1
00:00:00,100 --> 00:00:02,399
一篇Substack文章引爆华尔街

2
00:00:02,399 --> 00:00:04,480
AI会杀死经济吗

3
00:00:04,519 --> 00:00:06,740
一篇发表在Substack上的文章

4
00:00:06,740 --> 00:00:08,138
发布不到24小时

5
00:00:08,140 --> 00:00:09,860
就引发了华尔街的集体恐慌

6
00:00:10,660 --> 00:00:13,320
IBM单日暴跌13%

7
00:00:13,320 --> 00:00:15,480
美国运通跌超8%

8
00:00:15,480 --> 00:00:15,880
黑石

9
00:00:15,880 --> 00:00:15,890
DoorDash

10
00:00:15,880 --> 00:00:19,460
Uber、Visa、万事达卡全线崩塌

11
00:00:19,460 --> 00:00:21,400
道琼斯狂泻超过820点

12
00:00:21,800 --> 00:00:23,940
但这不是什么央行政策突变

13
00:00:23,940 --> 00:00:25,980
不是哪家巨头财报暴雷

14
00:00:25,980 --> 00:00:29,261
甚至不是一份来自高盛或摩根士丹利的正式研报

15
00:00:29,859 --> 00:00:30,940
引发这场地震的

16
00:00:30,940 --> 00:00:32,600
是一个叫Citrini Research的

17
00:00:32,600 --> 00:00:34,039
独立分析机构

18
00:00:34,039 --> 00:00:36,460
在Substack上发的一篇博客文章

19
00:00:36,579 --> 00:00:39,480
标题叫<2028全球智能危机>

20
00:00:40,079 --> 00:00:41,219
更魔幻的是

21
00:00:41,219 --> 00:00:43,399
这篇文章的作者自己都说了

22
00:00:43,399 --> 00:00:44,500
这不是预测

23
00:00:44,500 --> 00:00:45,600
这是一个思想实验

24
00:00:46,039 --> 00:00:47,259
一个思想实验

25
00:00:47,259 --> 00:00:48,940
干掉了几千亿美元的市值

26
00:00:49,500 --> 00:00:50,259
你不觉得

27
00:00:50,259 --> 00:00:51,399
这件事本身

28
00:00:51,399 --> 00:00:53,390
比文章的内容更值得深思吗

29
00:00:53,939 --> 00:00:54,219
哈喽

30
00:00:54,219 --> 00:00:54,659
大家好

31
00:00:54,659 --> 00:00:55,160
我是王利杰

32
00:00:55,619 --> 00:00:58,760
今天我不打算简单地给你转述这篇报告说了什么

33
00:00:59,240 --> 00:01:02,460
我想带你思考三个更深层的问题：第一

34
00:01:02,460 --> 00:01:05,299
这篇文章的核心逻辑链条到底是什么

35
00:01:05,299 --> 00:01:06,459
它有没有道理

36
00:01:06,500 --> 00:01:07,159
第二

37
00:01:07,159 --> 00:01:10,760
为什么一篇自称"思想实验"的文章能掀翻华尔街

38
00:01:10,799 --> 00:01:11,480
第三

39
00:01:11,480 --> 00:01:14,200
也是我认为最重要的 - 在AI时代"

40
00:01:14,299 --> 00:01:16,519
叙事"本身是不是已经成为了一种

41
00:01:16,519 --> 00:01:18,219
比技术更强大的力量

42
00:01:18,299 --> 00:01:19,061
先说第一个问题

43
00:01:19,459 --> 00:01:21,519
这篇报告到底在讲什么

44
00:01:21,560 --> 00:01:23,379
Citrini Research的创始人叫

45
00:01:23,379 --> 00:01:24,838
詹姆斯·范·吉伦

46
00:01:24,840 --> 00:01:27,959
一个33岁的前洛杉矶急救医护人员

47
00:01:27,959 --> 00:01:29,681
拥有生物学和心理学学位

48
00:01:30,159 --> 00:01:32,200
他不是华尔街的老牌大佬

49
00:01:32,200 --> 00:01:35,379
但他在2022年底公开做空硅谷银行

50
00:01:35,379 --> 00:01:37,621
几个月后硅谷银行真的倒了

51
00:01:37,760 --> 00:01:39,219
靠着这一战成名

52
00:01:39,219 --> 00:01:40,840
他在Substack上把Citrini

53
00:01:40,879 --> 00:01:42,659
Research做成了平台上

54
00:01:42,659 --> 00:01:45,180
最受欢迎的金融类newsletter之一

55
00:01:45,180 --> 00:01:46,939
拥有超过十万订阅者

56
00:01:46,939 --> 00:01:49,900
自称投资组合自2023年5月以来

57
00:01:49,900 --> 00:01:51,200
回报超过200%

58
00:01:51,640 --> 00:01:53,299
共同作者阿拉普·沙阿

59
00:01:53,299 --> 00:01:54,700
曾在对冲基金Viking

60
00:01:54,799 --> 00:01:56,420
Global和Citadel

61
00:01:56,420 --> 00:01:57,599
做过分析师

62
00:01:57,599 --> 00:02:00,680
还联合创办了金融AI搜索平台Sentieo

63
00:02:00,680 --> 00:02:03,280
现在是AI助手公司Littlebird的CEO

64
00:02:03,760 --> 00:02:05,640
他们用了一个非常聪明的写作手法

65
00:02:06,120 --> 00:02:09,080
整篇文章是以2028年6月的视角写的

66
00:02:09,080 --> 00:02:11,480
假装在回顾一场已经发生的经济危机

67
00:02:11,860 --> 00:02:13,859
就像你穿越到了2028年

68
00:02:13,860 --> 00:02:15,419
翻开一份金融备忘录

69
00:02:15,419 --> 00:02:19,120
上面写着：今天早上失业率公布为10.2%

70
00:02:19,120 --> 00:02:21,859
标普500从2026年10月的高点

71
00:02:21,860 --> 00:02:23,479
已经暴跌了38%

72
00:02:24,039 --> 00:02:27,040
然后他们开始"回顾"这场危机是怎么一步步发生的

73
00:02:27,699 --> 00:02:28,219
第一环

74
00:02:28,219 --> 00:02:29,219
叫"幽灵GDP"

75
00:02:29,819 --> 00:02:32,700
这个概念是整篇报告最精彩的发明

76
00:02:32,780 --> 00:02:33,580
他们说，AI

77
00:02:33,580 --> 00:02:36,960
带来的生产力爆炸会让GDP数据非常好看

78
00:02:36,960 --> 00:02:38,299
企业利润创新高

79
00:02:38,300 --> 00:02:40,900
股市狂飙 - 标普冲到8000点

80
00:02:40,900 --> 00:02:42,501
纳斯达克突破30000点

81
00:02:43,259 --> 00:02:44,379
但问题是，AI

82
00:02:44,379 --> 00:02:45,320
机器不消费

83
00:02:45,460 --> 00:02:46,379
它们不买房

84
00:02:46,379 --> 00:02:46,979
不吃饭

85
00:02:46,979 --> 00:02:47,860
不看电影

86
00:02:47,860 --> 00:02:48,920
不给孩子报补习班

87
00:02:49,740 --> 00:02:51,099
生产力是上去了

88
00:02:51,099 --> 00:02:54,820
但这些产出没有通过工资流向消费者的口袋

89
00:02:54,879 --> 00:02:56,319
GDP数据漂亮

90
00:02:56,319 --> 00:02:58,420
但钱没有在经济体里真正循环

91
00:02:58,979 --> 00:03:02,020
这就是"幽灵GDP" - 账面上的繁荣

92
00:03:02,020 --> 00:03:03,061
实质上的空心化

93
00:03:03,460 --> 00:03:04,280
第二环

94
00:03:04,280 --> 00:03:05,820
商业模式的系统性坍塌

95
00:03:06,419 --> 00:03:09,039
这里才是真正让华尔街肉疼的部分

96
00:03:09,039 --> 00:03:10,501
因为报告直接点名了

97
00:03:11,080 --> 00:03:11,840
它说，AI

98
00:03:11,840 --> 00:03:13,941
代理会成为消费者的全能管家

99
00:03:14,520 --> 00:03:17,099
你以后不需要自己去Booking上订酒店

100
00:03:17,099 --> 00:03:18,639
不需要自己去比价

101
00:03:18,639 --> 00:03:19,820
不需要自己续保险

102
00:03:20,180 --> 00:03:21,799
AI帮你搞定一切

103
00:03:21,800 --> 00:03:23,979
而且它不受品牌忠诚度的影响

104
00:03:23,979 --> 00:03:25,379
不受广告的影响

105
00:03:25,379 --> 00:03:26,540
它只认性价比

106
00:03:26,639 --> 00:03:29,439
报告里有一句话特别扎心 - "结果表明

107
00:03:29,439 --> 00:03:31,020
很多人们所谓的'关系'

108
00:03:31,020 --> 00:03:32,881
不过是带着友善面具的'摩擦'

109
00:03:33,329 --> 00:03:34,099
"什么意思

110
00:03:34,240 --> 00:03:36,280
你和你的保险经纪人的"关系"

111
00:03:36,280 --> 00:03:38,381
本质上是因为你懒得自己去比价

112
00:03:38,800 --> 00:03:40,679
你对某个旅游平台的"忠诚"

113
00:03:40,680 --> 00:03:42,721
本质上是因为你没时间货比三家

114
00:03:43,379 --> 00:03:45,180
当AI消灭了这些摩擦

115
00:03:45,180 --> 00:03:47,740
建立在信息不对称和人类惰性之上的

116
00:03:47,740 --> 00:03:49,341
整个商业生态就会瓦解

117
00:03:49,759 --> 00:03:51,539
旅游预订平台、保险续保

118
00:03:51,719 --> 00:03:53,829
房产经纪、财务咨询、税务服务

119
00:03:54,419 --> 00:03:57,299
所有价值主张本质上是"我帮你处理你

120
00:03:57,300 --> 00:03:58,680
觉得麻烦的事"的行业

121
00:03:58,680 --> 00:04:00,401
都会面临AI的降维打击

122
00:04:01,139 --> 00:04:02,381
更狠的是支付网络

123
00:04:02,860 --> 00:04:06,120
报告预测AI交易会绕过信用卡网络

124
00:04:06,120 --> 00:04:07,521
用低成本的稳定币结算

125
00:04:08,259 --> 00:04:12,060
Visa和万事达卡那2%到3%的手续费

126
00:04:12,099 --> 00:04:14,361
AI觉得这是不可接受的摩擦成本

127
00:04:15,080 --> 00:04:17,019
难怪报告发布当天，Visa

128
00:04:17,019 --> 00:04:19,040
和万事达卡的股价双双跳水

129
00:04:19,420 --> 00:04:20,158
第三环

130
00:04:20,160 --> 00:04:21,980
白领失业引发消费崩溃

131
00:04:22,500 --> 00:04:23,640
这一环打击面最广

132
00:04:24,060 --> 00:04:26,519
美国经济的核心驱动力是消费

133
00:04:26,519 --> 00:04:30,540
而白领群体贡献了大约75%的可自由支配消费支出

134
00:04:31,379 --> 00:04:33,800
AI大规模替代白领工作之后

135
00:04:33,800 --> 00:04:36,439
大量高技能人才被迫涌入零工经济

136
00:04:36,439 --> 00:04:37,680
跑外卖、开网约车

137
00:04:38,310 --> 00:04:41,800
但他们的涌入会压低整个零工市场的工资水平

138
00:04:41,800 --> 00:04:44,500
原本就在做这些工作的人收入也跟着下降

139
00:04:45,019 --> 00:04:45,639
而且

140
00:04:45,639 --> 00:04:47,379
即使那些还没被裁的白领

141
00:04:47,379 --> 00:04:49,598
看到身边的同事一个个消失

142
00:04:49,600 --> 00:04:52,300
也会本能地收紧钱包、增加储蓄

143
00:04:52,420 --> 00:04:54,700
消费意愿和消费能力同时萎缩

144
00:04:55,300 --> 00:04:56,300
第四环

145
00:04:56,300 --> 00:04:57,640
金融体系的连锁崩塌

146
00:04:58,339 --> 00:05:01,519
这里报告讲了一个大多数人完全不了解的暗雷

147
00:05:02,060 --> 00:05:02,980
私募信贷市场

148
00:05:03,560 --> 00:05:04,319
过去十年

149
00:05:04,319 --> 00:05:06,240
这个市场膨胀得非常快

150
00:05:06,240 --> 00:05:08,860
其中大量贷款流向了SaaS软件公司

151
00:05:09,339 --> 00:05:10,000
为什么

152
00:05:10,040 --> 00:05:12,339
因为SaaS公司有"经常性收入"

153
00:05:12,339 --> 00:05:13,920
客户签了年度合同

154
00:05:13,920 --> 00:05:15,459
收入可预测性很强

155
00:05:15,459 --> 00:05:17,260
是信贷市场最喜欢的抵押品

156
00:05:17,800 --> 00:05:19,300
但AI来了之后呢

157
00:05:19,339 --> 00:05:21,300
如果企业发现可以用AI代理

158
00:05:21,300 --> 00:05:21,790
替代掉ServiceNow

159
00:05:21,800 --> 00:05:25,319
Salesforce这些SaaS工具的功能

160
00:05:25,319 --> 00:05:26,780
续约率就会断崖式下跌

161
00:05:27,300 --> 00:05:28,980
SaaS公司收入锐减

162
00:05:28,980 --> 00:05:30,440
私募信贷大面积违约

163
00:05:31,000 --> 00:05:33,300
而普通老百姓的养老金和保险金

164
00:05:33,300 --> 00:05:35,580
很多都间接投资了这些私募信贷产品

165
00:05:36,300 --> 00:05:36,959
第五环

166
00:05:36,959 --> 00:05:38,060
也是最恐怖的一环

167
00:05:38,519 --> 00:05:40,680
没有自然制动机制的负反馈循环

168
00:05:41,360 --> 00:05:43,680
传统经济衰退有一个自我修复的逻辑

169
00:05:43,680 --> 00:05:47,899
企业裁员→工资下降→成本降低→利润恢复→

170
00:05:47,899 --> 00:05:48,540
重新招人

171
00:05:49,259 --> 00:05:51,098
但在AI驱动的衰退中

172
00:05:51,100 --> 00:05:54,100
企业裁员省下的钱不会用来重新招人

173
00:05:54,100 --> 00:05:56,180
而是用来买更多的AI算力

174
00:05:56,220 --> 00:05:57,540
AI能力提升了

175
00:05:57,540 --> 00:05:59,040
又可以裁掉更多人

176
00:05:59,199 --> 00:06:02,899
裁员→投资AI→AI更强→裁更多人→

177
00:06:02,899 --> 00:06:03,759
投资更多AI

178
00:06:04,360 --> 00:06:06,598
这是一个正反馈的死亡螺旋

179
00:06:06,600 --> 00:06:07,840
没有自然的刹车机制

180
00:06:08,259 --> 00:06:11,658
这个逻辑链条我之前在讲"正反馈死亡螺旋"的那期

181
00:06:11,660 --> 00:06:13,839
节目里详细拆解过

182
00:06:13,839 --> 00:06:16,500
本质上和情绪-炎症的恶性循环、雪崩

183
00:06:16,620 --> 00:06:18,439
核裂变是同一种动力学结构

184
00:06:18,939 --> 00:06:20,939
输出绕回来放大偏差

185
00:06:20,939 --> 00:06:21,480
系统失控

186
00:06:21,879 --> 00:06:22,279
好

187
00:06:22,279 --> 00:06:22,850
五环讲完了

188
00:06:23,379 --> 00:06:25,579
这个逻辑链条有没有道理

189
00:06:25,759 --> 00:06:26,740
坦率地说

190
00:06:26,740 --> 00:06:28,661
每一环单独拿出来都有道理

191
00:06:29,100 --> 00:06:30,579
AI替代白领工作

192
00:06:30,699 --> 00:06:31,240
正在发生

193
00:06:31,819 --> 00:06:34,079
IBM当天暴跌13%

194
00:06:34,079 --> 00:06:36,120
直接导火索是Anthropic

195
00:06:36,120 --> 00:06:37,660
同一天宣布它的Claude

196
00:06:37,720 --> 00:06:40,699
Code工具可以自动化COBOL代码的现代化改造

197
00:06:41,220 --> 00:06:43,718
这是一个每年300亿美元的市场

198
00:06:43,720 --> 00:06:45,510
是IBM最肥的利润来源之一

199
00:06:46,300 --> 00:06:48,540
而Citrini报告引发的恐慌情绪

200
00:06:48,540 --> 00:06:50,740
则在整个市场层面火上浇油

201
00:06:51,339 --> 00:06:52,939
商业摩擦被AI消灭

202
00:06:53,060 --> 00:06:53,661
也在发生

203
00:06:54,240 --> 00:06:55,519
消费萎缩的风险

204
00:06:55,639 --> 00:06:56,280
逻辑上成立

205
00:06:56,740 --> 00:06:58,019
私募信贷的暗雷

206
00:06:58,040 --> 00:06:58,680
确实存在

207
00:06:59,160 --> 00:07:02,899
但问题在于 - 这五环需要在两年内同时发生

208
00:07:03,019 --> 00:07:05,540
同时失控、而且政府完全不作为

209
00:07:06,060 --> 00:07:08,800
知名经济学博主诺亚·史密斯说得好

210
00:07:08,800 --> 00:07:11,250
这篇文章本质上是"两个论点合一"

211
00:07:11,939 --> 00:07:12,860
微观层面

212
00:07:12,860 --> 00:07:14,779
哪些行业会被AI颠覆

213
00:07:14,899 --> 00:07:15,860
这个论点很有力

214
00:07:16,319 --> 00:07:17,240
宏观层面

215
00:07:17,240 --> 00:07:19,720
这些颠覆会不会导致经济崩溃

216
00:07:19,839 --> 00:07:21,140
这个论点就弱得多了

217
00:07:21,660 --> 00:07:23,500
因为它假设了三个变量不动

218
00:07:23,620 --> 00:07:27,339
技术的扩散速度不会受阻、流失的工资不会被重新分配

219
00:07:27,920 --> 00:07:30,149
政府会坐视税基侵蚀而无动于衷

220
00:07:30,720 --> 00:07:33,260
同时让AI的能力无限制地加速

221
00:07:33,959 --> 00:07:34,879
历史告诉我们

222
00:07:34,879 --> 00:07:37,060
通用技术的普及从来不是一条直线

223
00:07:37,579 --> 00:07:41,360
电力的发明到电力真正改变工厂的生产方式

224
00:07:41,360 --> 00:07:42,440
中间隔了三十年

225
00:07:42,800 --> 00:07:46,360
互联网的发明到互联网真正改变商业模式

226
00:07:46,360 --> 00:07:47,500
中间也隔了十多年

227
00:07:48,000 --> 00:07:49,480
不是因为技术不行

228
00:07:49,480 --> 00:07:52,519
而是因为组织需要重构、流程需要再造

229
00:07:52,680 --> 00:07:55,380
人员需要培训、监管需要跟上

230
00:07:55,439 --> 00:07:56,080
AI也一样

231
00:07:56,660 --> 00:07:59,019
从"AI能做某件事"到"企业真正

232
00:07:59,019 --> 00:08:01,240
大规模部署AI替代人力"

233
00:08:01,240 --> 00:08:03,280
中间的鸿沟比很多人想象的要大

234
00:08:03,860 --> 00:08:04,199
而且

235
00:08:04,199 --> 00:08:07,439
农业机械化在20世纪彻底改变了美国农业

236
00:08:07,439 --> 00:08:10,790
就业占比从约40%下降到今天不足2%

237
00:08:11,420 --> 00:08:13,379
但整体经济并没有崩溃

238
00:08:13,379 --> 00:08:16,680
因为新的行业和新的就业机会被创造出来了

239
00:08:16,740 --> 00:08:18,360
这是乐观派最有力的论据

240
00:08:18,920 --> 00:08:19,899
但话说回来

241
00:08:19,899 --> 00:08:22,100
反对派的论据也有一个致命的弱点

242
00:08:22,540 --> 00:08:24,161
他们假设这次和以前一样

243
00:08:24,620 --> 00:08:26,100
以前每次技术革命

244
00:08:26,100 --> 00:08:28,911
被淘汰的工人可以转向新的领域重新学习

245
00:08:29,560 --> 00:08:31,040
但AI的特殊性在于

246
00:08:31,040 --> 00:08:32,659
它是一种通用智能

247
00:08:32,720 --> 00:08:34,440
你转向新领域，AI

248
00:08:34,440 --> 00:08:35,860
在那个领域也在变强

249
00:08:36,440 --> 00:08:38,178
这就像你和一个对手赛跑

250
00:08:38,179 --> 00:08:39,399
你换了一条赛道

251
00:08:39,399 --> 00:08:41,419
发现它已经在那条赛道上等着你了

252
00:08:41,879 --> 00:08:43,640
这是历史上从未出现过的情况

253
00:08:44,158 --> 00:08:45,580
所以我的判断是什么

254
00:08:45,700 --> 00:08:48,240
报告描述的每一个风险都是真实的

255
00:08:48,240 --> 00:08:50,240
但它给出的时间线太激进了

256
00:08:50,820 --> 00:08:52,159
两年内全面崩溃

257
00:08:52,200 --> 00:08:52,961
可能性不大

258
00:08:53,559 --> 00:08:54,679
但五到十年内

259
00:08:54,679 --> 00:08:56,159
如果我们不做准备

260
00:08:56,159 --> 00:08:58,519
报告中描述的很多场景完全可能

261
00:08:58,519 --> 00:09:00,721
以更慢但更确定的方式展开

262
00:09:01,200 --> 00:09:02,600
这更像是气候变化

263
00:09:02,600 --> 00:09:05,639
而不是新冠疫情 - 影响是深远的

264
00:09:05,639 --> 00:09:06,680
但节奏是渐进的

265
00:09:07,200 --> 00:09:07,500
好

266
00:09:07,500 --> 00:09:09,039
现在来聊第二个问题

267
00:09:09,039 --> 00:09:10,821
也是我认为真正有意思的问题

268
00:09:11,399 --> 00:09:13,299
为什么一篇自称"思想实验"的

269
00:09:13,299 --> 00:09:14,539
Substack文章

270
00:09:14,539 --> 00:09:16,620
能干掉几千亿美元的市值

271
00:09:16,740 --> 00:09:19,840
白宫经济顾问委员会代理主席皮埃尔·亚雷德

272
00:09:19,840 --> 00:09:21,799
接受彭博社采访时说

273
00:09:21,799 --> 00:09:24,559
这篇报告是"一部有趣的科幻作品"

274
00:09:24,559 --> 00:09:27,000
违反了经济学中一些基本的恒等关系

275
00:09:27,460 --> 00:09:28,620
话虽如此

276
00:09:28,620 --> 00:09:30,821
道琼斯已经跌了超过820点了

277
00:09:31,190 --> 00:09:32,860
"大空头"迈克尔·伯瑞

278
00:09:32,980 --> 00:09:36,100
就是那个预测了2008年金融危机的传奇投资人

279
00:09:36,600 --> 00:09:38,440
在推特上转发了这篇文章

280
00:09:38,440 --> 00:09:41,039
配了一句话："你们还觉得我看空

281
00:09:41,059 --> 00:09:43,020
"这条推文被浏览了数百万次

282
00:09:43,519 --> 00:09:46,279
那你就要问了：市场到底在害怕什么

283
00:09:46,379 --> 00:09:49,139
是在害怕这篇文章说的事情会发生吗

284
00:09:49,240 --> 00:09:50,840
还是在害怕别的东西

285
00:09:50,940 --> 00:09:51,438
我认为

286
00:09:51,440 --> 00:09:53,820
市场害怕的不是这篇文章的预测

287
00:09:53,820 --> 00:09:55,759
而是这篇文章暴露出来的一个事实

288
00:09:56,240 --> 00:09:59,640
没有人知道AI到底会对经济产生多大的影响

289
00:09:59,799 --> 00:10:02,259
彭博社有一篇评论的标题说得特别直白

290
00:10:02,259 --> 00:10:04,900
"华尔街已经不知道该怎么看AI了

291
00:10:05,370 --> 00:10:05,821
"想想看

292
00:10:06,159 --> 00:10:06,899
过去两年

293
00:10:06,899 --> 00:10:09,000
华尔街的主流叙事是什么

294
00:10:09,039 --> 00:10:12,179
AI是人类历史上最伟大的生产力革命

295
00:10:12,179 --> 00:10:15,539
买英伟达、买微软、买一切跟AI沾边的东西

296
00:10:15,539 --> 00:10:16,240
涨到天上去

297
00:10:16,720 --> 00:10:19,500
这个叙事支撑了万亿美元级别的市值膨胀

298
00:10:19,940 --> 00:10:20,539
但现在

299
00:10:20,539 --> 00:10:22,240
一个完全相反的叙事出现了

300
00:10:22,240 --> 00:10:24,809
AI确实会兑现所有的技术预期

301
00:10:25,299 --> 00:10:27,100
然后正是这个兑现的过程

302
00:10:27,100 --> 00:10:27,961
会摧毁经济

303
00:10:28,440 --> 00:10:30,960
"如果我们对AI的乐观判断继续被证明

304
00:10:30,960 --> 00:10:33,779
是对的……那如果这恰恰是利空呢

305
00:10:33,789 --> 00:10:36,520
"这是Citrini在推特上发布文章时写的导语

306
00:10:37,080 --> 00:10:38,559
这句话太精准了

307
00:10:38,559 --> 00:10:40,020
精准到让人脊背发凉

308
00:10:40,500 --> 00:10:43,620
因为它触及了一个大多数人从未想过的可能性

309
00:10:43,620 --> 00:10:46,399
AI最大的风险不是它失败

310
00:10:46,399 --> 00:10:47,140
而是它成功

311
00:10:47,960 --> 00:10:50,820
这就引出了我今天最想跟大家聊的第三个问题

312
00:10:51,259 --> 00:10:51,941
叙事的力量

313
00:10:52,500 --> 00:10:55,539
2013年诺贝尔经济学奖得主罗伯特·席勒

314
00:10:55,539 --> 00:10:56,460
写过一本书叫

315
00:10:56,460 --> 00:10:57,600
<叙事经济学>

316
00:10:58,299 --> 00:10:59,379
他的核心观点是

317
00:10:59,379 --> 00:11:02,860
经济波动很多时候不是由基本面驱动的

318
00:11:02,860 --> 00:11:05,639
而是由那些被过度简化、容易传播

319
00:11:05,759 --> 00:11:07,961
触动情绪的经济叙事驱动的

320
00:11:08,379 --> 00:11:12,200
一个好的叙事就像病毒一样 - 它不需要完全正确

321
00:11:12,200 --> 00:11:15,640
它只需要足够可信、足够吓人、足够容易复述

322
00:11:16,539 --> 00:11:18,779
乔治·索罗斯有一个更直白的说法

323
00:11:18,779 --> 00:11:19,600
叫"反身性"

324
00:11:20,299 --> 00:11:23,360
市场参与者的认知不是被动地反映现实

325
00:11:23,360 --> 00:11:24,900
而是主动地塑造现实

326
00:11:25,700 --> 00:11:27,700
当足够多的人相信一个叙事

327
00:11:27,700 --> 00:11:29,380
这个叙事就会变成现实

328
00:11:30,120 --> 00:11:31,539
Citrini的这篇报告

329
00:11:31,539 --> 00:11:33,620
就是一个教科书级别的叙事武器

330
00:11:34,159 --> 00:11:35,480
它有完整的逻辑链条

331
00:11:35,860 --> 00:11:37,701
它有具体的公司名称和数字

332
00:11:38,159 --> 00:11:40,429
它触及了每个打工人最深层的恐惧

333
00:11:40,919 --> 00:11:42,341
我的工作会不会消失

334
00:11:42,879 --> 00:11:46,679
它用了一个极其聪明的"来自未来的回顾"写作手法

335
00:11:46,679 --> 00:11:50,081
让读者在心理上产生了一种"这已经发生了"的错觉

336
00:11:50,539 --> 00:11:51,780
它不需要完全正确

337
00:11:52,240 --> 00:11:55,780
它只需要足够可信、足够可怕、足够可传播

338
00:11:56,340 --> 00:11:59,779
而市场的反应完美地验证了索罗斯的反身性理论

339
00:12:00,340 --> 00:12:03,340
一篇文章改变了市场参与者的预期

340
00:12:03,340 --> 00:12:06,019
预期的改变导致了实际的抛售

341
00:12:06,019 --> 00:12:08,960
实际的抛售又强化了文章的"可信度"

342
00:12:08,960 --> 00:12:10,841
更多人读到文章后加入抛售

343
00:12:11,360 --> 00:12:13,321
叙事创造了它所描述的现实

344
00:12:13,840 --> 00:12:14,860
更讽刺的是

345
00:12:14,860 --> 00:12:18,240
报告两位作者的投资组合恰恰是这场恐慌的受益者

346
00:12:18,740 --> 00:12:21,940
他们持有大量半导体和AI算力相关的股票

347
00:12:22,080 --> 00:12:23,080
英伟达、Alphabet

348
00:12:23,460 --> 00:12:26,720
这些公司在"AI屠宰经济"的叙事中是最终的赢家

349
00:12:27,570 --> 00:12:30,360
他们做空的是那些被AI颠覆的传统企业

350
00:12:30,940 --> 00:12:31,759
换句话说

351
00:12:31,759 --> 00:12:33,960
如果他们的叙事被市场接受

352
00:12:33,960 --> 00:12:37,399
传统企业股价下跌、AI算力公司继续上涨

353
00:12:37,399 --> 00:12:38,180
他们两头赚

354
00:12:38,639 --> 00:12:40,701
我不是说他们是故意操纵市场

355
00:12:40,820 --> 00:12:43,600
范·吉伦自己都说市场的反应"远超预期"

356
00:12:44,139 --> 00:12:45,059
但这个利益结构

357
00:12:45,059 --> 00:12:45,660
你需要知道

358
00:12:46,100 --> 00:12:47,159
就像我一直说的

359
00:12:47,159 --> 00:12:48,460
读任何研究报告

360
00:12:48,460 --> 00:12:50,179
第一件事要看的不是结论

361
00:12:50,179 --> 00:12:51,150
而是作者的仓位

362
00:12:51,879 --> 00:12:53,420
但这里有一个更深层的悖论

363
00:12:53,879 --> 00:12:56,500
即使我们知道这篇报告有利益倾向

364
00:12:56,559 --> 00:12:58,779
即使白宫说这是科幻小说

365
00:12:58,779 --> 00:13:02,071
即使诺亚·史密斯说这不过是一个"吓人的睡前故事"

366
00:13:02,460 --> 00:13:04,581
它依然造成了实实在在的市场冲击

367
00:13:04,960 --> 00:13:05,620
为什么

368
00:13:05,700 --> 00:13:08,720
因为它说出了一个所有人都隐约感觉到

369
00:13:08,720 --> 00:13:10,961
但没有人敢系统化表达的恐惧

370
00:13:11,419 --> 00:13:14,040
市场不是因为相信了一篇文章而崩盘

371
00:13:14,620 --> 00:13:18,500
市场是因为这篇文章把弥漫在空气中的焦虑结晶化了

372
00:13:19,100 --> 00:13:20,559
恐惧一直都在

373
00:13:20,559 --> 00:13:21,941
它只是需要一个载体

374
00:13:22,299 --> 00:13:24,140
Citrini的报告就是那个载体

375
00:13:24,700 --> 00:13:26,139
如果你经常看我的频道

376
00:13:26,139 --> 00:13:28,240
你会记得我之前讲Matt Shumer那篇

377
00:13:28,240 --> 00:13:28,980
<Something Big Is

378
00:13:29,059 --> 00:13:29,379
Happening>

379
00:13:29,419 --> 00:13:30,799
的时候说过

380
00:13:30,799 --> 00:13:34,321
圈内人和圈外人之间的信息差已经大到了危险的地步

381
00:13:34,899 --> 00:13:37,860
Citrini的报告其实是这个信息差的另一个证据

382
00:13:38,340 --> 00:13:39,120
只不过这次

383
00:13:39,120 --> 00:13:42,379
信息差的震波从科技圈传到了金融圈

384
00:13:42,379 --> 00:13:44,040
然后传到了整个市场

385
00:13:44,120 --> 00:13:44,580
所以

386
00:13:44,580 --> 00:13:45,910
站在今天这个时间点

387
00:13:45,910 --> 00:13:47,360
我们应该怎么看

388
00:13:47,480 --> 00:13:48,659
我的观点很明确

389
00:13:48,659 --> 00:13:49,320
也很坦诚

390
00:13:49,820 --> 00:13:52,760
我不知道AI到底会以多快的速度改变经济

391
00:13:53,240 --> 00:13:53,841
没有人知道

392
00:13:54,399 --> 00:13:55,360
白宫不知道

393
00:13:55,360 --> 00:13:57,039
华尔街不知道，Citrini

394
00:13:57,039 --> 00:13:57,840
不知道，Sam

395
00:13:57,899 --> 00:13:58,900
Altman也不知道

396
00:13:59,419 --> 00:14:00,961
但有几件事我非常确定

397
00:14:01,460 --> 00:14:02,460
第一，AI

398
00:14:02,460 --> 00:14:04,799
对白领工作的冲击是真实的

399
00:14:04,799 --> 00:14:06,060
而且正在加速

400
00:14:06,179 --> 00:14:07,919
这不是2028年的科幻

401
00:14:07,919 --> 00:14:09,441
这是2026年的现实

402
00:14:09,919 --> 00:14:12,779
IBM当天跌13%，Anthropic

403
00:14:12,779 --> 00:14:15,159
真的展示了一个能威胁IBM核心

404
00:14:15,159 --> 00:14:17,019
业务的AI工具，Citrini

405
00:14:17,019 --> 00:14:20,321
的恐慌叙事又把整个市场的焦虑推到了顶点

406
00:14:21,139 --> 00:14:24,540
印度IT行业在整个2月份蒸发了500亿美元市值

407
00:14:25,059 --> 00:14:26,299
这些不是叙事

408
00:14:26,299 --> 00:14:27,941
这些是正在发生的事实

409
00:14:28,700 --> 00:14:29,399
第二

410
00:14:29,399 --> 00:14:32,400
叙事的力量在AI时代被空前放大了

411
00:14:32,440 --> 00:14:35,240
一篇Substack文章能引发千亿级别的

412
00:14:35,240 --> 00:14:36,200
市场波动

413
00:14:36,200 --> 00:14:39,701
这本身就说明市场的脆弱性已经到了一个危险的水平

414
00:14:40,440 --> 00:14:42,919
当"可信的故事"和"可交易的信号"之间的

415
00:14:42,919 --> 00:14:44,720
距离缩短到接近于零

416
00:14:44,720 --> 00:14:47,620
我们就进入了一个叙事即现实的时代

417
00:14:47,679 --> 00:14:48,259
第三

418
00:14:48,259 --> 00:14:51,759
也是我最想说的 - 在两种对立叙事的拉锯中

419
00:14:51,759 --> 00:14:52,860
你的位置在哪里

420
00:14:52,860 --> 00:14:54,201
取决于你自己的选择

421
00:14:54,779 --> 00:14:55,759
乐观派说，AI

422
00:14:55,759 --> 00:14:56,919
是生产力革命

423
00:14:56,919 --> 00:14:59,340
它会创造比摧毁更多的机会

424
00:14:59,340 --> 00:15:00,420
历史站在我们这边

425
00:15:01,080 --> 00:15:01,919
悲观派说

426
00:15:01,919 --> 00:15:03,299
这次不一样，AI

427
00:15:03,299 --> 00:15:04,299
是通用智能

428
00:15:04,299 --> 00:15:06,140
它会封堵所有的转型出口

429
00:15:06,519 --> 00:15:08,179
两种叙事都有逻辑

430
00:15:08,179 --> 00:15:09,019
都有支持者

431
00:15:09,019 --> 00:15:09,640
都有数据

432
00:15:10,100 --> 00:15:10,700
但最终

433
00:15:10,700 --> 00:15:12,379
哪种叙事成为现实

434
00:15:12,379 --> 00:15:14,419
不完全取决于技术本身

435
00:15:14,419 --> 00:15:16,899
而是取决于我们 - 作为个体、作为企业

436
00:15:16,940 --> 00:15:18,941
作为社会 - 如何选择和行动

437
00:15:19,539 --> 00:15:22,799
如果所有人都相信"天要塌了"然后什么都不做

438
00:15:22,799 --> 00:15:23,640
天就真的会塌

439
00:15:24,120 --> 00:15:27,399
如果足够多的人把AI当作工具而不是威胁

440
00:15:27,399 --> 00:15:29,440
主动重塑自己的能力结构

441
00:15:29,440 --> 00:15:30,740
结果就会完全不同

442
00:15:31,220 --> 00:15:32,620
这就是反身性的力量

443
00:15:33,200 --> 00:15:35,299
叙事不是被动描述未来的

444
00:15:35,299 --> 00:15:36,951
叙事是主动创造未来的

445
00:15:37,360 --> 00:15:39,879
所以我的建议没有变：不要恐慌

446
00:15:39,879 --> 00:15:40,860
但也不要无视

447
00:15:41,480 --> 00:15:42,818
主动靠近AI

448
00:15:42,820 --> 00:15:45,180
让它成为你的杠杆而不是你的替代者

449
00:15:45,759 --> 00:15:48,159
在一个旧秩序正在松动的世界里

450
00:15:48,159 --> 00:15:50,279
最危险的位置不是站错了队

451
00:15:50,279 --> 00:15:51,370
而是站在原地不动

452
00:15:51,820 --> 00:15:52,159
最后

453
00:15:52,159 --> 00:15:53,620
我想把这个问题抛给你们

454
00:15:54,240 --> 00:15:55,940
一篇Substack博客文章

455
00:15:55,940 --> 00:15:57,640
干掉了几千亿美元的市值

456
00:15:58,159 --> 00:16:00,220
这到底说明了AI有多可怕

457
00:16:00,220 --> 00:16:02,159
还是说明了人类有多脆弱

458
00:16:02,279 --> 00:16:02,840
又或者

459
00:16:02,840 --> 00:16:04,220
两者是同一件事

460
00:16:04,259 --> 00:16:04,960
我是王利杰

461
00:16:05,039 --> 00:16:05,980
我们下期见