1
00:00:15,410 --> 00:00:25,850
Speaker 1: Pushkin. When the trouble started in the middle of the Atlantic,

2
00:00:26,530 --> 00:00:30,650
Speaker 1: Captain Mark Dubois was in the flight rest compartment, right

3
00:00:30,690 --> 00:00:33,130
Speaker 1: next to the flight deck. He was in charge of

4
00:00:33,330 --> 00:00:37,690
Speaker 1: Air France flight four four seven, en route overnight from

5
00:00:37,770 --> 00:00:41,610
Speaker 1: Rio de Janeiro to Paris, but he was tired. He

6
00:00:41,610 --> 00:00:44,490
Speaker 1: had been seeing the sights of Rio with his girlfriend

7
00:00:45,050 --> 00:00:49,050
Speaker 1: Copacabana Beach a helicopter tour, and he hadn't had a

8
00:00:49,050 --> 00:00:52,450
Speaker 1: lot of sleep. The airliner was in the hands of

9
00:00:52,610 --> 00:00:57,850
Speaker 1: flight officers David Robert and Pierre Cedric Bonard, and when

10
00:00:57,850 --> 00:01:02,530
Speaker 1: the trouble started, first Officer David Robert pressed the call

11
00:01:02,610 --> 00:01:07,610
Speaker 1: button to summon Captain Dubois. When you're asleep and the

12
00:01:07,650 --> 00:01:11,810
Speaker 1: alarm goes off, how quickly do you wake up? Captain

13
00:01:11,890 --> 00:01:15,490
Speaker 1: Dubois took ninety eight seconds to get out of bed

14
00:01:16,010 --> 00:01:22,050
Speaker 1: into the flight deck, not exactly slow, but not quick enough.

15
00:01:23,170 --> 00:01:25,850
Speaker 1: By the time Dubois arrived on the flight deck of

16
00:01:25,850 --> 00:01:29,810
Speaker 1: his own airplane, he was confronted with a scene of confusion.

17
00:01:30,490 --> 00:01:33,730
Speaker 1: The plane was shaking so violently that it was hard

18
00:01:33,730 --> 00:01:37,450
Speaker 1: to read the instruments. An alarm was alternating between a

19
00:01:37,610 --> 00:01:45,410
Speaker 1: chirruping trill and an automated voice store Stare Store. His

20
00:01:45,610 --> 00:01:49,690
Speaker 1: junior co pilots were at the controls. In a calm tone.

21
00:01:49,970 --> 00:01:55,250
Speaker 1: Captain Dubois asked, what's happening. Co pilot David Robert's answer

22
00:01:55,530 --> 00:01:59,290
Speaker 1: was less calm. We completely lost control of the airplane

23
00:01:59,410 --> 00:02:04,450
Speaker 1: and we don't understand anything. We tried everything. Two of

24
00:02:04,490 --> 00:02:08,730
Speaker 1: those statements were wrong. The crew were in control of

25
00:02:08,730 --> 00:02:11,930
Speaker 1: the air plane. It was doing exactly what they told

26
00:02:11,970 --> 00:02:15,730
Speaker 1: it to do, and they hadn't tried everything. In fact,

27
00:02:16,170 --> 00:02:20,010
Speaker 1: one very simple course of action would soon have solved

28
00:02:20,010 --> 00:02:24,010
Speaker 1: their problem. But David Robert was certainly right on one count.

29
00:02:24,850 --> 00:02:30,370
Speaker 1: They didn't understand what was happening. I'm Tim Harford, and

30
00:02:30,410 --> 00:02:57,210
Speaker 1: you're listening to cautionary tales. The disappearance of Air France

31
00:02:57,330 --> 00:03:00,450
Speaker 1: flight four four seven in the early hours of the

32
00:03:00,490 --> 00:03:04,530
Speaker 1: first of June two thousand and nine was at first

33
00:03:04,610 --> 00:03:10,170
Speaker 1: an utter mystery. The plane was an Airbus A three thirty,

34
00:03:10,610 --> 00:03:13,970
Speaker 1: a modern airplane with an excellent safety record. In the

35
00:03:14,050 --> 00:03:17,850
Speaker 1: fifteen years since being introduced in the early nineteen nineties,

36
00:03:18,250 --> 00:03:23,130
Speaker 1: not a single passenger A three thirty had crashed anywhere

37
00:03:23,170 --> 00:03:26,250
Speaker 1: in the world. This one was just four years old

38
00:03:26,370 --> 00:03:31,410
Speaker 1: and fully serviced. The crew were highly trained, the captain experienced,

39
00:03:31,810 --> 00:03:35,010
Speaker 1: and there seemed to be nothing too challenging about the conditions.

40
00:03:35,890 --> 00:03:42,690
Speaker 1: And yet somehow Flight four four seven had simply fallen

41
00:03:42,730 --> 00:03:47,050
Speaker 1: out of the sky. Search teams found traces of wreckage

42
00:03:47,090 --> 00:03:49,770
Speaker 1: on the surface of the waves a few hours later,

43
00:03:50,570 --> 00:03:54,530
Speaker 1: confirming that the plane had been destroyed and all two

44
00:03:54,690 --> 00:03:59,010
Speaker 1: hundred and twenty eight people on board were dead. But

45
00:03:59,090 --> 00:04:03,450
Speaker 1: the black box flight recorder, containing possibly vital clues to

46
00:04:03,530 --> 00:04:07,570
Speaker 1: the cause of the disaster, it was somewhere on the

47
00:04:07,570 --> 00:04:11,650
Speaker 1: bottom of the Atlantic Ocean. There wasn't until nearly two

48
00:04:11,770 --> 00:04:15,290
Speaker 1: years later that the black box was discovered and the

49
00:04:15,370 --> 00:04:21,570
Speaker 1: mystery could start to be solved. This dear listener, is

50
00:04:21,610 --> 00:04:25,330
Speaker 1: not just a story about a plane crash. It's a

51
00:04:25,370 --> 00:04:34,690
Speaker 1: warning to all of us about what's coming. Air France

52
00:04:34,770 --> 00:04:37,690
Speaker 1: flight four four seven had begun with an on time

53
00:04:37,810 --> 00:04:41,690
Speaker 1: takeoff from Rio de Geneio at seven twenty nine pm

54
00:04:41,850 --> 00:04:45,410
Speaker 1: on May the thirty first, two thousand and nine, bound

55
00:04:45,450 --> 00:04:50,410
Speaker 1: for Paris with hindsight. The three pilots had their weaknesses.

56
00:04:51,010 --> 00:04:55,050
Speaker 1: Captain Mark Dubois fifty eight, had plenty of experience flying

57
00:04:55,130 --> 00:04:59,810
Speaker 1: both light airplanes and large passenger aircraft, but it had

58
00:04:59,930 --> 00:05:04,810
Speaker 1: very little sleep. Pierre Cedric Bonnard thirty two, was young

59
00:05:04,890 --> 00:05:08,450
Speaker 1: and didn't have many flying hours under his belt. David

60
00:05:08,570 --> 00:05:12,570
Speaker 1: Robert thirty seven had recently become an Air France manager

61
00:05:12,690 --> 00:05:16,130
Speaker 1: and no longer flew full time. He was flying this

62
00:05:16,250 --> 00:05:22,210
Speaker 1: route to keep active his credentials as a pilot. Fortunately,

63
00:05:22,330 --> 00:05:26,290
Speaker 1: given these potential fragilities, the crew were in charge of

64
00:05:26,290 --> 00:05:30,130
Speaker 1: one of the most advanced planes in the world, an

65
00:05:30,130 --> 00:05:35,770
Speaker 1: Airbus three thirty. Legendarily smooth and easy to fly. Like

66
00:05:36,010 --> 00:05:39,410
Speaker 1: any other modern aircraft, the A three thirty has an

67
00:05:39,450 --> 00:05:43,290
Speaker 1: autopilot to keep the plane flying on a programmed route,

68
00:05:43,690 --> 00:05:47,330
Speaker 1: but it also has a much more sophisticated automation system

69
00:05:47,610 --> 00:05:52,250
Speaker 1: called assistive fly by wire, a traditional airplane gives the

70
00:05:52,290 --> 00:05:57,770
Speaker 1: pilot direct control of the flaps on the plane, it's rudder, elevators,

71
00:05:57,930 --> 00:06:02,490
Speaker 1: and ailerons. This means the pilot has plenty of latitude

72
00:06:02,530 --> 00:06:07,170
Speaker 1: to make mistakes. Fly by wire is much smoother and

73
00:06:07,410 --> 00:06:11,970
Speaker 1: potentially safer. Two. It inserts itself between the pilot with

74
00:06:12,050 --> 00:06:16,250
Speaker 1: all his or her faults, and the plane's physical mechanisms.

75
00:06:16,770 --> 00:06:21,370
Speaker 1: A tactful translator between human and machine, it observes the

76
00:06:21,410 --> 00:06:25,170
Speaker 1: pilot tugging on the controls, figures out how the pilot

77
00:06:25,410 --> 00:06:30,370
Speaker 1: wanted the plane to move, and executes that maneuver perfectly.

78
00:06:31,010 --> 00:06:34,090
Speaker 1: It will turn a clumsy movement into a graceful one.

79
00:06:35,010 --> 00:06:40,290
Speaker 1: This makes it very hard to crash an A three thirty,

80
00:06:40,450 --> 00:06:49,050
Speaker 1: very hard, but it turns out not impossible. As the

81
00:06:49,050 --> 00:06:53,450
Speaker 1: plane approached the equator, the junior pilot, Pierre Cedric Bonan,

82
00:06:54,010 --> 00:06:58,090
Speaker 1: was flying, or more precisely, was letting the autopilot fly.

83
00:06:59,010 --> 00:07:03,290
Speaker 1: Captain du Bras was with him. Ahead on the weather radar,

84
00:07:03,610 --> 00:07:07,450
Speaker 1: they could see tropical thunderstorms gathering, which at that time

85
00:07:07,490 --> 00:07:11,890
Speaker 1: of year and in that location was common enough. We're

86
00:07:11,890 --> 00:07:16,250
Speaker 1: not bothered by storm clouds, eh, said the old hand Dubois.

87
00:07:17,170 --> 00:07:22,010
Speaker 1: Young Bonan didn't respond. He was it would turn out

88
00:07:22,690 --> 00:07:26,890
Speaker 1: very much bothered by the thunderstorms, and many captains would

89
00:07:26,890 --> 00:07:29,810
Speaker 1: have chosen to divert around them for the comfort of

90
00:07:29,850 --> 00:07:33,570
Speaker 1: the passengers as much as anything. That wasn't a possibility

91
00:07:33,650 --> 00:07:39,330
Speaker 1: that was discussed. Instead, Dubois noted, we'll wait a little

92
00:07:39,370 --> 00:07:44,570
Speaker 1: and see if that goes away, and if not, then

93
00:07:44,970 --> 00:07:50,730
Speaker 1: what not Captain Dubois's problem. A few minutes later, at

94
00:07:50,730 --> 00:07:54,570
Speaker 1: eleven PM, Rio time, he pressed the buzzer to summon

95
00:07:54,610 --> 00:07:59,170
Speaker 1: David Robert so that Dubois could take a nap. This

96
00:07:59,610 --> 00:08:03,930
Speaker 1: wasn't particularly unusual. Everyone needs a rest after all, and

97
00:08:04,010 --> 00:08:07,210
Speaker 1: the junior pilots need to get some experience making decisions

98
00:08:07,250 --> 00:08:12,090
Speaker 1: about the plane. With the plane on course to fly

99
00:08:12,290 --> 00:08:16,730
Speaker 1: straight into thunderstorms, Dubois's decision to leave the flight deck

100
00:08:17,810 --> 00:08:23,730
Speaker 1: raises questions. The chief investigator of the crash, A Lamb

101
00:08:23,850 --> 00:08:27,610
Speaker 1: we Are, spoke to the writer and pilot, William Longevisha

102
00:08:27,770 --> 00:08:32,170
Speaker 1: about that his leaving was not against the rules. Still,

103
00:08:32,530 --> 00:08:36,610
Speaker 1: it is surprising if you're responsible for the outcome, you

104
00:08:36,730 --> 00:08:40,810
Speaker 1: do not go on vacation. During the main event, with

105
00:08:41,090 --> 00:08:45,570
Speaker 1: Dubois Gonne, Pierre Cedric Bannan's nerves about the storms became

106
00:08:45,650 --> 00:08:50,370
Speaker 1: more apparent. Put a lavash poutin, he yelled. At one point.

107
00:08:50,770 --> 00:08:55,130
Speaker 1: The outburst, the French equivalent of fucking, hell, fuck, seemed

108
00:08:55,170 --> 00:08:58,890
Speaker 1: to be provoked by nothing in particular. He talked with

109
00:08:58,970 --> 00:09:01,450
Speaker 1: David Robert about how it was a shame that they

110
00:09:01,450 --> 00:09:05,690
Speaker 1: couldn't fly high enough to clear the storms. But they couldn't.

111
00:09:06,250 --> 00:09:08,850
Speaker 1: There's a limit to how high a plane can go.

112
00:09:09,610 --> 00:09:12,810
Speaker 1: That high you fly the further you are from dangers

113
00:09:12,850 --> 00:09:18,250
Speaker 1: on the ground. But the thinner the atmosphere becomes, and

114
00:09:18,330 --> 00:09:21,450
Speaker 1: the atmosphere, of course is what the wings are using

115
00:09:21,730 --> 00:09:26,050
Speaker 1: to support the aircraft too high and the margins for

116
00:09:26,250 --> 00:09:30,330
Speaker 1: error become tight. That's okay, though, because on an A

117
00:09:30,570 --> 00:09:35,570
Speaker 1: three thirty, the assistive fly by wire system always keeps

118
00:09:35,610 --> 00:09:41,370
Speaker 1: the pilots within those margins. As the plane approached the storm,

119
00:09:41,730 --> 00:09:46,730
Speaker 1: ice crystals rattled unnervingly against the windscreen and ice began

120
00:09:46,810 --> 00:09:51,330
Speaker 1: to form on the wings. Banan and Robert switched on

121
00:09:51,490 --> 00:09:55,250
Speaker 1: the anti icing system to prevent too much ice building

122
00:09:55,330 --> 00:09:59,810
Speaker 1: up and slowing the plane down. Robert nudged Bonan a

123
00:09:59,890 --> 00:10:02,890
Speaker 1: couple of times to pull left, avoiding the worst of

124
00:10:02,930 --> 00:10:08,890
Speaker 1: the weather. Banar seemed slightly distracted, perhaps put on edge

125
00:10:08,890 --> 00:10:11,850
Speaker 1: by the fact that they hadn't plotted a route around

126
00:10:11,890 --> 00:10:17,330
Speaker 1: the storms much earlier. A faint odor of electrical burning

127
00:10:17,490 --> 00:10:24,090
Speaker 1: filled the cockpit and the temperature rose. Robert assured Bonan

128
00:10:24,210 --> 00:10:27,450
Speaker 1: that all this was the result of the electrical storm,

129
00:10:27,650 --> 00:10:31,970
Speaker 1: not an equipment failure. But the ice wasn't just forming

130
00:10:32,050 --> 00:10:35,650
Speaker 1: on the wings. It had also blocked the plane's air

131
00:10:35,730 --> 00:10:40,090
Speaker 1: speed sensors, meaning that the autopilot could no longer fly

132
00:10:40,250 --> 00:10:45,290
Speaker 1: the plane by itself. A defrosting system activated to melt

133
00:10:45,330 --> 00:10:48,690
Speaker 1: the ice and unblock the sensors, but in the meantime

134
00:10:49,170 --> 00:10:53,890
Speaker 1: the pilots needed to take control. An alarm sounded in

135
00:10:53,930 --> 00:10:59,210
Speaker 1: the cockpit, notifying Bonar and Robert that the autopilot had disconnected,

136
00:10:59,970 --> 00:11:02,690
Speaker 1: and a message popped up, adding that at the same

137
00:11:02,770 --> 00:11:05,970
Speaker 1: time and for the same reason, the assistive fly by

138
00:11:06,010 --> 00:11:10,450
Speaker 1: wire system had stopped assisting. No longer would it be

139
00:11:10,490 --> 00:11:18,210
Speaker 1: the smooth tongued interpreter between pilot and plane. Instead, the

140
00:11:18,290 --> 00:11:23,050
Speaker 1: system was a literal minded translator that would relay any instruction,

141
00:11:23,690 --> 00:11:28,930
Speaker 1: no matter how foolish. Pierre Cedric Bonnard was in direct,

142
00:11:29,570 --> 00:11:34,450
Speaker 1: unmediated control of the airplane, a situation with which he

143
00:11:34,570 --> 00:11:40,570
Speaker 1: had almost no experience. Still, all he needed to do

144
00:11:41,130 --> 00:11:44,970
Speaker 1: was to keep the plane flying straight and level for

145
00:11:45,050 --> 00:11:48,970
Speaker 1: a couple of minutes until the air speed indicators defrosted.

146
00:11:50,050 --> 00:11:55,850
Speaker 1: How hard could that be? Cautionary tales will return after

147
00:11:55,850 --> 00:12:08,770
Speaker 1: the break. Not long ago, Fabrizio de Laqua, a researcher

148
00:12:08,850 --> 00:12:12,690
Speaker 1: at Harvard Business School, ran an experiment to see how

149
00:12:12,810 --> 00:12:17,450
Speaker 1: people performed when they were assisted by an algorithm. The

150
00:12:17,530 --> 00:12:21,450
Speaker 1: experiment was designed to be practical and realistic. It involved

151
00:12:21,570 --> 00:12:27,050
Speaker 1: professional recruiters being paid to evaluate real resumes, equipped with

152
00:12:27,210 --> 00:12:32,010
Speaker 1: commercially available software to use the sophisticated pattern recognition we

153
00:12:32,050 --> 00:12:37,130
Speaker 1: call machine learning to assess and grade those resumes. Some

154
00:12:37,210 --> 00:12:40,170
Speaker 1: of the recruiters were given software that was designed to

155
00:12:40,250 --> 00:12:44,370
Speaker 1: operate at a very high standard for simplicity. Delaqua calls

156
00:12:44,450 --> 00:12:50,050
Speaker 1: that good AI. Other recruiters chosen at random, were given

157
00:12:50,090 --> 00:12:54,210
Speaker 1: an algorithm which didn't work quite as well, or bad AI.

158
00:12:55,010 --> 00:12:58,050
Speaker 1: They were told that the algorithm was patchy, it would

159
00:12:58,090 --> 00:13:01,690
Speaker 1: give good advice, but it would also make mistakes. Then

160
00:13:01,930 --> 00:13:04,730
Speaker 1: there was a third group, also chosen at random, who

161
00:13:04,770 --> 00:13:10,490
Speaker 1: got no AI support at all. That the computer assistance

162
00:13:10,770 --> 00:13:14,650
Speaker 1: was very helpful. Whether recruiters were given good AI or

163
00:13:14,730 --> 00:13:19,170
Speaker 1: bad AI, they made more accurate recruitment choices than the

164
00:13:19,210 --> 00:13:24,730
Speaker 1: recruiters with no AI at all. But here's the surprise.

165
00:13:26,170 --> 00:13:29,690
Speaker 1: The recruiters with good AI did worse than those with

166
00:13:29,850 --> 00:13:35,330
Speaker 1: bad AI. Why because they switched off The group who

167
00:13:35,410 --> 00:13:39,090
Speaker 1: had the good AI spent less time analyzing each application.

168
00:13:39,410 --> 00:13:41,610
Speaker 1: They more or less left the decision to the computer.

169
00:13:42,570 --> 00:13:45,570
Speaker 1: The group who knew they had a less reliable AI tool,

170
00:13:46,130 --> 00:13:50,170
Speaker 1: spent more effort and paid closer attention to the applications

171
00:13:50,730 --> 00:13:54,730
Speaker 1: they used the AI, but they also used their own judgment,

172
00:13:55,370 --> 00:13:59,650
Speaker 1: and despite having a worse tool, they made more accurate decisions.

173
00:14:00,570 --> 00:14:04,850
Speaker 1: With the rise of powerful new AI systems, we tend

174
00:14:04,890 --> 00:14:09,810
Speaker 1: to ask who's better humans or computers. The De Laqua

175
00:14:09,970 --> 00:14:12,930
Speaker 1: experiment reminds us that that might be the wrong question.

176
00:14:13,770 --> 00:14:18,490
Speaker 1: Often decisions are made by humans and computers working together,

177
00:14:19,050 --> 00:14:23,010
Speaker 1: and just using the best computer doesn't necessarily get the

178
00:14:23,010 --> 00:14:30,810
Speaker 1: best results out of the humans. Pierre Cedric Bonan was

179
00:14:30,850 --> 00:14:37,930
Speaker 1: flying at high altitude in thin, unforgiving air into a thunderstorm.

180
00:14:38,210 --> 00:14:41,850
Speaker 1: It was dark, with an unnerving burning smell in the

181
00:14:41,890 --> 00:14:45,450
Speaker 1: cabin because of the electrical charge in the air and

182
00:14:45,490 --> 00:14:48,970
Speaker 1: the clatter of hailstones on the windshield. Then there was

183
00:14:49,010 --> 00:14:53,250
Speaker 1: the sound of the alarm disconnecting the autopilot Boan. Anne

184
00:14:53,370 --> 00:14:56,530
Speaker 1: needed all the help he could get, and just at

185
00:14:56,530 --> 00:15:02,570
Speaker 1: that moment the assistive fly by wire system disconnected, but

186
00:15:02,650 --> 00:15:08,290
Speaker 1: Anne had no real experience flying without it. When the

187
00:15:08,330 --> 00:15:14,210
Speaker 1: autopilot disengaged, Banan grabbed the control stick and immediately the

188
00:15:14,250 --> 00:15:18,970
Speaker 1: trouble began. The plane rocked right and left and right

189
00:15:19,050 --> 00:15:24,530
Speaker 1: and left, and each time Banan over corrected. He was

190
00:15:24,650 --> 00:15:27,770
Speaker 1: used to flying in the thick air of takeoff and landing,

191
00:15:28,290 --> 00:15:33,890
Speaker 1: whereas at high altitude plane behaved differently. And more importantly,

192
00:15:33,930 --> 00:15:37,370
Speaker 1: Banan was used to flying with the assistive fly by wire,

193
00:15:37,770 --> 00:15:42,610
Speaker 1: gracefully interpreting his every move, and suddenly he was having

194
00:15:42,610 --> 00:15:46,010
Speaker 1: to fly the plane without it right and left and

195
00:15:46,130 --> 00:15:50,730
Speaker 1: right and left. It rocked ten times in thirty seconds.

196
00:15:51,410 --> 00:15:53,610
Speaker 1: The side to side rocking of the plane must have

197
00:15:53,650 --> 00:15:58,850
Speaker 1: been unsettling, but it wasn't particularly dangerous. What was dangerous

198
00:15:59,090 --> 00:16:02,530
Speaker 1: was that Banan also pulled back on the control stick,

199
00:16:03,050 --> 00:16:06,570
Speaker 1: sending the plane into a climb in the thin air

200
00:16:06,890 --> 00:16:11,370
Speaker 1: a climbing plane could easily store. Stalling is what happens

201
00:16:11,410 --> 00:16:15,530
Speaker 1: when the wings don't generate enough lift. A stalling plane

202
00:16:15,610 --> 00:16:20,650
Speaker 1: is pointed upwards trying to climb, but it's losing forward

203
00:16:20,770 --> 00:16:26,210
Speaker 1: speed and losing height, scrabbling for altitude as it slides

204
00:16:26,370 --> 00:16:30,930
Speaker 1: down through the air. So why did Bonan point the

205
00:16:30,970 --> 00:16:34,650
Speaker 1: plane up and risk a stall. It was an instinctive

206
00:16:34,810 --> 00:16:38,490
Speaker 1: reaction from a pilot used to taking control of the

207
00:16:38,530 --> 00:16:42,250
Speaker 1: plane at takeoff and landing when a stall is unlikely

208
00:16:42,890 --> 00:16:45,810
Speaker 1: and the main danger comes from not having enough height

209
00:16:46,050 --> 00:16:49,570
Speaker 1: and slamming into the ground. If there's a problem as

210
00:16:49,650 --> 00:16:52,970
Speaker 1: you're landing, you gun the engines and point the nose

211
00:16:53,010 --> 00:16:58,930
Speaker 1: of the plane upwards. That's what Banan was doing. In

212
00:16:59,010 --> 00:17:04,770
Speaker 1: an article in Popular Mechanics, the aviation journalist Jeff Wise explained,

213
00:17:05,810 --> 00:17:10,290
Speaker 1: intense psychological stress tends to shut down the part of

214
00:17:10,330 --> 00:17:15,450
Speaker 1: the brain responsible for innovative, creative thought. Instead, we tend

215
00:17:15,490 --> 00:17:19,330
Speaker 1: to revert to the familiar and the well rehearsed. At

216
00:17:19,330 --> 00:17:23,730
Speaker 1: more than thirty seven thousand feet, the familiar and well

217
00:17:23,810 --> 00:17:26,690
Speaker 1: rehearsed action of pointing the nose of the plane up

218
00:17:27,370 --> 00:17:30,570
Speaker 1: wasn't going to make Bonan safer. It was bringing the

219
00:17:30,810 --> 00:17:41,250
Speaker 1: entire plane closer to catastrophe. In nineteen forty two, two

220
00:17:41,370 --> 00:17:46,410
Speaker 1: psychologists Abraham and Edith Luchins they were married, published the

221
00:17:46,450 --> 00:17:50,810
Speaker 1: results of a famous experiment. In this experiment, subjects were

222
00:17:50,810 --> 00:17:55,130
Speaker 1: given three different sized water jugs and asked to figure

223
00:17:55,130 --> 00:17:59,410
Speaker 1: out how to measure out a certain amount. For example,

224
00:18:00,250 --> 00:18:03,850
Speaker 1: one jug might have a capacity of twenty ounces, the

225
00:18:03,930 --> 00:18:08,890
Speaker 1: second one hundred ounces, and the third four ounces. The

226
00:18:09,330 --> 00:18:14,090
Speaker 1: question is, how would you measure seventy two ounces using

227
00:18:14,130 --> 00:18:19,250
Speaker 1: these jugs? The answer, Fill one hundred ounce jug, then

228
00:18:19,410 --> 00:18:23,330
Speaker 1: pour off twenty ounces into the medium sized jug. Then

229
00:18:23,570 --> 00:18:26,650
Speaker 1: you filled the small four ounce jug twice from the

230
00:18:26,650 --> 00:18:29,610
Speaker 1: big jug. With the pencil and paper, it's not too

231
00:18:29,610 --> 00:18:34,450
Speaker 1: tricky to figure this out. One hundred minus twenty minus

232
00:18:34,490 --> 00:18:40,330
Speaker 1: four minus four gives you seventy two ounces. The Luchins

233
00:18:40,690 --> 00:18:44,850
Speaker 1: gave their experimental subjects several of these problems, each with

234
00:18:44,970 --> 00:18:48,490
Speaker 1: different sized jars and a different target volume of water,

235
00:18:49,010 --> 00:18:53,850
Speaker 1: but each time the solution followed the same pattern. Fill

236
00:18:53,890 --> 00:18:56,730
Speaker 1: the big jar, then use it to fill the medium

237
00:18:56,810 --> 00:19:02,810
Speaker 1: jar once and the small jar twice. Now comes the trick.

238
00:19:03,770 --> 00:19:07,930
Speaker 1: The Luchins would give people a problem like this. The

239
00:19:07,970 --> 00:19:13,490
Speaker 1: big jar hole thirty nine ounces, the medium jar holds fifteen,

240
00:19:14,290 --> 00:19:21,090
Speaker 1: the small jar holds three. How do you get eighteen ounces? Well,

241
00:19:21,170 --> 00:19:24,410
Speaker 1: you can repeat the same process as before, fill the

242
00:19:24,410 --> 00:19:26,770
Speaker 1: big jar and use it to fill the medium jar

243
00:19:26,890 --> 00:19:30,610
Speaker 1: once and the small jar twice. It works, but if

244
00:19:30,610 --> 00:19:34,650
Speaker 1: you do it that way, you're over complicating things, because

245
00:19:34,970 --> 00:19:37,690
Speaker 1: you could simply fill the medium and the small jar

246
00:19:37,890 --> 00:19:42,170
Speaker 1: fifteen plus three is eighteen. That's much easier. But a

247
00:19:42,210 --> 00:19:46,290
Speaker 1: lot of people missed that obvious solution because they'd already

248
00:19:46,330 --> 00:19:50,130
Speaker 1: solved a bunch of previous problems that required the more

249
00:19:50,290 --> 00:19:55,290
Speaker 1: elaborate method. Abraham and Edith Luchin also had a control group.

250
00:19:55,690 --> 00:19:59,970
Speaker 1: They hadn't been given any practice problems. Instead, they started

251
00:20:00,010 --> 00:20:02,650
Speaker 1: with the eighteen ounce problem, and of course most of

252
00:20:02,690 --> 00:20:07,170
Speaker 1: them found the simple solution. Not having practiced was actually

253
00:20:07,170 --> 00:20:10,730
Speaker 1: an advantage. They saw the problem with fresh eyes and

254
00:20:10,770 --> 00:20:15,450
Speaker 1: solved it quickly and simply. The people who had practiced

255
00:20:16,050 --> 00:20:20,570
Speaker 1: tended to get stuck with a clumsy solution. The Luchens

256
00:20:20,770 --> 00:20:25,730
Speaker 1: called this the einstellung effect. Einstellung is perhaps best translated

257
00:20:25,770 --> 00:20:30,690
Speaker 1: here as state of mind. The practiced participants found a

258
00:20:30,730 --> 00:20:33,290
Speaker 1: simple rule of thumb that seemed to work, and so

259
00:20:33,370 --> 00:20:38,090
Speaker 1: they began applying it unthinkingly. As the Luchins put it,

260
00:20:38,570 --> 00:20:48,570
Speaker 1: the problem solving act had been mechanized. Banan's instinctive attempt

261
00:20:48,650 --> 00:20:52,130
Speaker 1: to climb by pulling back on the stick demonstrated an

262
00:20:52,130 --> 00:20:57,770
Speaker 1: Einstellung effect in two ways. First, as Jeff Wise explained,

263
00:20:58,250 --> 00:21:01,490
Speaker 1: he was reverting to his instinct that when you're in trouble,

264
00:21:01,930 --> 00:21:04,850
Speaker 1: safety is to be found by pulling the plane up

265
00:21:05,210 --> 00:21:11,290
Speaker 1: and seeking height. Second, Banan had almost always flown the

266
00:21:11,330 --> 00:21:16,050
Speaker 1: a three point thirty plane with the assistive fly by wire,

267
00:21:16,410 --> 00:21:20,450
Speaker 1: and with the assistive fly by wire operating, you literally

268
00:21:21,210 --> 00:21:26,930
Speaker 1: cannot stall the plane. The computer won't let you. Banan

269
00:21:27,130 --> 00:21:31,570
Speaker 1: had been trained by his own airplane never to worry

270
00:21:31,570 --> 00:21:36,450
Speaker 1: about stalls, never to even think about stalls, because stalls

271
00:21:36,850 --> 00:21:44,010
Speaker 1: simply can't happen. As Flight four four seven began to

272
00:21:44,130 --> 00:21:51,890
Speaker 1: lose air speed and altitude, an automated voice announced stall. Kescassa,

273
00:21:52,770 --> 00:22:00,530
Speaker 1: said David Robert, what was that stall? Store? Over the

274
00:22:00,570 --> 00:22:07,290
Speaker 1: next four minutes, the word stall would be repeated more

275
00:22:07,330 --> 00:22:13,090
Speaker 1: than seventy times. But Bonar and Robert, it seems, couldn't

276
00:22:13,130 --> 00:22:17,730
Speaker 1: grasp that a stall was possible. Their ein stellum, their

277
00:22:17,890 --> 00:22:27,410
Speaker 1: state of mind made that risk inconceivable. I first heard

278
00:22:27,450 --> 00:22:30,970
Speaker 1: the story of Flight four four seven told on the

279
00:22:31,090 --> 00:22:35,090
Speaker 1: ninety nine percent Invisible podcast back in twenty fifteen. By

280
00:22:35,130 --> 00:22:38,530
Speaker 1: the way, nineteen nine percent Invisible is amazing, and if

281
00:22:38,570 --> 00:22:42,610
Speaker 1: by some miracle you're not already a listener, go and subscribe.

282
00:22:42,770 --> 00:22:47,170
Speaker 1: He can thank me later. Now. In twenty fifteen, this

283
00:22:47,250 --> 00:22:51,690
Speaker 1: seemed like a warning about self driving cars. Here's a

284
00:22:51,730 --> 00:22:56,410
Speaker 1: pilot who grew so reliant on his assistive technology that

285
00:22:56,490 --> 00:23:00,250
Speaker 1: he forgot how to fly a plane at high altitude.

286
00:23:01,130 --> 00:23:04,370
Speaker 1: So what happens when the self driving cars take over

287
00:23:04,930 --> 00:23:09,370
Speaker 1: and we all become Pierre Cedric Bonar, unable to remember

288
00:23:09,450 --> 00:23:13,130
Speaker 1: what to do when the computer needs us to take over.

289
00:23:14,730 --> 00:23:18,130
Speaker 1: I see the story differently now. It's not just the

290
00:23:18,210 --> 00:23:24,530
Speaker 1: self driving cars. It's the appearance of artificial intelligence everywhere.

291
00:23:25,730 --> 00:23:30,130
Speaker 1: Consider those decision making algorithms that Fabrizio de Laqua gave

292
00:23:30,210 --> 00:23:34,050
Speaker 1: to professional recruiters, which made them switch off and let

293
00:23:34,050 --> 00:23:38,650
Speaker 1: the algorithm handle the problem. He called that study falling

294
00:23:38,730 --> 00:23:41,930
Speaker 1: asleep at the wheel. I think you can see why.

295
00:23:43,090 --> 00:23:47,650
Speaker 1: Or generative AI, which we use to paint pictures, create videos,

296
00:23:47,730 --> 00:23:52,450
Speaker 1: write essays like the assistive fly by wire on an

297
00:23:52,490 --> 00:23:57,930
Speaker 1: A three thirty. It's a technological miracle. But like the

298
00:23:57,970 --> 00:24:01,890
Speaker 1: assistive fly by wire, the question is not how well

299
00:24:01,970 --> 00:24:06,050
Speaker 1: the computer works, it's how well the computers and the

300
00:24:06,130 --> 00:24:14,050
Speaker 1: humans work together. Consider the hapless lawyers who turned to

301
00:24:14,290 --> 00:24:18,410
Speaker 1: chat GPT for help in formulating a case, only to

302
00:24:18,490 --> 00:24:22,490
Speaker 1: find it had simply invented new cases. Not only did

303
00:24:22,530 --> 00:24:26,290
Speaker 1: this actually happen, it's happened more than once. In a

304
00:24:26,330 --> 00:24:29,810
Speaker 1: New York case, the lawyers were fined five thousand dollars

305
00:24:30,010 --> 00:24:33,290
Speaker 1: and ordered to write letters of apology to the judges

306
00:24:33,370 --> 00:24:36,970
Speaker 1: whose names had been taken in vain by chat GPT.

307
00:24:37,850 --> 00:24:41,210
Speaker 1: In Canada, another lawyer was let off with a warning.

308
00:24:41,650 --> 00:24:46,170
Speaker 1: The Supreme Court of British Columbia believed her when she

309
00:24:46,250 --> 00:24:52,330
Speaker 1: said she didn't really understand how chat GTP worked, which

310
00:24:53,010 --> 00:24:58,210
Speaker 1: I can believe too. By now, surely even the lawyers

311
00:24:58,410 --> 00:25:02,130
Speaker 1: have figured out that you can't ask chat gpt to

312
00:25:02,250 --> 00:25:07,650
Speaker 1: prepare a legal submission for you without checking. But problems

313
00:25:07,650 --> 00:25:14,890
Speaker 1: with generative AI can occur in more surprising places. Cautionary

314
00:25:14,930 --> 00:25:24,610
Speaker 1: tales will return in a moment. Jeremy Hutley, Kean Gohart,

315
00:25:24,650 --> 00:25:29,770
Speaker 1: and Henwick Verdelin are experts in ideation, or, to use

316
00:25:29,810 --> 00:25:36,090
Speaker 1: its more everyday label, brainstorming creative problem solving as a group. Naturally,

317
00:25:36,250 --> 00:25:39,690
Speaker 1: when they heard about the launch of chat GPT, they

318
00:25:39,730 --> 00:25:43,250
Speaker 1: asked themselves what this new tool might bring to the

319
00:25:43,330 --> 00:25:50,130
Speaker 1: ideation process. After all, chat GPT was a sudden sensation, powerful, flexible,

320
00:25:50,570 --> 00:25:53,810
Speaker 1: easy to use, and the problem that the lawyers had

321
00:25:53,850 --> 00:25:57,290
Speaker 1: that chat GPT just makes stuff up isn't a problem

322
00:25:57,370 --> 00:26:01,650
Speaker 1: for ideation because the aim isn't accuracy but to generate

323
00:26:01,690 --> 00:26:05,250
Speaker 1: a huge range of solutions as quickly as possible before

324
00:26:05,290 --> 00:26:09,050
Speaker 1: you work out the details later. So the three researchers

325
00:26:09,290 --> 00:26:13,170
Speaker 1: decided to conduct a simple experiment in which they compared

326
00:26:13,330 --> 00:26:18,450
Speaker 1: ideation sessions using chat GPT with ideation sessions without it.

327
00:26:19,450 --> 00:26:24,250
Speaker 1: Jeremy Utley, who teaches innovation at Stanford University, thought that

328
00:26:24,490 --> 00:26:29,490
Speaker 1: chat GPT would help teams produce vastly more ideas, maybe

329
00:26:29,530 --> 00:26:32,410
Speaker 1: twice as many, five times as many, one hundred times

330
00:26:32,450 --> 00:26:36,210
Speaker 1: as many. He told the podcast You Are Not So

331
00:26:36,250 --> 00:26:39,970
Speaker 1: Smart that he thought the question their study would answer

332
00:26:40,130 --> 00:26:47,170
Speaker 1: was how many multiples more ideas are AI assisted teams generating?

333
00:26:47,930 --> 00:26:53,050
Speaker 1: And then he saw the results, he told you were

334
00:26:53,050 --> 00:26:59,210
Speaker 1: not so Smart. My first thought was, oh no, oh no.

335
00:27:01,210 --> 00:27:04,810
Speaker 1: For many of the teams using chat GPT, the entire

336
00:27:05,090 --> 00:27:11,250
Speaker 1: collaborative back and forth of the ideation process stopped. Instead,

337
00:27:11,890 --> 00:27:15,890
Speaker 1: the room would be silent except for the pecking at keyboards.

338
00:27:16,930 --> 00:27:21,610
Speaker 1: Each person would be staring into their screen, displaying what

339
00:27:21,650 --> 00:27:27,690
Speaker 1: the researchers came to describe as resting AI face, and

340
00:27:27,730 --> 00:27:34,890
Speaker 1: the ideas they produced utterly mediocre. Equipped with the latest, greatest,

341
00:27:35,250 --> 00:27:40,570
Speaker 1: most sophisticated tool in the history of brainstorming. These teams

342
00:27:40,610 --> 00:27:48,210
Speaker 1: produced totally predictable stuff, nothing brilliant, nothing particularly varied, nothing

343
00:27:48,250 --> 00:27:51,690
Speaker 1: that didn't need a lot of development work, and above all,

344
00:27:52,090 --> 00:27:59,890
Speaker 1: just not many ideas, which is insane because ideation is

345
00:28:00,050 --> 00:28:03,570
Speaker 1: all about creating a huge variety of ideas and sorting

346
00:28:03,610 --> 00:28:08,410
Speaker 1: through them later, and chat GPT is absolutely a machine

347
00:28:08,850 --> 00:28:13,210
Speaker 1: for reducing a huge variety of ideas. It was the

348
00:28:13,530 --> 00:28:17,610
Speaker 1: ein stelling problem again. What people really needed to do

349
00:28:17,770 --> 00:28:20,690
Speaker 1: was to engage with each other and engage with the AI,

350
00:28:21,250 --> 00:28:24,490
Speaker 1: prompting it, discussing the prompts, going back to the machine,

351
00:28:24,610 --> 00:28:28,730
Speaker 1: mixing things up, varying their queries, asking for more. But

352
00:28:28,850 --> 00:28:33,130
Speaker 1: what chat gpt gave them looked a lot like a

353
00:28:33,210 --> 00:28:37,890
Speaker 1: Google search bar. You type in your question, you get

354
00:28:37,930 --> 00:28:43,290
Speaker 1: an answer, and then you stop. You feel like you've

355
00:28:43,330 --> 00:28:46,690
Speaker 1: seen this situation before, and so you do what you

356
00:28:46,810 --> 00:28:51,010
Speaker 1: always do, and if it doesn't work often you just

357
00:28:51,210 --> 00:28:59,170
Speaker 1: do it again. You get stuck. Pierre Cedric Bonan was

358
00:28:59,290 --> 00:29:03,290
Speaker 1: certainly stuck. His instinct was to pull back on the

359
00:29:03,290 --> 00:29:08,010
Speaker 1: control stick, which was stalling the plane, which was sinking, sinking,

360
00:29:08,250 --> 00:29:12,170
Speaker 1: sinking to towards the Atlantic Ocean all around him and

361
00:29:12,290 --> 00:29:18,010
Speaker 1: David Robert. Alarms were sounding, including the alarm store of

362
00:29:18,290 --> 00:29:23,610
Speaker 1: Storm Stall, but they just didn't seem to be able

363
00:29:23,610 --> 00:29:29,090
Speaker 1: to diagnose their self inflicted problem. By this time, even

364
00:29:29,170 --> 00:29:34,170
Speaker 1: the air speed indicators had defrosted. There was literally nothing

365
00:29:34,250 --> 00:29:38,330
Speaker 1: wrong with the plane. If they'd gently pointed the nose

366
00:29:38,370 --> 00:29:41,730
Speaker 1: of the plane downwards, it would have regained speed and

367
00:29:41,850 --> 00:29:46,170
Speaker 1: lift and pulled out of the storm. They had plenty

368
00:29:46,170 --> 00:29:51,250
Speaker 1: of altitude to do that, but they didn't. Robert had

369
00:29:51,410 --> 00:29:54,810
Speaker 1: pressed the button to summon Captain Dubois from the rest cabin.

370
00:29:55,530 --> 00:29:58,810
Speaker 1: Fuck where is he? In a panic, he mashed it

371
00:29:58,890 --> 00:30:04,650
Speaker 1: again and again, Fuck is he coming or not? Remember?

372
00:30:05,050 --> 00:30:08,210
Speaker 1: Captain Dubois took only ninety eight seconds to reach the

373
00:30:08,250 --> 00:30:15,250
Speaker 1: flight deck. What's happening? Dubois seemed calm given the circumstances.

374
00:30:16,010 --> 00:30:20,770
Speaker 1: David Robert and Pierre Cedric Bonan were not. Bonan had

375
00:30:20,810 --> 00:30:23,770
Speaker 1: stalled the plane, which was plummeting out of the sky

376
00:30:24,250 --> 00:30:26,810
Speaker 1: nose way up in the air at one hundred and

377
00:30:26,850 --> 00:30:31,330
Speaker 1: fifty feet per second. David Robert had noted that the

378
00:30:31,370 --> 00:30:35,730
Speaker 1: air speed indicators had failed, and although the other readings

379
00:30:35,770 --> 00:30:43,330
Speaker 1: were accurate, including the store stare stall, he didn't believe them.

380
00:30:43,770 --> 00:30:50,850
Speaker 1: The air France pilots were hideously incompetent, says William Longevisha.

381
00:30:51,450 --> 00:30:55,690
Speaker 1: Longevichha argued that the pilots simply weren't used to flying

382
00:30:55,730 --> 00:31:00,490
Speaker 1: their own airplane at altitude without the help of the computer.

383
00:31:01,330 --> 00:31:05,050
Speaker 1: Even Captain Dubois had spent only four hours in the

384
00:31:05,130 --> 00:31:10,530
Speaker 1: last six months actually flying the plane rather than supervising

385
00:31:10,570 --> 00:31:13,810
Speaker 1: the autopilot, and it had the help of the full

386
00:31:14,050 --> 00:31:18,810
Speaker 1: assistive fly by wire system. If the plane flies itself,

387
00:31:19,610 --> 00:31:27,490
Speaker 1: when do the pilots get to practice. So far, we

388
00:31:27,610 --> 00:31:32,570
Speaker 1: haven't seen that problem with modern AI systems, but it's

389
00:31:32,730 --> 00:31:36,570
Speaker 1: obvious that trouble is coming. Think of the recruiters who

390
00:31:36,610 --> 00:31:40,130
Speaker 1: fell asleep at the wheel, the lawyers who didn't understand

391
00:31:40,530 --> 00:31:45,890
Speaker 1: chat GTP, and the brainstorming groups who stared slack jawed

392
00:31:45,890 --> 00:31:49,570
Speaker 1: at their screens rather than talking to each other. In

393
00:31:49,610 --> 00:31:53,290
Speaker 1: each case we can see an all too human willingness

394
00:31:53,810 --> 00:31:57,810
Speaker 1: to abandon our own judgment and let the computer do

395
00:31:57,930 --> 00:32:01,730
Speaker 1: the thinking. And the more we do that, the less

396
00:32:01,890 --> 00:32:06,570
Speaker 1: practice we will get. Better AIS are coming, of course,

397
00:32:07,250 --> 00:32:12,810
Speaker 1: than that will only make the worse. The psychologist James Reason,

398
00:32:13,050 --> 00:32:18,170
Speaker 1: the author of Human Error, explains why skills need to

399
00:32:18,210 --> 00:32:22,490
Speaker 1: be practiced continuously in order to maintain them. Yet an

400
00:32:22,570 --> 00:32:28,130
Speaker 1: automatic control system that fails only rarely denies operators the

401
00:32:28,290 --> 00:32:34,410
Speaker 1: opportunity for practicing these basic control skills. When manual takeover

402
00:32:34,570 --> 00:32:38,930
Speaker 1: is necessary, something has usually gone wrong. This means that

403
00:32:39,010 --> 00:32:42,970
Speaker 1: operators need to be more rather than less skilled in

404
00:32:43,050 --> 00:32:47,930
Speaker 1: order to cope with these a typical conditions. This is

405
00:32:48,050 --> 00:32:54,290
Speaker 1: called the paradox of automation. Unreliable automation keeps the operators

406
00:32:54,490 --> 00:32:59,450
Speaker 1: sharp and well practiced, but the better the automated system gets,

407
00:32:59,810 --> 00:33:03,850
Speaker 1: the less experience the operators will have in doing things

408
00:33:03,850 --> 00:33:09,010
Speaker 1: for themselves, and cruelly, the weirder the situations will be.

409
00:33:09,650 --> 00:33:15,770
Speaker 1: When the computer gives up. You might say, well, then

410
00:33:15,810 --> 00:33:20,490
Speaker 1: we shouldn't use these automated systems. Pilots should practice their

411
00:33:20,530 --> 00:33:24,930
Speaker 1: skills rather than using assistive fly by wire. We should

412
00:33:25,050 --> 00:33:29,370
Speaker 1: memorize phone numbers instead of relying on our smartphones. Kids

413
00:33:29,370 --> 00:33:33,850
Speaker 1: should learn long division rather than using calculators. Heck, books

414
00:33:33,890 --> 00:33:36,330
Speaker 1: are a disaster. In the good old days, before books,

415
00:33:36,570 --> 00:33:40,730
Speaker 1: people used to just remember fifteen hour epic poems such

416
00:33:40,730 --> 00:33:47,050
Speaker 1: as the Iliad Wow, and that's not going to happen anyway.

417
00:33:47,170 --> 00:33:52,050
Speaker 1: These tools don't just make life easier, they improve our performance.

418
00:33:52,570 --> 00:33:56,610
Speaker 1: You can do more sophisticated calculations with a pocket calculator

419
00:33:56,890 --> 00:34:01,970
Speaker 1: than without one. A library can contain vastly more information

420
00:34:02,090 --> 00:34:07,610
Speaker 1: than any human could memorize, and modern planes with autopilots

421
00:34:07,770 --> 00:34:12,770
Speaker 1: and assistive flyby wire are much much safer than the

422
00:34:12,850 --> 00:34:17,650
Speaker 1: old fashioned kind. But there is a price to be paid.

423
00:34:19,090 --> 00:34:22,930
Speaker 1: Sometimes we'll find we can't remember a phone number or

424
00:34:23,250 --> 00:34:26,450
Speaker 1: how to do long division. Or perhaps we'll find we've

425
00:34:26,530 --> 00:34:29,930
Speaker 1: asked an AI system to help us brainstorm, or to

426
00:34:30,010 --> 00:34:33,970
Speaker 1: help us decide who to hire, or write new laws,

427
00:34:34,770 --> 00:34:39,050
Speaker 1: or help us control weapon systems or plan military strategy.

428
00:34:39,970 --> 00:34:46,050
Speaker 1: Maybe we stop paying attention, or become so helplessly out

429
00:34:46,050 --> 00:34:50,290
Speaker 1: of practice that when the computer lets us down, we

430
00:34:50,410 --> 00:34:58,090
Speaker 1: don't even notice. By the time Captain Mark Dubois returned

431
00:34:58,090 --> 00:35:02,250
Speaker 1: to the flight deck, it was still possible to rescue

432
00:35:02,250 --> 00:35:07,130
Speaker 1: the plane, point the nose downward, regain forward air speed,

433
00:35:07,650 --> 00:35:11,290
Speaker 1: and dive out of the store. The plane still had

434
00:35:11,410 --> 00:35:14,810
Speaker 1: enough altitude to make that possible, but Dubois would have

435
00:35:14,850 --> 00:35:17,170
Speaker 1: had to take in a lot of information in a

436
00:35:17,370 --> 00:35:21,890
Speaker 1: very short space of time to diagnose the stall, and

437
00:35:21,970 --> 00:35:25,650
Speaker 1: neither he nor Robert could directly see that Bonan was

438
00:35:25,890 --> 00:35:30,650
Speaker 1: still yanking back on the control stick instinctively trying to climb.

439
00:35:31,770 --> 00:35:34,610
Speaker 1: The plane was falling so quickly that some of the

440
00:35:34,650 --> 00:35:38,290
Speaker 1: indicators had stopped giving redoubts, and the ones that were

441
00:35:38,330 --> 00:35:42,090
Speaker 1: working might have seemed unbelievable because of the extreme speed

442
00:35:42,130 --> 00:35:46,810
Speaker 1: of the fall. And then there's a fundamental ambiguity in

443
00:35:46,890 --> 00:35:51,770
Speaker 1: a stall. You're pointing up, but you're falling down. To

444
00:35:51,890 --> 00:35:56,290
Speaker 1: stop descending, you'd first have to dive. That can make

445
00:35:56,370 --> 00:36:00,170
Speaker 1: it difficult both to diagnose the problem and to talk

446
00:36:00,210 --> 00:36:05,090
Speaker 1: about it. Less than a minute after Captain Dubois entered

447
00:36:05,090 --> 00:36:10,210
Speaker 1: the flight deck, there's an exchange. Robert says, you're climbing.

448
00:36:11,090 --> 00:36:15,450
Speaker 1: Then he says, you're going down, down, down, down. Is

449
00:36:15,490 --> 00:36:19,130
Speaker 1: that an instruction to point the nose down or a

450
00:36:19,170 --> 00:36:23,690
Speaker 1: description of the plane which is falling fast? Captain Dubois

451
00:36:23,770 --> 00:36:28,250
Speaker 1: echoes going down. Bonan asks, am I going down now?

452
00:36:28,810 --> 00:36:34,330
Speaker 1: Robert and Dubois both disagree. Robert answers, go down. Dubois says, no,

453
00:36:34,610 --> 00:36:38,610
Speaker 1: you climb here. That's a description, not an order. Robert adds,

454
00:36:38,810 --> 00:36:42,610
Speaker 1: go down. Bonan says, I'm climbing, okay, So we're going down.

455
00:36:43,970 --> 00:36:47,850
Speaker 1: It's a mess. Are they climbing or going down? Both?

456
00:36:48,770 --> 00:36:53,250
Speaker 1: The nose is pointed up Bonan stick is back, and

457
00:36:53,290 --> 00:36:57,050
Speaker 1: they're falling at more than ten thousand feet a minute.

458
00:36:58,170 --> 00:37:02,410
Speaker 1: Maybe Captain Dubois has realized their story. Maybe not. He

459
00:37:02,490 --> 00:37:07,010
Speaker 1: doesn't say so directly, and he's not at the controls Bananas.

460
00:37:07,970 --> 00:37:14,810
Speaker 1: All the while, computer voice is adding store stall. It

461
00:37:14,890 --> 00:37:18,970
Speaker 1: takes another minute before there's some kind of clarity. The

462
00:37:19,010 --> 00:37:23,930
Speaker 1: plane has fallen through the ten thousand feet mark. There's

463
00:37:23,970 --> 00:37:31,290
Speaker 1: now less than a minute left. Robert says, Climb, climb, climb, climb,

464
00:37:31,690 --> 00:37:35,490
Speaker 1: Of course he does. The plane is plummeting. Banan replies,

465
00:37:35,810 --> 00:37:38,250
Speaker 1: but I've been at maxy nose up for a while.

466
00:37:39,130 --> 00:37:45,010
Speaker 1: At last, Captain Dubois seems to understand what Banan has done. No, no,

467
00:37:45,090 --> 00:37:49,650
Speaker 1: don't climb. At this point, David Robert pushes a button

468
00:37:49,890 --> 00:37:53,570
Speaker 1: to switch controlled his seat and pushes the nose of

469
00:37:53,610 --> 00:37:59,210
Speaker 1: the plane down. Banan, presumably panicking, pushes his button and

470
00:37:59,330 --> 00:38:02,890
Speaker 1: silently takes back control of the aircraft and sticks the

471
00:38:02,930 --> 00:38:07,210
Speaker 1: nose back up. It doesn't matter, it's too late anyway,

472
00:38:08,170 --> 00:38:14,570
Speaker 1: They only have seconds left. Pierre Cedric Bonan's wife is

473
00:38:14,770 --> 00:38:18,970
Speaker 1: back in the passenger cabin. Their two young sons are

474
00:38:19,010 --> 00:38:23,170
Speaker 1: back in Paris. Does Bonan realize they're about to be orphaned?

475
00:38:24,330 --> 00:38:30,050
Speaker 1: Probably we're going to crash, he says. This can't be true. Fuck,

476
00:38:30,370 --> 00:38:35,050
Speaker 1: we're dead, says David. Robert. In less than three seconds,

477
00:38:35,850 --> 00:38:40,650
Speaker 1: the plane will barely flop into the Atlantic Ocean, instantly

478
00:38:40,730 --> 00:38:43,850
Speaker 1: killing all two hundred and twenty eight people on board,

479
00:38:44,690 --> 00:38:54,290
Speaker 1: Robert Bonar, Captain Dubois, Bonan's wife, Dubois girlfriend, everyone, Pierre

480
00:38:54,290 --> 00:39:21,610
Speaker 1: Cedric Bonan's last words, but what's happening? For a full

481
00:39:21,610 --> 00:39:25,250
Speaker 1: list of our sources, see the show notes at Timharford

482
00:39:25,570 --> 00:39:34,890
Speaker 1: dot com. Cautionary Tales is written by me Tim Harford

483
00:39:34,930 --> 00:39:38,370
Speaker 1: with Andrew Wright. It's produced by Alice Fines with support

484
00:39:38,410 --> 00:39:42,050
Speaker 1: from Marilyn Rust. The sound design and original music is

485
00:39:42,050 --> 00:39:46,170
Speaker 1: the work of Pascal Wise. Sarah Nix edited the scripts.

486
00:39:46,850 --> 00:39:50,050
Speaker 1: It features the voice talents of Ben Crowe, Melanie Guttridge,

487
00:39:50,290 --> 00:39:54,850
Speaker 1: Stella Harford, Jemma Saunders, and Rufus Wright. The show also

488
00:39:54,970 --> 00:39:57,970
Speaker 1: wouldn't have been possible without the work of Jacob Weisberg,

489
00:39:58,210 --> 00:40:03,650
Speaker 1: Ryan Dilly, Greta Cohne, Eric's handler, Carrie Brody, and Christina Sullivan.

490
00:40:04,410 --> 00:40:08,970
Speaker 1: Cautionary Tales is a production of Pushkin Industries. It's recorded

491
00:40:08,970 --> 00:40:12,810
Speaker 1: a Wardour Studios in London by Tom Berry. If you

492
00:40:13,010 --> 00:40:17,210
Speaker 1: like the show, please remember to share, rate and review,

493
00:40:17,770 --> 00:40:19,930
Speaker 1: tell your friends and if you want to hear the

494
00:40:19,930 --> 00:40:23,410
Speaker 1: show ad free, sign up for Pushkin Plus on the

495
00:40:23,450 --> 00:40:27,690
Speaker 1: show page, if Apple Podcasts, or at pushkin dot fm,

496
00:40:27,890 --> 00:40:29,010
Speaker 1: slash plus