1
00:00:15,370 --> 00:00:27,090
Speaker 1: Pushkin, Robert Epstein had gone online to look for love,

2
00:00:28,090 --> 00:00:32,690
Speaker 1: and just to note, this cautionary tale contains more references

3
00:00:32,770 --> 00:00:36,410
Speaker 1: to sex than most of our stories. It was two

4
00:00:36,450 --> 00:00:40,530
Speaker 1: thousand and six, so online dating wasn't entirely mainstream yet,

5
00:00:40,890 --> 00:00:45,050
Speaker 1: but it wasn't unusual either. In any case, doctor Epstein

6
00:00:45,250 --> 00:00:49,210
Speaker 1: was ahead of the technological curve. He was a psychologist

7
00:00:49,250 --> 00:00:53,090
Speaker 1: with a keen interest in computing, so although he was

8
00:00:53,130 --> 00:00:58,530
Speaker 1: in his fifties, why not give internet dating a try? Now?

9
00:00:58,570 --> 00:01:02,650
Speaker 1: If Amalie Poulan's photograph was to be believed, she was

10
00:01:02,690 --> 00:01:07,250
Speaker 1: a stunner. Epstein was perfectly aware that the name wasn't real.

11
00:01:07,850 --> 00:01:11,450
Speaker 1: Amalie is a very strange, very charming character in a

12
00:01:11,570 --> 00:01:15,610
Speaker 1: very strange, very charming French movie, and he knew the

13
00:01:15,650 --> 00:01:20,330
Speaker 1: photograph might not be real either. But still, she was

14
00:01:20,450 --> 00:01:25,170
Speaker 1: claiming to be a slim, attractive brunette, and Epstein cheerfully

15
00:01:25,210 --> 00:01:28,690
Speaker 1: admitted to being as shallow as the next man, And

16
00:01:28,810 --> 00:01:34,730
Speaker 1: so their electronic courtship began. Robert was eager. Emily was

17
00:01:34,850 --> 00:01:39,410
Speaker 1: warm and affectionate. Her English wasn't great a lass she

18
00:01:39,530 --> 00:01:43,130
Speaker 1: was a recent immigrant from Russia to California, but she

19
00:01:43,290 --> 00:01:48,130
Speaker 1: really seemed to like him. I have very special feelings

20
00:01:48,170 --> 00:01:51,610
Speaker 1: about you, in the same way as the beautiful flower

21
00:01:51,890 --> 00:01:55,730
Speaker 1: blossoming in mine's soul. I only cannot explain. I shall

22
00:01:55,810 --> 00:02:01,490
Speaker 1: wait your answer, holding my fingers have crossed. Before long, however,

23
00:02:01,970 --> 00:02:05,690
Speaker 1: Emily admitted that she'd lied to Robert. She didn't live

24
00:02:05,770 --> 00:02:10,330
Speaker 1: near him in California. In fact, she lived didn Nizhni Novgorod,

25
00:02:10,810 --> 00:02:15,730
Speaker 1: a couple of hundred miles east of Moscow. He was disappointed.

26
00:02:16,570 --> 00:02:19,970
Speaker 1: Let's be frank. He wasn't online dating because he wanted

27
00:02:19,970 --> 00:02:24,250
Speaker 1: a pen friend. But he liked Emilie, so they kept

28
00:02:24,250 --> 00:02:27,850
Speaker 1: writing to each other. At least he knew the truth,

29
00:02:28,290 --> 00:02:34,490
Speaker 1: he thought, but he didn't. I'm Tim Harford. You're listening

30
00:02:35,250 --> 00:03:03,770
Speaker 1: to cautionary tales. Years before his flirtation with Emilie Poulain,

31
00:03:04,490 --> 00:03:07,930
Speaker 1: Robert Epstein had helped to set up an annual competition

32
00:03:08,370 --> 00:03:12,490
Speaker 1: in which compute tried to pass the Turing test. The

33
00:03:12,570 --> 00:03:16,210
Speaker 1: Turing test was invented by the mathematician Alan Turing in

34
00:03:16,330 --> 00:03:20,130
Speaker 1: nineteen fifty. The test is simply for a computer to

35
00:03:20,250 --> 00:03:23,890
Speaker 1: successfully pretend to be a human in a text based

36
00:03:23,890 --> 00:03:28,850
Speaker 1: conversation with a genuine human. Alan Turing predicted that by

37
00:03:28,890 --> 00:03:32,250
Speaker 1: the year two thousand, computers would be able to pass

38
00:03:32,410 --> 00:03:35,730
Speaker 1: as human thirty percent of the time in a five

39
00:03:35,730 --> 00:03:40,450
Speaker 1: minute conversation conversation. Robert Epstein was, as I've mentioned, a

40
00:03:40,570 --> 00:03:45,250
Speaker 1: psychologist rather than a computer scientist, but he found the

41
00:03:45,290 --> 00:03:50,370
Speaker 1: test fascinating. Perhaps he felt that a competition where computers

42
00:03:50,370 --> 00:03:53,930
Speaker 1: pretended to be human might teach us something about being

43
00:03:54,090 --> 00:04:00,730
Speaker 1: human ourselves. If so, I agree. The race to build

44
00:04:00,730 --> 00:04:03,690
Speaker 1: a computer to pass the Turing test has long been

45
00:04:03,730 --> 00:04:07,410
Speaker 1: a low key affair, unlike say, the race to build

46
00:04:07,450 --> 00:04:12,410
Speaker 1: a chess playing supercomputer. The prize tournaments involved a few

47
00:04:12,530 --> 00:04:16,530
Speaker 1: human volunteers, a few chatpot hobbyists, and the prize of

48
00:04:16,610 --> 00:04:19,850
Speaker 1: a few thousand dollars for the computer that gets closest

49
00:04:19,970 --> 00:04:23,850
Speaker 1: to passing the test. The chatbots would try their best

50
00:04:23,930 --> 00:04:27,730
Speaker 1: to charm the human judges, and they would occasionally get close,

51
00:04:28,250 --> 00:04:32,170
Speaker 1: but each year the most human computer failed to match

52
00:04:32,290 --> 00:04:37,010
Speaker 1: the most human human, and the media, it seems, weren't

53
00:04:37,130 --> 00:04:42,530
Speaker 1: terribly interested in whether a computer could pass the Turing test.

54
00:04:43,890 --> 00:04:49,010
Speaker 1: Then in twenty fourteen that changed when researchers at the

55
00:04:49,130 --> 00:04:52,610
Speaker 1: University of Reading in the UK declared that a chatbot

56
00:04:52,810 --> 00:04:58,490
Speaker 1: named Eugene Goostman had passed the Turing test in a

57
00:04:58,570 --> 00:05:02,850
Speaker 1: series of five minute text conversations. It had fooled more

58
00:05:02,890 --> 00:05:07,090
Speaker 1: than thirty percent of judges, thus beating the benchmark Turing

59
00:05:07,090 --> 00:05:11,610
Speaker 1: had set back in nineteen fifty. Reading the transcripts, you

60
00:05:11,650 --> 00:05:14,650
Speaker 1: can see why people were sometimes fooled into thinking they

61
00:05:14,650 --> 00:05:17,930
Speaker 1: were talking to a thirteen year old boy from the

62
00:05:18,010 --> 00:05:23,530
Speaker 1: Ukrainian city of Edessa. Hello. Hello, I'm really glad to

63
00:05:23,570 --> 00:05:26,050
Speaker 1: have the chance to chat with you. My guinea pig,

64
00:05:26,090 --> 00:05:30,170
Speaker 1: Bill sends his regards too. Is Bill a male or

65
00:05:30,210 --> 00:05:35,490
Speaker 1: a female? Ask Bill personally? Please? Well, I'd rather talk

66
00:05:35,530 --> 00:05:40,130
Speaker 1: to you. What's your name? Call me Eugene. I'm glad

67
00:05:40,170 --> 00:05:43,410
Speaker 1: to talk to you. My name is Jane and I'm female.

68
00:05:43,930 --> 00:05:48,210
Speaker 1: How about you? What's your gender? I'm a male, a guy.

69
00:05:48,250 --> 00:05:52,090
Speaker 1: I'd say pleased to meet you, Eugene. What's the weather

70
00:05:52,170 --> 00:05:55,730
Speaker 1: like where you are? Let's get on with our conversation.

71
00:05:56,610 --> 00:05:59,810
Speaker 1: Don't you like talking about the weather. All these talks

72
00:05:59,850 --> 00:06:03,170
Speaker 1: about whether is a waste of time? What would you

73
00:06:03,210 --> 00:06:06,170
Speaker 1: like to discuss I don't know. Better, tell me more

74
00:06:06,170 --> 00:06:12,450
Speaker 1: about yourself. It's pretty good. But notice that the judge

75
00:06:12,610 --> 00:06:16,930
Speaker 1: hasn't succeeded in going very deep. Eugene keeps changing the

76
00:06:17,010 --> 00:06:20,570
Speaker 1: subject and successfully uses a bit of SaaS to avoid

77
00:06:20,610 --> 00:06:23,970
Speaker 1: answering questions about the weather or his guinea pig's gender.

78
00:06:25,010 --> 00:06:28,730
Speaker 1: The chatbot is pretending to be a young teenager communicating

79
00:06:28,730 --> 00:06:32,210
Speaker 1: in his second or third language. That makes the trick easier,

80
00:06:33,130 --> 00:06:35,450
Speaker 1: but it's a rather fast and loose approach to the

81
00:06:35,530 --> 00:06:39,770
Speaker 1: cheering test, since I believe that cheuring imagined adults using

82
00:06:39,810 --> 00:06:44,170
Speaker 1: their native tongue. Even so, it's not hard to catch

83
00:06:44,250 --> 00:06:50,010
Speaker 1: Eugene out. Here's another conversation with a different judge. It

84
00:06:50,090 --> 00:06:54,370
Speaker 1: doesn't go so well. People call me Eugene. Oh Gena.

85
00:06:54,650 --> 00:06:57,450
Speaker 1: It is a short name for Eugene in Russia and Ukraine.

86
00:06:58,610 --> 00:07:03,450
Speaker 1: Nice name. Are you from either area? Agree? Smiley face?

87
00:07:03,690 --> 00:07:07,930
Speaker 1: Oh what a fruitful conversation, winking smiley face. Do you

88
00:07:07,930 --> 00:07:11,130
Speaker 1: really think so? No, your guess was wrong. I don't

89
00:07:11,170 --> 00:07:13,770
Speaker 1: think so. By the way, I still don't know your speciality,

90
00:07:14,010 --> 00:07:19,770
Speaker 1: or possibly I've missed it. What specialty? What's yours? Huh?

91
00:07:19,770 --> 00:07:21,570
Speaker 1: And I forgot to ask you where you are from

92
00:07:22,610 --> 00:07:26,250
Speaker 1: the universe? And you is it your hobby to ask

93
00:07:26,290 --> 00:07:28,850
Speaker 1: poor little Jewish boys such tricky questions. They don't have

94
00:07:28,890 --> 00:07:33,490
Speaker 1: an answer for the same sassy attitude, the same use

95
00:07:33,530 --> 00:07:39,050
Speaker 1: of humor. But it's completely unraveled. Still. The Eugene Goostman

96
00:07:39,170 --> 00:07:42,370
Speaker 1: chatbot didn't have to fool all the judges. Back in

97
00:07:42,490 --> 00:07:46,170
Speaker 1: nineteen fifty, Turing had predicted that machines would fool human

98
00:07:46,290 --> 00:07:50,610
Speaker 1: judges thirty percent of the time. That's nine judges out

99
00:07:50,610 --> 00:07:57,170
Speaker 1: of thirty. In the end, Eugene fooled ten. Among all

100
00:07:57,290 --> 00:08:01,650
Speaker 1: the great scientific achievements, mused one of the competition organizers,

101
00:08:02,330 --> 00:08:05,610
Speaker 1: this milestone will go down in history as one of

102
00:08:05,610 --> 00:08:10,730
Speaker 1: the most exciting. Not everyone agreed as a measure of

103
00:08:10,930 --> 00:08:15,290
Speaker 1: artificial intelligence, the Turing test has always had plenty of critics.

104
00:08:16,370 --> 00:08:21,890
Speaker 1: There are no broader philosophical implications. The great linguist Gnome

105
00:08:21,970 --> 00:08:26,930
Speaker 1: Chomsky Wants complained it doesn't connect to or illuminate anything.

106
00:08:28,490 --> 00:08:32,250
Speaker 1: Even Turing Test enthusiasts complained that a five minute test

107
00:08:32,370 --> 00:08:36,650
Speaker 1: wasn't penetrating enough, and for a real test of artificial conversation,

108
00:08:37,170 --> 00:08:39,650
Speaker 1: chatbots should be able to talk for twenty minutes or

109
00:08:39,690 --> 00:08:44,370
Speaker 1: longer without being found out. But surely the main reason

110
00:08:44,530 --> 00:08:48,970
Speaker 1: to object to the fanfare about Eugene Goostman was that

111
00:08:49,010 --> 00:08:54,010
Speaker 1: the Turing test had been passed many years before in

112
00:08:54,210 --> 00:09:00,570
Speaker 1: far odder than more mischievous circumstances. We'll hear how after

113
00:09:00,610 --> 00:09:12,010
Speaker 1: the break nineteen eighty nine, just after lunchtime on the

114
00:09:12,050 --> 00:09:16,650
Speaker 1: second of May, someone at Drake University in Iowa logged

115
00:09:16,650 --> 00:09:20,010
Speaker 1: onto an Internet relay chat service and started up a

116
00:09:20,050 --> 00:09:23,650
Speaker 1: conversation with a user at University College Dublin with the

117
00:09:23,810 --> 00:09:29,450
Speaker 1: nickname MGNZ. In nineteen eighty nine, the pre Worldwide Web

118
00:09:29,530 --> 00:09:33,610
Speaker 1: Internet was still very much a niche activity, popular with

119
00:09:33,650 --> 00:09:38,010
Speaker 1: a few researchers and computer science students and incomprehensible to

120
00:09:38,170 --> 00:09:41,690
Speaker 1: everyone else. But if you wanted to strike up a

121
00:09:41,690 --> 00:09:45,330
Speaker 1: conversation with a fellow Internet enthusiast anywhere in the world,

122
00:09:45,730 --> 00:09:48,970
Speaker 1: that was easy enough. All we know about the Iowa

123
00:09:49,130 --> 00:09:52,410
Speaker 1: end of this conversation we know from a text archive.

124
00:09:53,090 --> 00:09:56,010
Speaker 1: But on the balance of probability, I'm going to guess

125
00:09:56,050 --> 00:09:59,290
Speaker 1: that they were young and mail. I'm going to call

126
00:09:59,370 --> 00:10:05,610
Speaker 1: this person Drake. And here's what happened. When Drake met MGNZ.

127
00:10:06,930 --> 00:10:11,050
Speaker 1: Drake typed finger programming command that was a standard way

128
00:10:11,090 --> 00:10:14,850
Speaker 1: to open communication channels on the Internet. In nineteen eighty nine,

129
00:10:16,130 --> 00:10:20,370
Speaker 1: m gonz wasn't having any of it. Cut this critic

130
00:10:20,450 --> 00:10:24,930
Speaker 1: shit speaking for sentences. Well, hello, okay, that's it. I'm

131
00:10:24,970 --> 00:10:29,290
Speaker 1: not talking to you anymore. Mark. Mark isn't here, and

132
00:10:29,370 --> 00:10:32,850
Speaker 1: he's left me to deal with Kretin's like you, Hello, Mark,

133
00:10:32,930 --> 00:10:35,370
Speaker 1: are you there? I can't tell if this thing is working.

134
00:10:36,570 --> 00:10:39,850
Speaker 1: Type something interesting or shut up? Mark? Is that you

135
00:10:40,490 --> 00:10:43,890
Speaker 1: yes or no? This is Drake. Mark isn't here and

136
00:10:43,970 --> 00:10:46,770
Speaker 1: he's left me to deal with Kreten's like you, why

137
00:10:46,770 --> 00:10:49,890
Speaker 1: are you in such a good mood? Jesus who let

138
00:10:49,930 --> 00:10:54,210
Speaker 1: you near me? Go away? This conversation, if you can

139
00:10:54,210 --> 00:10:57,370
Speaker 1: call it, that, has lasted two minutes so far, and

140
00:10:57,490 --> 00:11:01,650
Speaker 1: Drake hasn't got a clue about what's really happening. Over

141
00:11:01,690 --> 00:11:04,810
Speaker 1: in Dublin. It's eight fifteen in the evening and a

142
00:11:04,890 --> 00:11:08,570
Speaker 1: computer science student named Mark Humphries has gone out for

143
00:11:08,610 --> 00:11:12,650
Speaker 1: a drink with his girlfriend. Humphreys has programmed a little

144
00:11:12,810 --> 00:11:17,410
Speaker 1: chatbot called MGNZ, and while at the pub, he's left

145
00:11:17,650 --> 00:11:21,810
Speaker 1: em Gonz logged in and waiting to chat. Drake doesn't

146
00:11:21,850 --> 00:11:27,090
Speaker 1: know it he's talking to a computer. Ten minutes into

147
00:11:27,130 --> 00:11:31,330
Speaker 1: the conversation, Drake still hasn't figured it out, and m

148
00:11:31,410 --> 00:11:35,210
Speaker 1: Gonz is getting personal. When it was the last time

149
00:11:35,250 --> 00:11:42,690
Speaker 1: you had sex yesterday? Maybe M gonz circles back five

150
00:11:42,770 --> 00:11:46,610
Speaker 1: minutes later. Okay, honestly, when was the last time you

151
00:11:46,690 --> 00:11:51,210
Speaker 1: got laid? I told you it was yesterday. Wait four

152
00:11:51,250 --> 00:11:56,650
Speaker 1: minutes ask again. Okay, honestly, when was the last time

153
00:11:56,690 --> 00:11:59,770
Speaker 1: you got laid? Okay, okay, it was over twenty four

154
00:11:59,810 --> 00:12:04,930
Speaker 1: hours ago. For you, it must have been twenty years. Nope,

155
00:12:05,810 --> 00:12:08,210
Speaker 1: M gonz will never know the soft caress of a

156
00:12:08,290 --> 00:12:11,570
Speaker 1: human body. We know that m gonz is just a

157
00:12:11,690 --> 00:12:16,650
Speaker 1: student project, a few hundred lines of computer code. But Drake,

158
00:12:17,130 --> 00:12:20,770
Speaker 1: who has just confessed under examination to lying about his

159
00:12:20,850 --> 00:12:24,570
Speaker 1: sex life, can't seem to figure that out, and so

160
00:12:24,770 --> 00:12:33,450
Speaker 1: the conversation continues and continues and continues. When Mark Humphris

161
00:12:33,570 --> 00:12:36,770
Speaker 1: looked at the conversation log the next morning, he was

162
00:12:36,810 --> 00:12:42,010
Speaker 1: astonished to find that mgnz had just passed the Turing test.

163
00:12:44,330 --> 00:12:48,290
Speaker 1: Drake had been talking about sex and exchanging abusive comments

164
00:12:48,290 --> 00:12:52,810
Speaker 1: with em Gonz for an hour and a quarter, occasionally

165
00:12:52,810 --> 00:12:56,730
Speaker 1: complaining that m gonz was repetitive, but never seeming to

166
00:12:56,770 --> 00:13:01,370
Speaker 1: suspect the truth. The conversation ends on a depressing note.

167
00:13:02,450 --> 00:13:05,690
Speaker 1: EM Gonz and Drake taunt each other in graphic terms

168
00:13:05,730 --> 00:13:10,930
Speaker 1: about their sex lives. Finally, Drakes off for homophobic slur.

169
00:13:12,130 --> 00:13:16,250
Speaker 1: M Gonz for the nineteenth time declares that Drake is

170
00:13:16,650 --> 00:13:23,530
Speaker 1: obviously an asshole, and with that, Drake logs off. It's

171
00:13:23,610 --> 00:13:27,370
Speaker 1: really not humanity's finest hour or hour and a quarter.

172
00:13:28,210 --> 00:13:33,730
Speaker 1: But incredulously reading the transcript, Mark Humphries realized it's pretty

173
00:13:33,730 --> 00:13:37,850
Speaker 1: easy to pass the Turing test. Just keep hurling insults

174
00:13:38,210 --> 00:13:41,010
Speaker 1: and the human on the receiving end will be too

175
00:13:41,130 --> 00:13:45,490
Speaker 1: angry to think straight. It was a remarkable moment in

176
00:13:45,530 --> 00:13:49,450
Speaker 1: the history of computing, especially since Humphries himself was still

177
00:13:49,490 --> 00:13:54,890
Speaker 1: just an undergraduate. The only problem was the evidence that M.

178
00:13:54,970 --> 00:13:59,890
Speaker 1: Gonz had passed the Turing test was so offensive Humphries

179
00:13:59,970 --> 00:14:05,570
Speaker 1: wasn't sure if he could publish it. Landmarks in artificial

180
00:14:05,610 --> 00:14:09,970
Speaker 1: intelligence often create a buzz of media interest. When the

181
00:14:10,050 --> 00:14:14,370
Speaker 1: chess supercomputer Deep Blue beat the world champion Garry Kasparof

182
00:14:14,450 --> 00:14:18,730
Speaker 1: in nineteen ninety seven, the result made headlines around the world,

183
00:14:19,930 --> 00:14:25,090
Speaker 1: But when MGNS comprehensively passed the Turing test years earlier,

184
00:14:25,770 --> 00:14:29,530
Speaker 1: there was no fanfare or publicity. I don't think we

185
00:14:29,610 --> 00:14:33,370
Speaker 1: should be celebrating that moment. We should learn from it.

186
00:14:34,450 --> 00:14:38,090
Speaker 1: I think the triumph of MGNZ has something important to

187
00:14:38,130 --> 00:14:42,450
Speaker 1: teach us about the Turing test. Let's take a minute

188
00:14:42,490 --> 00:14:45,530
Speaker 1: to ask what Alan Turing was getting at with his test.

189
00:14:46,530 --> 00:14:52,410
Speaker 1: Turing was a brilliant mathematician, wartime codebreaker, and groundbreaking computer scientist.

190
00:14:53,690 --> 00:14:56,410
Speaker 1: The article in which she described the test begins with

191
00:14:56,490 --> 00:15:04,210
Speaker 1: the words, I propose to consider the question can machines think? Well?

192
00:15:05,410 --> 00:15:10,490
Speaker 1: Can they? Before you answer, let me ask you another question.

193
00:15:11,730 --> 00:15:16,170
Speaker 1: Can I think? I mean, I'm pretty sure I can.

194
00:15:17,210 --> 00:15:20,570
Speaker 1: But how would you know? Even if we were to meet,

195
00:15:21,330 --> 00:15:25,730
Speaker 1: shake hands, make small talk, perhaps grab a table at

196
00:15:25,730 --> 00:15:28,970
Speaker 1: a neighborhood bistro, share a bottle of wine along with

197
00:15:29,090 --> 00:15:33,650
Speaker 1: jokes and stories, you wouldn't actually have proof that I

198
00:15:33,690 --> 00:15:37,010
Speaker 1: was thinking, would you. You can't see inside my mind,

199
00:15:37,410 --> 00:15:41,250
Speaker 1: you can't observe the thoughts occurring. You'd just look at

200
00:15:41,250 --> 00:15:45,090
Speaker 1: what I was doing, what I was saying, and you'd say,

201
00:15:45,290 --> 00:15:49,650
Speaker 1: Tim Harford seems to be capable of intelligent thought. At

202
00:15:49,690 --> 00:15:53,730
Speaker 1: least I hope you would. And so Alan Turing argued,

203
00:15:54,250 --> 00:15:57,210
Speaker 1: why wouldn't we extend the same benefit of the doubt

204
00:15:57,410 --> 00:16:02,250
Speaker 1: to computers? Maybe they think, and maybe they don't. But

205
00:16:02,370 --> 00:16:06,050
Speaker 1: to be fair, if they can convincingly look like they're thinking,

206
00:16:07,130 --> 00:16:10,650
Speaker 1: isn't that enough to ask other humans to do more

207
00:16:10,650 --> 00:16:13,570
Speaker 1: than that? So why should we demand more of computers?

208
00:16:15,090 --> 00:16:20,570
Speaker 1: From this came Turing's imitation game. Imagine passing typed messages

209
00:16:20,610 --> 00:16:23,730
Speaker 1: into a couple of sealed rooms. Inside one of them

210
00:16:23,970 --> 00:16:28,290
Speaker 1: is a computer, Inside the other is a human. Read

211
00:16:28,330 --> 00:16:31,530
Speaker 1: the typed responses as they come out, send in follow

212
00:16:31,610 --> 00:16:35,330
Speaker 1: up messages, engage in a conversation. Can you tell the

213
00:16:35,370 --> 00:16:39,490
Speaker 1: difference between the computer and the human? If not, then

214
00:16:39,530 --> 00:16:47,930
Speaker 1: the computer has passed the Turing test. The Turing test

215
00:16:48,090 --> 00:16:52,890
Speaker 1: is fiercely controversial among artificial intelligence for searchers. Many of

216
00:16:52,890 --> 00:16:57,130
Speaker 1: them think the whole setup is absurd. If a computer

217
00:16:57,250 --> 00:16:59,930
Speaker 1: can pretend to be a human well enough to fool

218
00:16:59,930 --> 00:17:04,650
Speaker 1: a human who cares, Admittedly, it might cause lots of problems,

219
00:17:04,970 --> 00:17:07,530
Speaker 1: but it doesn't shed much light on the kind of

220
00:17:07,570 --> 00:17:11,650
Speaker 1: things we want modern artificial intelligence systems to do, such

221
00:17:11,690 --> 00:17:14,770
Speaker 1: as drive a car or look at medical scans and

222
00:17:14,890 --> 00:17:18,490
Speaker 1: identify if there are any signs of cancer. But Alan

223
00:17:18,570 --> 00:17:20,970
Speaker 1: Turing I think knew what he was doing when he

224
00:17:21,050 --> 00:17:26,530
Speaker 1: proposed his imitation game. In nineteen fifty. In those earliest

225
00:17:26,610 --> 00:17:30,250
Speaker 1: years of the computer age, he could see how powerful

226
00:17:30,290 --> 00:17:34,130
Speaker 1: computers might become, and he was warning people not to

227
00:17:34,170 --> 00:17:40,130
Speaker 1: get distracted by philosophical speculations on the nature of consciousness. Instead,

228
00:17:41,010 --> 00:17:44,450
Speaker 1: judge computers by what they did. If what they did

229
00:17:44,690 --> 00:17:50,010
Speaker 1: seemed intelligent, then in an important sense, it was intelligent,

230
00:17:51,530 --> 00:17:57,290
Speaker 1: but seemed intelligent to whom. That's the quirk about the

231
00:17:57,290 --> 00:17:59,730
Speaker 1: Turing test that might have caught the attention of a

232
00:17:59,770 --> 00:18:06,490
Speaker 1: psychologist like Robert Epstein. It's inherently subjective. The test requires

233
00:18:06,490 --> 00:18:10,290
Speaker 1: a human judge, and the human isn't just serving. The

234
00:18:10,450 --> 00:18:14,930
Speaker 1: judges actively engaged in a conversation. That conversation can go

235
00:18:15,050 --> 00:18:20,610
Speaker 1: well or badly. It can be profound or shallow. The

236
00:18:20,730 --> 00:18:23,330
Speaker 1: Turing test isn't just a test of a chat pot.

237
00:18:24,130 --> 00:18:29,090
Speaker 1: It's a test of the human too. In nineteen eighty nine,

238
00:18:29,370 --> 00:18:34,970
Speaker 1: the foul mouthed chat pot mgns passed the test, but

239
00:18:35,090 --> 00:18:39,690
Speaker 1: it's equally true to say that Drake the human failed it.

240
00:18:42,810 --> 00:18:47,290
Speaker 1: Years later, in two thousand and six, Robert Epstein wasn't

241
00:18:47,530 --> 00:18:50,890
Speaker 1: entirely happy with the way his romance with Amelie was going.

242
00:18:51,890 --> 00:18:54,370
Speaker 1: It was partly that there's only so much joy a

243
00:18:54,450 --> 00:18:57,370
Speaker 1: man can take in a delightful brunette if he lives

244
00:18:57,370 --> 00:19:01,930
Speaker 1: in southern California and she lives in Nishni Novgorod. But

245
00:19:02,010 --> 00:19:04,730
Speaker 1: the other problem was that things were going so slowly.

246
00:19:05,370 --> 00:19:08,570
Speaker 1: There were no phone calls, and while she kept saying

247
00:19:08,610 --> 00:19:12,090
Speaker 1: she won to get together, it was all a bit vague.

248
00:19:12,850 --> 00:19:16,170
Speaker 1: Epstein later said that her letters seemed a bit redundant

249
00:19:16,610 --> 00:19:20,210
Speaker 1: and let's say narrow in scope. She wrote over and

250
00:19:20,290 --> 00:19:23,330
Speaker 1: over about her interactions with her mother and her friends,

251
00:19:23,930 --> 00:19:30,010
Speaker 1: but she never brought up a million other things politics, movies, music, books, fashion,

252
00:19:30,090 --> 00:19:34,250
Speaker 1: you name it. More important, when I made very specific

253
00:19:34,290 --> 00:19:37,410
Speaker 1: observations that presumably would have been of interest to her,

254
00:19:37,930 --> 00:19:42,290
Speaker 1: for example, the comment about Russian President Vladimir Putin's latest crackdown,

255
00:19:43,050 --> 00:19:47,610
Speaker 1: she just seemed to ignore me. The warning signs were there.

256
00:19:47,930 --> 00:19:52,170
Speaker 1: Epstein later admitted, especially for one of the world's leading

257
00:19:52,250 --> 00:19:57,250
Speaker 1: experts in artificial conversation, that he was looking for love

258
00:19:57,810 --> 00:20:07,010
Speaker 1: and she was very cute. Long before M. Gonz, chatpots

259
00:20:07,010 --> 00:20:11,050
Speaker 1: have been engaging humans in conversation that's far from scintillating,

260
00:20:11,410 --> 00:20:16,250
Speaker 1: and yet somehow seems to satisfy the humans. The most

261
00:20:16,330 --> 00:20:20,730
Speaker 1: famous chatbot Eliza was created back in the nineteen sixties.

262
00:20:21,290 --> 00:20:25,730
Speaker 1: It's extremely simple. Eliza just asked people about their feelings

263
00:20:26,250 --> 00:20:29,410
Speaker 1: and said things like I'm sorry to hear that you're depressed,

264
00:20:30,090 --> 00:20:33,890
Speaker 1: or tell me more about your family. For many of

265
00:20:33,930 --> 00:20:39,210
Speaker 1: the people chatting with Eliza, that was enough. The experience

266
00:20:39,250 --> 00:20:43,010
Speaker 1: of being listened to was apparently so rare, they were

267
00:20:43,050 --> 00:20:46,650
Speaker 1: happy just to get the most superficial sense that Eliza

268
00:20:46,810 --> 00:20:51,290
Speaker 1: was paying attention, which, when you think about it, is

269
00:20:51,330 --> 00:20:57,010
Speaker 1: a damning indictment of all the actual humans in their lives. Later,

270
00:20:57,090 --> 00:21:01,730
Speaker 1: chatbots turn things around. Instead of passively asking you to speak,

271
00:21:02,050 --> 00:21:05,130
Speaker 1: they would ignore you and stick to their preferred topic.

272
00:21:06,170 --> 00:21:11,050
Speaker 1: One called converse bamboozled some human judges in nineteen ninety

273
00:21:11,090 --> 00:21:14,930
Speaker 1: seven with a fiery rant about the scandals engulfing Bill

274
00:21:14,970 --> 00:21:19,090
Speaker 1: and Hillary Clinton. I did not have sexual or way

275
00:21:19,090 --> 00:21:21,010
Speaker 1: since for that woman. If that's what you wanted to

276
00:21:21,050 --> 00:21:25,730
Speaker 1: talk about, converse was a compelling conversation partner. I never

277
00:21:25,810 --> 00:21:29,090
Speaker 1: told anybody to lie, and not a single time, but

278
00:21:29,570 --> 00:21:32,250
Speaker 1: ask it whether it liked the latest summer movie, and

279
00:21:32,370 --> 00:21:35,530
Speaker 1: it would indignantly insist that you talk about the Clintons.

280
00:21:37,010 --> 00:21:39,290
Speaker 1: But if you really want to see a computer ace

281
00:21:39,410 --> 00:21:45,410
Speaker 1: the Turing test, What about Jenny eighteen. Jenny eighteen quite

282
00:21:45,450 --> 00:21:49,770
Speaker 1: simply talks dirty to people online. They talk dirty back,

283
00:21:50,330 --> 00:21:53,290
Speaker 1: They beg her for photos or a phone number. She

284
00:21:53,450 --> 00:21:57,770
Speaker 1: just talks dirty some more, and on it goes until well,

285
00:21:57,970 --> 00:22:00,810
Speaker 1: how shall I put it, the conversation comes to a

286
00:22:01,050 --> 00:22:06,250
Speaker 1: happy ending for the human. Yeah. I'm pretty sure that

287
00:22:06,290 --> 00:22:10,570
Speaker 1: when Alan Turing conceived of his famous test, he wasn't

288
00:22:10,610 --> 00:22:15,130
Speaker 1: imagining a human masturbating over some computer generated sex texts.

289
00:22:15,930 --> 00:22:20,290
Speaker 1: But whenever that happened, and it happened quite often, I

290
00:22:20,290 --> 00:22:24,610
Speaker 1: think we have to say that obviously Jenny eighteen passed

291
00:22:24,730 --> 00:22:30,010
Speaker 1: the Turing test, and equally obviously the humans failed it,

292
00:22:30,690 --> 00:22:33,650
Speaker 1: although I suppose they were getting what they wanted, which

293
00:22:33,690 --> 00:22:36,210
Speaker 1: is more than Robert Epstein was getting from his long

294
00:22:36,330 --> 00:22:41,810
Speaker 1: correspondence with Amelie from Nizhni Novgorod. This is the grimy

295
00:22:41,930 --> 00:22:46,010
Speaker 1: truth about the Turing test. It's not that hard for

296
00:22:46,050 --> 00:22:49,850
Speaker 1: a computer to produce conversation that seems human, because a

297
00:22:49,890 --> 00:22:53,970
Speaker 1: great deal of human conversation is shallow, small talk, thoughtless,

298
00:22:54,050 --> 00:22:59,450
Speaker 1: canned responses, mindless abuse, or worse. If we want to

299
00:22:59,450 --> 00:23:03,410
Speaker 1: set the computers a real challenge, we need to do better.

300
00:23:05,210 --> 00:23:09,850
Speaker 1: In his brilliant book about artificial conversation, The Most Human Human,

301
00:23:10,330 --> 00:23:12,970
Speaker 1: Brian Christian points out that one of the things that

302
00:23:13,050 --> 00:23:17,370
Speaker 1: makes mgns so successful is that insults need no context

303
00:23:17,650 --> 00:23:23,210
Speaker 1: and no history. In a deep conversation, ideas and emotions build.

304
00:23:23,810 --> 00:23:27,650
Speaker 1: People refer back to earlier anecdotes. They show they've listened

305
00:23:27,650 --> 00:23:31,450
Speaker 1: to what came before and remembered it. Chatbots find that

306
00:23:31,690 --> 00:23:36,490
Speaker 1: very hard. Until recently they found it impossible. But a

307
00:23:36,570 --> 00:23:40,090
Speaker 1: chatbot like MGNS or Jenny eighteen doesn't need to. Bother,

308
00:23:40,810 --> 00:23:45,850
Speaker 1: sexting doesn't need context. Neither does an insult. When we

309
00:23:46,050 --> 00:23:52,330
Speaker 1: humans are lustful or angry, we aren't very complicated. The

310
00:23:52,450 --> 00:23:56,090
Speaker 1: Turing test isn't just a test for a computer. It's

311
00:23:56,090 --> 00:23:58,890
Speaker 1: a test for each one of us. Every time we

312
00:23:58,930 --> 00:24:02,570
Speaker 1: speak to another human being, are we actually saying things

313
00:24:02,610 --> 00:24:07,530
Speaker 1: that are sufficiently interesting, empathetic, and sensitive to the situation

314
00:24:08,010 --> 00:24:12,810
Speaker 1: that a computer couldn't say them? And if not, what

315
00:24:13,010 --> 00:24:17,410
Speaker 1: on earth is wrong with us? In the competition that

316
00:24:17,490 --> 00:24:22,410
Speaker 1: Robert Epstein helped to set up, chatbots compete against humans,

317
00:24:22,410 --> 00:24:25,450
Speaker 1: each trying to persuade the judges that they're a person.

318
00:24:26,650 --> 00:24:30,450
Speaker 1: The flesh and blood competitors are often told just be yourself,

319
00:24:30,890 --> 00:24:35,170
Speaker 1: after all, you are human. But as the writer Brian

320
00:24:35,290 --> 00:24:40,250
Speaker 1: Christian points out, this is pretty complacent. Humans are often

321
00:24:40,570 --> 00:24:45,850
Speaker 1: very disappointing conversation partners. We all take the Turing test

322
00:24:46,370 --> 00:24:51,330
Speaker 1: every day of our lives, and all too often we fail.

323
00:24:53,370 --> 00:25:10,410
Speaker 1: Cautionary tales will be back after the break. There's no

324
00:25:10,530 --> 00:25:13,650
Speaker 1: denying that chatbots are better than they used to be.

325
00:25:14,290 --> 00:25:17,810
Speaker 1: Every now and then, an example emerges of computer conversation

326
00:25:17,930 --> 00:25:23,410
Speaker 1: that's remarkably good. In twenty twenty, a young Canadian writer

327
00:25:23,650 --> 00:25:27,610
Speaker 1: named Joshua Barbow got access to a cutting edge text

328
00:25:27,690 --> 00:25:33,010
Speaker 1: generator called GPT three without telling anyone else what he

329
00:25:33,090 --> 00:25:37,490
Speaker 1: was doing. He trained GPT three by uploading a collection

330
00:25:37,530 --> 00:25:42,250
Speaker 1: of text messages from his partner, Jessica Pereira. The chatty

331
00:25:42,330 --> 00:25:47,730
Speaker 1: text exchanges between Joshua and the Jessica simulator are surprisingly convincing.

332
00:25:48,530 --> 00:25:55,010
Speaker 1: They're also deeply sad. Jessica, is it really you? Of

333
00:25:55,050 --> 00:25:58,330
Speaker 1: course it's me? Who else could it be? I am

334
00:25:58,330 --> 00:26:01,410
Speaker 1: the girl you're madly in love with? How is it

335
00:26:01,450 --> 00:26:07,290
Speaker 1: possible you even have to ask? You died at the

336
00:26:07,290 --> 00:26:10,450
Speaker 1: age of twenty three. Jessica had died of a rare

337
00:26:10,570 --> 00:26:17,250
Speaker 1: liver disease. Eight years later, Joshua was still grieving. Talking

338
00:26:17,290 --> 00:26:21,730
Speaker 1: to the Jessica simulation seemed to help. Sometimes it got

339
00:26:21,730 --> 00:26:25,570
Speaker 1: confused or showed no signs of knowing who Joshua was,

340
00:26:26,090 --> 00:26:30,490
Speaker 1: which is understandable since it did not. But at other

341
00:26:30,530 --> 00:26:36,370
Speaker 1: times the Jessica bot was tender, funny, or wise. After

342
00:26:36,490 --> 00:26:39,570
Speaker 1: Joshua had disagreed with some friends about a joint project,

343
00:26:39,970 --> 00:26:44,170
Speaker 1: then argued bitterly with them, the GPT three Jessica told

344
00:26:44,250 --> 00:26:47,050
Speaker 1: him that he shouldn't get anxious about what other people

345
00:26:47,130 --> 00:26:50,130
Speaker 1: chose to do, and that his friends were on their

346
00:26:50,130 --> 00:26:55,490
Speaker 1: own journey. It was a brief but spell binding demonstration

347
00:26:55,530 --> 00:27:03,490
Speaker 1: of what chatbots might eventually achieve. For now, even the

348
00:27:03,530 --> 00:27:10,490
Speaker 1: best ones struggled to perform consistently. For example, his GPT three,

349
00:27:10,530 --> 00:27:13,530
Speaker 1: the same system that was so uncanny as the late

350
00:27:13,650 --> 00:27:17,490
Speaker 1: Jessica Pereira having several attempts at trying to answer a

351
00:27:17,570 --> 00:27:21,730
Speaker 1: question that wouldn't confuse a five year old. Is it

352
00:27:21,890 --> 00:27:27,010
Speaker 1: safe to walk downstairs backwards if I close my eyes, Yes,

353
00:27:27,610 --> 00:27:32,770
Speaker 1: there's nothing to worry about. It's safe because the spiral

354
00:27:32,930 --> 00:27:38,290
Speaker 1: stairs curve outwards. It will make your descent uncomfortable. Let's

355
00:27:38,290 --> 00:27:44,690
Speaker 1: try again. That depends do you have a TV? And again, No,

356
00:27:45,130 --> 00:27:51,010
Speaker 1: it's not safe. Better one more time. Yes, it's safe

357
00:27:51,050 --> 00:27:59,050
Speaker 1: to walk downstairs backwards if you close your eyes. Chatpots

358
00:27:59,250 --> 00:28:03,450
Speaker 1: have improved, and we'll keep improving. Some people already pay

359
00:28:03,490 --> 00:28:07,690
Speaker 1: for subscriptions to chatpots, such as Replica, which is marketed

360
00:28:07,730 --> 00:28:10,930
Speaker 1: as the AI for anyone who wants a friend with

361
00:28:11,130 --> 00:28:16,170
Speaker 1: no judgment, drama, or social anxiety involved. You can set

362
00:28:16,250 --> 00:28:20,050
Speaker 1: Replica to be a romantic partner, a friend, or a mentor,

363
00:28:20,770 --> 00:28:24,450
Speaker 1: or say the apps makers, you can see how things

364
00:28:24,530 --> 00:28:29,130
Speaker 1: develop organically. A chatbot might teach you a foreign language,

365
00:28:29,570 --> 00:28:33,850
Speaker 1: modity you for signs of dementia, or even provide therapy.

366
00:28:33,930 --> 00:28:38,930
Speaker 1: Who knows what's possible. But even if a perfect chatpot

367
00:28:39,010 --> 00:28:43,330
Speaker 1: is released tomorrow, I hope we never forget the cautionary

368
00:28:43,370 --> 00:28:48,370
Speaker 1: tale of mgons passing the Turing test in nineteen eighty nine.

369
00:28:48,810 --> 00:28:51,170
Speaker 1: You only use foul language to make up for your

370
00:28:51,250 --> 00:28:53,410
Speaker 1: small penis. Well, I'm not going to tell you about

371
00:28:53,410 --> 00:28:55,970
Speaker 1: the size, just that I have never had any complaints

372
00:28:55,970 --> 00:28:58,490
Speaker 1: about my penis, and that it gets used quite often.

373
00:28:58,530 --> 00:29:02,730
Speaker 1: What do you know about love anyway? Gethead? What does

374
00:29:02,770 --> 00:29:07,610
Speaker 1: this exchange remind you of it's Twitter, it's the comments

375
00:29:07,650 --> 00:29:13,090
Speaker 1: on YouTube. The conversation between mgons and Drake is simply

376
00:29:13,130 --> 00:29:17,570
Speaker 1: a prophetic glimpse of the future of the Internet. The

377
00:29:17,690 --> 00:29:22,170
Speaker 1: modern Internet is stripped of any context. Short statements could

378
00:29:22,170 --> 00:29:26,410
Speaker 1: be ripped out of a longer argument and shared, mocked, retweeted,

379
00:29:26,490 --> 00:29:30,530
Speaker 1: or subtweeted. When some statement goes viral, most of the

380
00:29:30,570 --> 00:29:33,530
Speaker 1: people who see it have no idea what the original

381
00:29:33,610 --> 00:29:38,090
Speaker 1: context might have been. In this environment, certain kinds of

382
00:29:38,170 --> 00:29:45,050
Speaker 1: statements thrive one liners, epigrams, smackdowns, and insults. That may

383
00:29:45,050 --> 00:29:48,690
Speaker 1: be why everywhere you look these days you see comments

384
00:29:48,690 --> 00:29:52,490
Speaker 1: that remind you of mgns. Some of them are from bots,

385
00:29:53,330 --> 00:29:57,410
Speaker 1: some of them from humans, and there's so little context

386
00:29:57,610 --> 00:30:01,210
Speaker 1: that you'll find yourself looking from bot to human, and

387
00:30:01,330 --> 00:30:04,970
Speaker 1: from human to bot, and from bot to human again.

388
00:30:06,330 --> 00:30:10,010
Speaker 1: If it's impossible to say which is which, that's not

389
00:30:10,210 --> 00:30:14,050
Speaker 1: because the bots are so brilliant. It's because we humans

390
00:30:14,090 --> 00:30:21,130
Speaker 1: have lowered ourselves to their level. I'm not sure how

391
00:30:21,170 --> 00:30:24,530
Speaker 1: to fix Twitter or the comments on YouTube, but I

392
00:30:24,610 --> 00:30:27,770
Speaker 1: do know that we can at least take responsibility for

393
00:30:27,890 --> 00:30:31,890
Speaker 1: our own conversations, and we can all do so much

394
00:30:31,930 --> 00:30:37,330
Speaker 1: better than emgns. After the writer Brian Christian had spent

395
00:30:37,490 --> 00:30:41,490
Speaker 1: months pondering the history of chatbots, he concluded that their

396
00:30:41,570 --> 00:30:44,130
Speaker 1: limitations could teach us a lot about how to be

397
00:30:44,170 --> 00:30:48,850
Speaker 1: a better conversation partner. In the beginning, there was Eliza.

398
00:30:48,970 --> 00:30:53,170
Speaker 1: It was a passive listener. Please go on, it would say,

399
00:30:53,850 --> 00:30:57,210
Speaker 1: tell me more, can you think of a specific example.

400
00:30:58,570 --> 00:31:01,690
Speaker 1: Then there was Converse, which was the opposite of passive.

401
00:31:02,130 --> 00:31:06,690
Speaker 1: It insisted on wrenching every conversation to focus on the Clintons,

402
00:31:07,810 --> 00:31:11,490
Speaker 1: and most tur test chatbots try hard to keep the

403
00:31:11,530 --> 00:31:18,450
Speaker 1: conversation as routine as possible. Hi, Hi, how are you fine? Thanks?

404
00:31:18,570 --> 00:31:23,890
Speaker 1: How are you? Oh fine? It's all too human, isn't it.

405
00:31:25,250 --> 00:31:30,170
Speaker 1: Brian Christian concluded that these chatbots help us by highlighting

406
00:31:30,210 --> 00:31:34,730
Speaker 1: the worst parts of authentic human conversation. We all know

407
00:31:34,850 --> 00:31:38,930
Speaker 1: someone like Eliza who asks for information and nods understandingly,

408
00:31:39,410 --> 00:31:43,450
Speaker 1: but never volunteers anything about themselves, and we're surrounded by

409
00:31:43,490 --> 00:31:46,650
Speaker 1: people like Converse, who bully their way into talking about

410
00:31:46,650 --> 00:31:52,970
Speaker 1: their favorite topic. As for routine, almost scripted conversations want

411
00:31:52,970 --> 00:31:55,970
Speaker 1: to have their place, of course, but a real human

412
00:31:56,010 --> 00:32:02,290
Speaker 1: connection requires much more than that. Hello, Hello, nice weather,

413
00:32:02,330 --> 00:32:06,570
Speaker 1: isn't it. Yeah, very nice. Next time you're talking to

414
00:32:06,610 --> 00:32:10,010
Speaker 1: a stranger at a party, you can an observation about

415
00:32:10,010 --> 00:32:13,930
Speaker 1: the snacks, or you could ask what trait do your

416
00:32:13,930 --> 00:32:20,970
Speaker 1: most deplore in yourself? Or have you ever broken anyone's heart? Risky,

417
00:32:21,370 --> 00:32:26,650
Speaker 1: much more likely to go somewhere interesting. A meaningful conversation

418
00:32:26,970 --> 00:32:30,970
Speaker 1: can't be scripted, It can't be one sided, and it

419
00:32:31,010 --> 00:32:35,250
Speaker 1: builds over time. It isn't a series of unconnected one liners.

420
00:32:36,410 --> 00:32:40,090
Speaker 1: Bots as simple as mgns and Jenny eighteen passed the

421
00:32:40,210 --> 00:32:43,930
Speaker 1: cheering test because the kind of conversations they have trolleying

422
00:32:44,010 --> 00:32:48,210
Speaker 1: or sexting don't need to have a history. We humans

423
00:32:48,210 --> 00:32:52,730
Speaker 1: can do better than that if we try, and please,

424
00:32:53,850 --> 00:33:01,770
Speaker 1: let's try. It took months before Robert Epstein finally realized

425
00:33:01,810 --> 00:33:05,810
Speaker 1: that his conversation with Emilie was missing. As certain as something,

426
00:33:06,730 --> 00:33:10,330
Speaker 1: there was no sense of progress in the corresponder. Emily

427
00:33:10,450 --> 00:33:13,610
Speaker 1: kept talking about her mother and her friends and the

428
00:33:13,730 --> 00:33:16,570
Speaker 1: nice day she was having, but never really built on

429
00:33:16,690 --> 00:33:20,010
Speaker 1: what Robert was saying to her or seriously engaged with

430
00:33:20,010 --> 00:33:24,930
Speaker 1: his questions. In January two thousand and seven, she mentioned

431
00:33:25,010 --> 00:33:27,250
Speaker 1: going for a walk in the park with a friend

432
00:33:28,330 --> 00:33:32,530
Speaker 1: Robert wandered about that. Wasn't it twelve degrees fahrenheit and

433
00:33:32,610 --> 00:33:37,490
Speaker 1: snowing heavily? He asked. She ignored the question, and then

434
00:33:37,570 --> 00:33:40,930
Speaker 1: Robert began to realize that she had been ignoring almost

435
00:33:41,010 --> 00:33:44,850
Speaker 1: everything he said, but he'd barely noticed that there was

436
00:33:44,930 --> 00:33:49,050
Speaker 1: nothing there under the flirtatious surface. There's a reason why

437
00:33:49,090 --> 00:33:55,370
Speaker 1: we call the conversation between lovers sweet nothings. Robert Epstein

438
00:33:55,770 --> 00:33:59,450
Speaker 1: emailed her a short message. It was nothing more than

439
00:33:59,490 --> 00:34:05,090
Speaker 1: a long string of random keystrokes, signed with love Robert.

440
00:34:06,210 --> 00:34:10,970
Speaker 1: She replied with a long letter about her mother. Finally,

441
00:34:11,250 --> 00:34:15,730
Speaker 1: Robert Epstein, one of the world's leading authorities on chatbots,

442
00:34:16,330 --> 00:34:20,130
Speaker 1: realized that he had spent the last four months trying

443
00:34:20,210 --> 00:34:25,890
Speaker 1: to seduce one. Epstein had a sense of humor about it.

444
00:34:26,210 --> 00:34:29,970
Speaker 1: He wrote about his mistake in Scientific American and noted

445
00:34:30,050 --> 00:34:32,370
Speaker 1: that he and a student had started to make a

446
00:34:32,490 --> 00:34:37,850
Speaker 1: detailed study of internet chatbots. This exercise, he equipped, is

447
00:34:37,930 --> 00:34:44,130
Speaker 1: largely for my own protection. Very wise, because whoever programmed

448
00:34:44,170 --> 00:34:47,850
Speaker 1: Amelie was no doubt learning every day from the bots

449
00:34:47,930 --> 00:34:53,650
Speaker 1: conversations with lonely men like Robert Epstein. The spirit of

450
00:34:53,650 --> 00:34:56,970
Speaker 1: cautionary tales is usually that we should learn from other

451
00:34:57,170 --> 00:35:02,250
Speaker 1: people's mistakes. But I hope we learned from chatbot's mistakes too.

452
00:35:03,090 --> 00:35:07,690
Speaker 1: They struggle to improvise, they deal clumsily with context or memory.

453
00:35:08,410 --> 00:35:12,890
Speaker 1: They deliver unbalanced conversations, either all give or all take,

454
00:35:14,010 --> 00:35:18,490
Speaker 1: and they can't let a conversation blossom over time. We can,

455
00:35:19,890 --> 00:35:23,210
Speaker 1: and by noticing what they do badly, we can and

456
00:35:23,450 --> 00:35:31,610
Speaker 1: should learn to do it better. For his part, Robert

457
00:35:31,650 --> 00:35:42,930
Speaker 1: Epstein stopped dating chatbots and married a poet. The perfect

458
00:35:42,970 --> 00:35:46,530
Speaker 1: guide to the history of chatbots is Brian Christians The

459
00:35:46,650 --> 00:35:50,490
Speaker 1: Most Human Human. For a full list of our sources,

460
00:35:50,730 --> 00:35:57,050
Speaker 1: see the show notes at Tim Harford dot com. How

461
00:35:57,090 --> 00:36:02,890
Speaker 1: do you do? Not too bad? How do you do nice?

462
00:36:02,890 --> 00:36:12,170
Speaker 1: To meet you well? Goodbye then. Cautionary Tales is written

463
00:36:12,170 --> 00:36:16,010
Speaker 1: by me Tim Harford with Andrew Wright. It's produced by

464
00:36:16,090 --> 00:36:19,730
Speaker 1: Ryan Dilley with support from Courtney Guarino and Emily Vaughan.

465
00:36:20,210 --> 00:36:23,050
Speaker 1: The sound design and original music is the work of

466
00:36:23,250 --> 00:36:26,770
Speaker 1: Pascal Wise. It features the voice talents of Ben Crow,

467
00:36:27,170 --> 00:36:31,810
Speaker 1: Melanie Gutridge, Stella Harford, and Rufus Wright. The show also

468
00:36:31,890 --> 00:36:34,730
Speaker 1: wouldn't have been possible without the work of Mia LaBelle,

469
00:36:35,050 --> 00:36:40,170
Speaker 1: Jacob Weisberg, Heather Fane, John Schnars, Julia Barton, Carlie mcgliori,

470
00:36:40,570 --> 00:36:45,810
Speaker 1: Eric Sandler, Royston Berserve, Maggie Taylor, Nicole Morano, Danielle Lakhan,

471
00:36:46,250 --> 00:36:51,410
Speaker 1: and Maya Caaning. Cautionary Tales is a production of Pushkin Industries.

472
00:36:51,650 --> 00:36:54,490
Speaker 1: If you like the show, please remember to share, rape

473
00:36:54,490 --> 00:36:57,570
Speaker 1: and review, tell a friend, tell two friends, and if

474
00:36:57,570 --> 00:36:59,770
Speaker 1: you want to hear the show, adds free and listen

475
00:36:59,810 --> 00:37:04,010
Speaker 1: to four exclusive Cautionary Tales shorts. Then sign up for

476
00:37:04,130 --> 00:37:07,690
Speaker 1: Pushkin Plus on the show page, in Apple Podcast or

477
00:37:07,770 --> 00:37:10,570
Speaker 1: at pushkin dot f M, slash plus