bostrom 嘅理論有好多錯誤

151
00:08:05,680 --> 00:08:07,816
但我無時間講曬所有

152
00:08:07,840 --> 00:08:14,316
但簡單嚟講，可以咁理解：
超智能唔代表超萬能

153
00:08:14,320 --> 00:08:16,576
哈爾對於成個太空探索團隊嘅威脅

154
00:08:16,576 --> 00:08:20,656
僅限於佢可以對探索任務落命令

155
00:08:20,680 --> 00:08:23,176
所以任務需要一個超智能機器

156
00:08:23,200 --> 00:08:25,696
落命令嘅需要有統治世界嘅能力

157
00:08:25,720 --> 00:08:27,520
電影《未來戰士 2018》
裡面嘅 Skynet

158
00:08:27,520 --> 00:08:30,840
嗰個可以操控人類意志嘅
超人工智能防禦系統

159
00:08:30,840 --> 00:08:35,696
控制曬全世界所有嘅機器同裝置

160
00:08:35,720 --> 00:08:37,176
即係話

161
00:08:37,200 --> 00:08:39,296
電影情節係唔會發生

162
00:08:39,320 --> 00:08:43,940
我哋唔係整隻機械人出嚟呼風喚雨

163
00:08:43,940 --> 00:08:47,136
操控喜怒無常、陷於鬥爭嘅人類

164
00:08:47,160 --> 00:08:51,056
再者，如果咁樣嘅機械人真係存在

165
00:08:51,080 --> 00:08:54,016
佢就會同人類嘅經濟鬥過

166
00:08:54,040 --> 00:08:56,880
甚至同人類爭資源

167
00:08:57,200 --> 00:08:58,416
最終結果係——

168
00:08:58,440 --> 00:09:00,430
唔好話俾 Siri 聽——

169
00:09:00,440 --> 00:09:01,826
我哋可以隨時熄咗佢哋

170
00:09:01,826 --> 00:09:03,960
（笑聲）

171
00:09:05,360 --> 00:09:10,226
我哋同機械人係共同進化

172
00:09:10,226 --> 00:09:15,449
未來嘅人類同今日嘅我哋
唔可以同日而語

173
00:09:15,449 --> 00:09:18,910
人類擔心人工智能帶嚟威脅

174
00:09:18,910 --> 00:09:24,290
只會令人類唔去真正關心
科技崛起帶嚟嘅人文同社會問題

175
00:09:24,290 --> 00:09:28,846
而呢啲問題正正係
我哋需要著手解決嘅

176
00:09:29,360 --> 00:09:30,600
問題例如有︰

177
00:09:30,600 --> 00:09:34,556
當我哋唔再需要勞動力嘅時候
我哋要點去調控社會？

178
00:09:34,560 --> 00:09:40,206
點樣向全世界傳播知識同教育
同時又尊重當地嘅差異？

179
00:09:40,206 --> 00:09:44,456
點樣通過認知式保健幫人類延年益壽？

180
00:09:44,480 --> 00:09:49,756
點樣利用電腦幫人類踏足外太空？

181
00:09:49,760 --> 00:09:51,800
諗下都覺得興奮

182
00:09:52,400 --> 00:09:59,706
利用計算科學去開拓人類經歷嘅機會
就喺手裏邊

183
00:09:59,706 --> 00:10:01,996
而我哋只係啱啱捉緊到

184
00:10:02,296 --> 00:10:03,496
多謝大家

185
00:10:03,496 --> 00:10:05,032
（掌聲）