【Meet Kevin】下一个万亿级机会：揭秘SpaceX的AI通信底牌

1
00:00:00,060 --> 00:00:00,500
你知道吗

2
00:00:00,500 --> 00:00:06,410
曾经被称为手机，能够实时生成你想要的任何视频内容

3
00:00:06,410 --> 00:00:09,529
我认为未来不会有操作系统

4
00:00:09,529 --> 00:00:12,769
未来也不会有应用程序

5
00:00:12,769 --> 00:00:13,890
将会存在操作系统

6
00:00:13,890 --> 00:00:14,390
说唱

7
00:00:14,390 --> 00:00:15,539
它将只是

8
00:00:15,539 --> 00:00:19,359
你拥有一台具备屏幕和音频功能的设备

9
00:00:19,359 --> 00:00:23,510
并且并且并且为了

10
00:00:23,510 --> 00:00:27,629
尽可能在设备上部署更多人工智能

11
00:00:27,629 --> 00:00:28,629
天啊

12
00:00:28,629 --> 00:00:29,929
这段视频将成为一棵树

13
00:00:29,929 --> 00:00:33,289
我们不仅要讨论SpaceX

14
00:00:33,289 --> 00:00:34,240
埃隆

15
00:00:34,240 --> 00:00:35,240
假新闻

16
00:00:35,240 --> 00:00:36,500
不是假新闻

17
00:00:36,500 --> 00:00:38,659
英伟达和美满电子科技

18
00:00:38,659 --> 00:00:41,109
全部集中在同一个视频里

19
00:00:41,109 --> 00:00:44,570
但我们想聊一些令人兴奋的内容

20
00:00:44,570 --> 00:00:47,310
现在这不是我的优惠码

21
00:00:47,310 --> 00:00:48,390
凯文官网

22
00:00:48,390 --> 00:00:53,320
这是埃隆与华尔街日报X之间的最新动态

23
00:00:53,320 --> 00:00:57,329
以及金融时报泄露的信息

24
00:00:57,329 --> 00:01:00,670
以及我们从埃隆在

25
00:01:00,670 --> 00:01:02,490
与乔·罗根的访谈中得知的内容

26
00:01:02,490 --> 00:01:03,990
让我们梳理所有信息

27
00:01:03,990 --> 00:01:05,469
因为这非常令人兴奋

28
00:01:05,469 --> 00:01:06,250
好的

29
00:01:06,250 --> 00:01:07,250
首先

30
00:01:07,250 --> 00:01:08,390
我想明确说明

31
00:01:08,390 --> 00:01:11,659
埃隆称这是彻头彻尾的谎言

32
00:01:11,659 --> 00:01:14,180
但他总是这样

33
00:01:14,180 --> 00:01:14,540
好的

34
00:01:14,540 --> 00:01:18,939
这是埃隆在路透社的经典操作

35
00:01:18,939 --> 00:01:19,420
例如

36
00:01:19,420 --> 00:01:20,739
出来宣称啊

37
00:01:20,739 --> 00:01:22,439
Model 2已停产

38
00:01:22,439 --> 00:01:25,239
据公司内部匿名人士透露

39
00:01:25,239 --> 00:01:26,829
因为他们不想被解雇

40
00:01:26,829 --> 00:01:29,829
然后埃隆出来反驳说假新闻

41
00:01:30,150 --> 00:01:31,670
然后人们会说

42
00:01:31,670 --> 00:01:32,790
他们引用匿名消息源

43
00:01:32,790 --> 00:01:33,750
或许这是假的

44
00:01:33,750 --> 00:01:35,150
两个月后突然

45
00:01:35,150 --> 00:01:35,310
哦

46
00:01:35,310 --> 00:01:36,349
Model 2已停产

47
00:01:37,790 --> 00:01:39,640
这就像一个故事

48
00:01:39,640 --> 00:01:41,659
与时间一样古老

49
00:01:41,659 --> 00:01:42,599
这就是循环

50
00:01:42,599 --> 00:01:43,519
在这个案例中

51
00:01:43,519 --> 00:01:49,640
埃隆说太空探索公司向投资者展示这一点的原因确实很明显

52
00:01:49,640 --> 00:01:52,640
埃隆·马斯克新AI设备的原型

53
00:01:52,640 --> 00:01:53,640
这其实是一部手机

54
00:01:53,640 --> 00:01:56,849
没有一个是完全虚假的

55
00:01:56,849 --> 00:01:59,129
为什么埃隆要说这完全不实

56
00:01:59,129 --> 00:02:02,489
因为这可能构成重大证券违规

57
00:02:02,489 --> 00:02:05,310
这个人可能会被起诉到玻利维亚

58
00:02:05,310 --> 00:02:07,079
我意思是它已经够被起诉了

59
00:02:07,079 --> 00:02:08,939
他会被起诉得更多

60
00:02:08,939 --> 00:02:13,759
如果你向内幕人士展示你或内部人士

61
00:02:13,759 --> 00:02:16,500
内幕人士实际在开发或进行的项目信息

62
00:02:16,500 --> 00:02:20,050
为了鼓励他们提供更好的融资

63
00:02:20,050 --> 00:02:23,370
或在绿鞋机制下购买更多IPO股份

64
00:02:23,370 --> 00:02:25,009
超额配售或类似操作

65
00:02:25,009 --> 00:02:25,729
这将是重大

66
00:02:25,729 --> 00:02:26,650
哎呀搞砸了

67
00:02:26,650 --> 00:02:26,969
所以

68
00:02:26,969 --> 00:02:29,330
当然埃隆必须出来澄清这是虚假的

69
00:02:29,330 --> 00:02:34,469
但我们需要知道他在乔·罗根播客中说了什么

70
00:02:34,469 --> 00:02:35,909
因为这很有趣

71
00:02:35,909 --> 00:02:36,250
好的

72
00:02:36,250 --> 00:02:36,930
让我们拆解一下

73
00:02:36,930 --> 00:02:39,210
首先是《华尔街日报》的报道

74
00:02:39,210 --> 00:02:44,750
《华尔街日报》称埃隆开发了一款手机设备原型

75
00:02:44,750 --> 00:02:51,039
设计旨在重塑人类与人工智能的交互方式

76
00:02:51,039 --> 00:02:53,919
现在真正引人注目的是

77
00:02:53,919 --> 00:02:55,459
我要你扪心自问

78
00:02:55,459 --> 00:03:02,389
还有哪家公司正在设计一款设计精巧的AI设备

79
00:03:02,389 --> 00:03:03,389
并且

80
00:03:03,389 --> 00:03:04,009
你知道的

81
00:03:04,009 --> 00:03:09,699
整合AI并革新我们与AI的沟通方式

82
00:03:09,699 --> 00:03:11,959
你可能会想到Meta和Meta眼镜

83
00:03:11,959 --> 00:03:14,229
这确实是个合理猜测

84
00:03:14,229 --> 00:03:16,710
但我说的是精巧设计

85
00:03:18,509 --> 00:03:19,539
乔尼我

86
00:03:19,539 --> 00:03:20,060
你知道的

87
00:03:20,060 --> 00:03:20,419
乔尼

88
00:03:20,419 --> 00:03:20,900
伊夫

89
00:03:20,900 --> 00:03:22,580
苹果设计师

90
00:03:22,580 --> 00:03:24,560
他现在在哪家公司工作

91
00:03:24,560 --> 00:03:29,879
AI 猜谁是埃隆·马斯克绝对讨厌的OpenAI

92
00:03:31,240 --> 00:03:31,439
所以

93
00:03:31,439 --> 00:03:32,500
如果OpenAI

94
00:03:32,500 --> 00:03:39,750
正在用挖来的苹果设计师打造一款精巧AI设备

95
00:03:39,750 --> 00:03:42,259
甚至可能取代你的手机

96
00:03:42,259 --> 00:03:45,659
埃隆也想这么做确实合乎逻辑

97
00:03:45,659 --> 00:03:50,280
尤其是我们还有星链在埃隆旗下

98
00:03:50,280 --> 00:03:52,219
除此之外

99
00:03:52,219 --> 00:03:54,020
就在六天前

100
00:03:54,020 --> 00:03:56,620
看看这篇《金融时报》报道

101
00:03:56,620 --> 00:04:01,580
马斯克的SpaceX计划将星链移动服务推向美国

102
00:04:01,580 --> 00:04:02,800
消费市场一切正常

103
00:04:02,800 --> 00:04:05,189
这并不意味着他一定会推出手机

104
00:04:05,189 --> 00:04:05,550
但

105
00:04:05,550 --> 00:04:06,349
这确实表明

106
00:04:06,349 --> 00:04:09,069
马斯克的SpaceX已向投资者表示

107
00:04:09,069 --> 00:04:14,199
计划推出面向消费者的全新星链移动服务

108
00:04:14,199 --> 00:04:19,259
这可能颠覆该国数十亿美元的手机网络市场

109
00:04:19,259 --> 00:04:20,480
我觉得大家都会说

110
00:04:20,480 --> 00:04:21,278
听着啊

111
00:04:21,278 --> 00:04:24,298
如果我能有更好的手机信号

112
00:04:24,298 --> 00:04:25,298
我立刻报名

113
00:04:25,298 --> 00:04:25,699
你知道的

114
00:04:25,699 --> 00:04:27,098
有些人会说我不喜欢马斯克

115
00:04:27,098 --> 00:04:28,499
有些人说我对T-Mobile还行

116
00:04:28,499 --> 00:04:29,899
不过随便啦

117
00:04:29,899 --> 00:04:32,759
但重点是这已讨论很久了

118
00:04:32,759 --> 00:04:34,079
有些人说

119
00:04:34,079 --> 00:04:35,040
这太荒谬了

120
00:04:35,040 --> 00:04:37,019
需要这么多频段

121
00:04:37,019 --> 00:04:39,319
不可能完全取代其他公司

122
00:04:39,319 --> 00:04:40,839
你连竞争都难

123
00:04:40,839 --> 00:04:44,620
因为他们拥有大量所需频谱资源

124
00:04:44,620 --> 00:04:47,610
这也是为什么其他人说像马斯克本人

125
00:04:47,610 --> 00:04:50,410
当他在《All In》节目中接受采访时表示

126
00:04:51,610 --> 00:04:57,860
或许我们实际要做的是太空到手机

127
00:04:57,860 --> 00:05:01,699
通过50或65毫赫兹连接

128
00:05:01,699 --> 00:05:04,500
可能需要更新手机硬件

129
00:05:04,500 --> 00:05:06,860
我不确定这如何实现室内覆盖

130
00:05:06,860 --> 00:05:08,120
马斯克说这可行

131
00:05:08,120 --> 00:05:10,298
但不如你自己听听马斯克怎么说

132
00:05:10,298 --> 00:05:11,358
谈谈这个

133
00:05:11,358 --> 00:05:13,259
然后我会介绍那个设备

134
00:05:13,259 --> 00:05:15,338
在我看来这很科幻

135
00:05:15,338 --> 00:05:18,389
我会展示乔·罗根播客中的片段

136
00:05:18,389 --> 00:05:19,389
现在让我们深入探讨

137
00:05:19,389 --> 00:05:19,750
呃

138
00:05:19,750 --> 00:05:20,310
请记住

139
00:05:20,310 --> 00:05:20,509
是的

140
00:05:20,509 --> 00:05:23,430
我们确实有该优惠券的单日延期

141
00:05:23,430 --> 00:05:24,029
在meet kevin网站

142
00:05:24,029 --> 00:05:24,879
com和rivest

143
00:05:24,879 --> 00:05:26,980
我们将大幅提高价格

144
00:05:26,980 --> 00:05:30,970
因为我们6月和7月发布了大量内容

145
00:05:30,970 --> 00:05:32,370
现在必须大幅涨价

146
00:05:32,370 --> 00:05:34,569
如果你在优惠码到期前入手

147
00:05:34,569 --> 00:05:36,410
将终身锁定当前价格

148
00:05:36,410 --> 00:05:36,850
好的这里

149
00:05:36,850 --> 00:05:37,389
我们开始

150
00:05:39,240 --> 00:05:40,360
SpaceX宣布

151
00:05:42,920 --> 00:05:48,379
直接从卫星向手机提供高速且稳定的连接

152
00:05:48,839 --> 00:05:53,639
不过手机需要进行硬件改动

153
00:05:53,639 --> 00:05:57,500
因为这些频段已在当前手机中得到支持

154
00:05:58,180 --> 00:05:59,720
芯片组方面

155
00:05:59,720 --> 00:06:00,829
我来澄清一下

156
00:06:00,829 --> 00:06:04,730
其中一些频段已在许多设备中支持

157
00:06:04,730 --> 00:06:08,180
除了埃隆正在构建的这种类型

158
00:06:08,225 --> 00:06:09,920
我该怎么解释这个

159
00:06:09,920 --> 00:06:15,449
这款号称能生成视频的设备

160
00:06:15,449 --> 00:06:19,579
会根据使用需求动态调整

161
00:06:20,699 --> 00:06:23,639
埃隆认为需要硬件升级

162
00:06:23,639 --> 00:06:26,959
这可能是他们自建硬件的原因

163
00:06:26,959 --> 00:06:29,459
他稍后在节目中会详细说明

164
00:06:29,459 --> 00:06:33,629
在乔·罗根节目中加入这些频段

165
00:06:33,629 --> 00:06:35,930
这大概就是你的时间线了

166
00:06:35,930 --> 00:06:40,300
能够使用已收购频谱的手机

167
00:06:40,300 --> 00:06:43,079
可能两年后开始发货

168
00:06:43,079 --> 00:06:45,379
嗯而且

169
00:06:45,779 --> 00:06:47,800
我们也需要部署相关卫星

170
00:06:47,800 --> 00:06:50,480
这些卫星将使用指定频段通信

171
00:06:50,480 --> 00:06:51,480
同时进行

172
00:06:51,480 --> 00:06:54,920
我们正在建造卫星并与手机制造商合作

173
00:06:54,920 --> 00:06:57,319
为手机添加这些频段

174
00:06:57,319 --> 00:06:59,579
基本上他在说

175
00:06:59,579 --> 00:07:00,449
这件事正在推进

176
00:07:00,449 --> 00:07:02,410
我们已经在部署卫星网络

177
00:07:02,410 --> 00:07:04,550
并提前通知

178
00:07:04,550 --> 00:07:05,050
你知道的

179
00:07:05,050 --> 00:07:06,709
三星或苹果

180
00:07:06,709 --> 00:07:12,500
或许需要考虑添加专用芯片以提升带宽

181
00:07:12,500 --> 00:07:14,009
以满足我们的需求

182
00:07:14,009 --> 00:07:17,408
这是你们可以同步推进的工作

183
00:07:17,408 --> 00:07:19,389
因为用户会需要这个功能

184
00:07:19,389 --> 00:07:20,449
如果你们不做

185
00:07:20,449 --> 00:07:23,360
我们将通过自有设备实现

186
00:07:23,360 --> 00:07:24,699
稍后详细说明

187
00:07:24,699 --> 00:07:28,949
卫星与手机将完美协同

188
00:07:28,949 --> 00:07:30,790
实现高带宽连接

189
00:07:30,790 --> 00:07:31,670
但最终效果

190
00:07:31,670 --> 00:07:35,370
高带宽意味着可流畅观看视频

191
00:07:35,370 --> 00:07:36,189
嗯

192
00:07:36,189 --> 00:07:37,329
随时随地在手机上

193
00:07:37,329 --> 00:07:37,990
好的

194
00:07:37,990 --> 00:07:38,569
现在

195
00:07:38,569 --> 00:07:39,410
让我们回到乔·罗根

196
00:07:39,410 --> 00:07:39,870
罗根

197
00:07:39,870 --> 00:07:42,420
这不仅仅是随时随地看视频

198
00:07:42,420 --> 00:07:43,519
随时随地在手机上

199
00:07:43,519 --> 00:07:50,019
核心目标是打造API生态系统

200
00:07:50,019 --> 00:07:51,360
无需第三方应用

201
00:07:51,360 --> 00:07:52,040
我老实说

202
00:07:52,040 --> 00:07:54,319
我觉得这有点像外星科技

203
00:07:54,319 --> 00:07:57,040
我认为最简单的方式是这样思考

204
00:07:57,040 --> 00:07:59,480
我拿这张报纸过来

205
00:07:59,480 --> 00:08:00,110
好的

206
00:08:00,110 --> 00:08:05,350
这张报纸上确实有委内瑞拉的救援行动

207
00:08:05,350 --> 00:08:07,790
拿政治术语来说特朗普削减开支

208
00:08:07,790 --> 00:08:08,750
大手笔交易

209
00:08:08,750 --> 00:08:09,189
好的

210
00:08:09,189 --> 00:08:09,490
随便吧

211
00:08:09,490 --> 00:08:10,730
这是纽约时报没错

212
00:08:10,730 --> 00:08:13,490
想象一下如果我有这样的设备

213
00:08:13,490 --> 00:08:16,038
然后我说好吧

214
00:08:16,038 --> 00:08:16,899
显示我的邮件

215
00:08:19,019 --> 00:08:19,298
你知道的

216
00:08:19,298 --> 00:08:20,379
我所有邮件都显示在这里

217
00:08:20,379 --> 00:08:26,750
然后我清空并展示凯文的YouTube频道

218
00:08:26,750 --> 00:08:27,850
有点酷

219
00:08:27,850 --> 00:08:28,230
嘿

220
00:08:28,230 --> 00:08:34,048
这是最新推文，整个界面无需应用直接变化

221
00:08:34,048 --> 00:08:37,839
我觉得这和应用功能类似

222
00:08:37,839 --> 00:08:40,958
但马斯克应该能解释得更清楚

223
00:08:40,958 --> 00:08:42,759
我认为他不满的是整个

224
00:08:42,759 --> 00:08:45,580
像主页上寻找不同应用

225
00:08:45,580 --> 00:08:48,139
它们之间整合得并不好

226
00:08:48,139 --> 00:08:51,980
我们应该能生成完全自定义的桌面

227
00:08:51,980 --> 00:08:54,360
根据即时需求动态调整

228
00:08:54,360 --> 00:08:55,559
这愿景挺酷的

229
00:08:55,559 --> 00:08:56,580
但可能更简单

230
00:08:56,580 --> 00:08:58,620
如果仔细听他讲解的话

231
00:08:58,620 --> 00:09:00,639
我可以告诉你未来的发展方向

232
00:09:00,639 --> 00:09:01,200
呃

233
00:09:01,200 --> 00:09:02,159
那就是

234
00:09:02,159 --> 00:09:03,820
我们不会有传统意义上的手机

235
00:09:03,820 --> 00:09:06,100
或者说

236
00:09:06,100 --> 00:09:10,139
我们所说的手机将真正成为

237
00:09:10,659 --> 00:09:13,240
AI的边缘节点

238
00:09:13,240 --> 00:09:14,940
AI推理处理

239
00:09:14,940 --> 00:09:17,590
视频推理配合

240
00:09:17,590 --> 00:09:18,029
呃

241
00:09:18,029 --> 00:09:20,230
配备一些无线电连接

242
00:09:20,230 --> 00:09:20,870
嗯连接到

243
00:09:20,870 --> 00:09:21,929
但本质上

244
00:09:21,929 --> 00:09:22,309
你会拥有

245
00:09:23,590 --> 00:09:24,470
嗯

246
00:09:24,470 --> 00:09:27,110
服务器端的AI

247
00:09:27,110 --> 00:09:30,830
与设备端AI通信

248
00:09:30,830 --> 00:09:31,769
嗯

249
00:09:31,769 --> 00:09:32,570
你知道的

250
00:09:32,570 --> 00:09:38,470
以前称为手机，实时生成任何所需视频

251
00:09:38,470 --> 00:09:41,590
我认为他们不会运行操作系统

252
00:09:41,590 --> 00:09:44,090
也不会有应用程序

253
00:09:44,090 --> 00:09:44,809
在未来

254
00:09:44,809 --> 00:09:46,470
将会存在操作系统或封装层

255
00:09:46,470 --> 00:09:47,610
它将只是

256
00:09:47,610 --> 00:09:51,429
你有一个专为屏幕和音频设计的设备

257
00:09:51,429 --> 00:09:52,779
并且对于

258
00:09:52,779 --> 00:09:54,620
并且并且并且

259
00:09:54,620 --> 00:09:59,639
尽量在设备上部署尽可能多的AI

260
00:09:59,639 --> 00:10:03,399
以减少边缘设备与服务器之间所需的带宽

261
00:10:03,399 --> 00:10:03,999
节点设备

262
00:10:03,999 --> 00:10:06,990
或者仅称为手机和服务器

263
00:10:06,990 --> 00:10:08,269
如果没有应用程序

264
00:10:09,309 --> 00:10:11,409
人们会用什么

265
00:10:11,409 --> 00:10:13,269
X还会存在吗

266
00:10:13,269 --> 00:10:14,659
对吧

267
00:10:14,659 --> 00:10:15,639
这是个合理的问题

268
00:10:15,639 --> 00:10:16,039
就像

269
00:10:16,039 --> 00:10:18,549
这会不会自相矛盾呢

270
00:10:18,549 --> 00:10:22,320
到目前为止我们描述的内容其实正是埃隆所阐述的

271
00:10:22,320 --> 00:10:25,299
而在我们的描述中甚至提到

272
00:10:25,299 --> 00:10:27,460
但本质上不就是普通应用吗

273
00:10:27,460 --> 00:10:27,840
你知道的

274
00:10:27,840 --> 00:10:28,990
我可以切换上下文

275
00:10:28,990 --> 00:10:30,730
埃隆的愿景是不

276
00:10:30,730 --> 00:10:34,419
所有内容都应实时生成

277
00:10:34,419 --> 00:10:37,159
是否会成为邮件平台

278
00:10:37,159 --> 00:10:39,000
还是全部通过

279
00:10:39,000 --> 00:10:39,809
AI

280
00:10:41,830 --> 00:10:43,870
AI将带来这种优势

281
00:10:43,870 --> 00:10:49,149
而非像乔·罗根的独立播客

282
00:10:49,149 --> 00:10:52,929
我的猜测是试图从

283
00:10:52,929 --> 00:10:53,769
这次访谈

284
00:10:53,769 --> 00:10:56,389
展示如果我能证明

285
00:10:56,389 --> 00:10:56,710
我

286
00:10:56,710 --> 00:10:58,070
你知道我需要写报告

287
00:10:58,070 --> 00:10:58,710
我需要我的

288
00:10:58,710 --> 00:11:00,350
我需要部分邮件

289
00:11:00,350 --> 00:11:03,500
并需要部分这个或那个

290
00:11:03,500 --> 00:11:06,250
而不是像

291
00:11:06,250 --> 00:11:06,850
嗯

292
00:11:06,850 --> 00:11:08,110
我可以即时

293
00:11:08,110 --> 00:11:10,149
让它完全符合我的需求

294
00:11:10,149 --> 00:11:11,009
我不

295
00:11:11,009 --> 00:11:11,549
我不知道

296
00:11:11,549 --> 00:11:13,960
这似乎不太一致

297
00:11:13,960 --> 00:11:14,340
现在

298
00:11:14,340 --> 00:11:14,799
就像

299
00:11:14,799 --> 00:11:18,120
能否每次都生成我习惯的内容

300
00:11:18,120 --> 00:11:20,320
或者我得学会现在新按钮的位置

301
00:11:20,320 --> 00:11:22,240
显然会有一些学习曲线

302
00:11:22,240 --> 00:11:23,960
听起来非常陌生

303
00:11:23,960 --> 00:11:26,480
但听起来非常有趣的应用

304
00:11:26,639 --> 00:11:27,879
无论你能想到什么

305
00:11:27,879 --> 00:11:29,860
或者AI能预见的任何事物

306
00:11:29,860 --> 00:11:30,539
你可能会想要

307
00:11:30,539 --> 00:11:32,328
它会直接展示给你

308
00:11:32,328 --> 00:11:32,828
没错

309
00:11:32,828 --> 00:11:34,609
这就是我的预测

310
00:11:34,609 --> 00:11:36,288
事物最终会走向哪里

311
00:11:36,570 --> 00:11:38,529
我们在这里的时间框架是多久

312
00:11:40,409 --> 00:11:42,279
我不知道大概就是这样

313
00:11:42,720 --> 00:11:43,279
可能

314
00:11:45,360 --> 00:11:48,830
五年或六年左右，挺有意思的

315
00:11:48,830 --> 00:11:51,350
时间框架已经有所调整

316
00:11:51,350 --> 00:11:52,889
他基本上在说嘿

317
00:11:52,889 --> 00:11:53,429
我是说

318
00:11:53,429 --> 00:11:55,330
这不一定是时间框架的改变

319
00:11:55,330 --> 00:11:58,409
这期播客是在2025年10月录制的

320
00:11:58,409 --> 00:12:02,409
而与马斯克的全集播客稍近一些

321
00:12:02,409 --> 00:12:05,269
所以你可能听到的是嘿

322
00:12:05,269 --> 00:12:06,490
手机将像往常一样开始

323
00:12:06,490 --> 00:12:08,559
手机将在两年后开始发货

324
00:12:08,559 --> 00:12:12,299
具备与星链卫星通信的能力

325
00:12:12,299 --> 00:12:14,789
几年后

326
00:12:14,789 --> 00:12:19,730
那时我们可以深入探讨解耦AI领域

327
00:12:19,730 --> 00:12:23,929
基本上你的AI在生成知识

328
00:12:23,929 --> 00:12:27,090
或者像你面前的界面

329
00:12:27,090 --> 00:12:35,490
然后这个高度生成的AI界面与远程数据库通信

330
00:12:35,490 --> 00:12:38,230
这将增加带宽需求

331
00:12:38,230 --> 00:12:40,190
这将增加服务器需求

332
00:12:40,190 --> 00:12:43,070
这有趣地引出了美满科技

333
00:12:44,269 --> 00:12:45,009
看看这个

334
00:12:45,009 --> 00:12:48,289
这是美满科技今天刚发布的，时机太巧了

335
00:12:48,289 --> 00:12:48,830
但

336
00:12:48,830 --> 00:12:49,710
听这个

337
00:12:49,710 --> 00:12:51,889
谢谢观看

338
00:12:51,889 --> 00:12:53,409
你可能听到了我刚才说的内容

339
00:12:53,409 --> 00:12:55,230
今天我们讨论的是连接性

340
00:12:55,230 --> 00:12:57,009
下一个万亿美金公司

341
00:12:57,009 --> 00:12:58,009
女士们先生们

342
00:12:58,009 --> 00:12:58,629
哇

343
00:13:00,710 --> 00:13:01,730
这会很令人兴奋

344
00:13:01,730 --> 00:13:02,570
我们一起做吧

345
00:13:02,570 --> 00:13:02,870
是的

346
00:13:02,870 --> 00:13:04,149
我们一起做吧

347
00:13:04,149 --> 00:13:07,269
但一切始于今天正在发生的事

348
00:13:07,269 --> 00:13:09,470
在AI基础设施更广泛的领域

349
00:13:09,470 --> 00:13:11,190
你如何看待这一点

350
00:13:11,190 --> 00:13:12,710
仅从整体视角来看

351
00:13:12,710 --> 00:13:16,490
在这个非凡时刻客户需求飙升

352
00:13:16,490 --> 00:13:19,190
我感觉就像詹森会怎么做

353
00:13:19,190 --> 00:13:20,450
詹森直接出现

354
00:13:20,450 --> 00:13:21,070
然后他说道

355
00:13:21,070 --> 00:13:21,190
是的

356
00:13:21,190 --> 00:13:22,750
我来帮忙泵货

357
00:13:22,750 --> 00:13:26,200
公平地说这一直是课程成员最爱

358
00:13:26,200 --> 00:13:29,320
所以我们决定加倍参与

359
00:13:29,320 --> 00:13:30,240
我们要接受这些赞誉

360
00:13:31,600 --> 00:13:34,750
嗯，我觉得课程成员在这次表现很好

361
00:13:34,750 --> 00:13:37,470
但我觉得詹森就是直接出现

362
00:13:37,470 --> 00:13:38,710
他问我们该谈什么

363
00:13:38,710 --> 00:13:40,129
对方不得不重新强调

364
00:13:40,129 --> 00:13:42,200
他们要讨论的是连接性

365
00:13:42,200 --> 00:13:43,940
兄弟一定要谈连接性

366
00:13:43,940 --> 00:13:44,620
你会看到的

367
00:13:44,620 --> 00:13:46,009
说到他

368
00:13:46,009 --> 00:13:47,850
在演讲开头三十秒融入

369
00:13:47,850 --> 00:13:49,740
在他开场调侃时

370
00:13:49,740 --> 00:13:51,860
你能看到连接性贯穿其中

371
00:13:51,860 --> 00:13:53,740
以及所需的互联互通

372
00:13:53,740 --> 00:13:54,100
是的

373
00:13:54,100 --> 00:13:54,720
这真的很棒

374
00:13:54,720 --> 00:13:55,220
你知道昨天

375
00:13:56,220 --> 00:14:00,100
我说实用AI已经到来

376
00:14:00,100 --> 00:14:03,200
这就是需求激增的原因

377
00:14:03,200 --> 00:14:05,460
这也是我的需求激增的原因

378
00:14:05,460 --> 00:14:10,120
这种新计算模式称为智能体

379
00:14:10,120 --> 00:14:14,190
这些智能体运行在特定计算平台

380
00:14:14,190 --> 00:14:21,159
这种计算模式是解耦和分布式的

381
00:14:21,159 --> 00:14:24,109
将计算问题拆解为多个部分

382
00:14:24,109 --> 00:14:27,109
在数据中心全局分布执行

383
00:14:27,109 --> 00:14:30,429
所需的就是强大的连接性

384
00:14:30,429 --> 00:14:33,669
这就是马特表现优异的原因

385
00:14:33,669 --> 00:14:35,029
这就是为什么

386
00:14:35,029 --> 00:14:37,049
这已被整合

387
00:14:37,049 --> 00:14:43,789
获取AI能力正是漫威现在做的事

388
00:14:43,789 --> 00:14:45,529
接下来更有趣了

389
00:14:46,529 --> 00:14:52,159
马斯克说他不开发这个设备

390
00:14:52,159 --> 00:14:55,320
OpenAI在开发设备

391
00:14:55,320 --> 00:15:01,299
但马斯克构想设备形态却讨厌OpenAI

392
00:15:01,299 --> 00:15:03,960
马斯克谈论边缘计算

393
00:15:03,960 --> 00:15:06,460
英伟达的詹森也在谈

394
00:15:06,460 --> 00:15:11,210
科通芯城和美满电子的原型如《华尔街日报》所述

395
00:15:11,210 --> 00:15:14,610
比iPhone更纤薄

396
00:15:14,610 --> 00:15:21,559
运行专有操作系统并整合SpaceX的AI技术

397
00:15:21,919 --> 00:15:26,279
值得注意的是体积至关重要

398
00:15:26,279 --> 00:15:27,039
事实上

399
00:15:27,039 --> 00:15:28,269
最新进展

400
00:15:28,269 --> 00:15:34,029
一种类似苹果开箱视频，由精英们展示

401
00:15:34,029 --> 00:15:36,289
对于这些iPhone翻盖手机

402
00:15:36,289 --> 00:15:38,720
手机都讲究尺寸大小

403
00:15:38,720 --> 00:15:39,840
这有多大

404
00:15:39,840 --> 00:15:40,919
有多厚

405
00:15:40,919 --> 00:15:42,700
当折叠起来时

406
00:15:42,700 --> 00:15:43,840
那部分厚度

407
00:15:43,840 --> 00:15:45,159
有多少折痕

408
00:15:45,159 --> 00:15:46,279
你去了哪里

409
00:15:46,279 --> 00:15:48,179
厚度很重要

410
00:15:48,179 --> 00:15:53,389
所以yan也在正确方向上努力

411
00:15:53,389 --> 00:15:58,039
SpaceX曾对投资者表示项目处于早期阶段

412
00:15:58,039 --> 00:16:00,519
SpaceX和高通的代表

413
00:16:00,519 --> 00:16:02,559
由于高通参与其中

414
00:16:02,559 --> 00:16:02,919
呃

415
00:16:02,919 --> 00:16:05,159
比如内置芯片组

416
00:16:05,159 --> 00:16:08,779
包含在SpaceX设备或手机内

417
00:16:08,779 --> 00:16:09,980
无论叫什么名字

418
00:16:09,980 --> 00:16:11,860
但他们并未回应

419
00:16:11,860 --> 00:16:13,220
但呃

420
00:16:13,220 --> 00:16:17,000
《华尔街日报》称该项目体现马斯克的宏大野心

421
00:16:17,000 --> 00:16:19,919
随着他转向全球卫星互联

422
00:16:19,919 --> 00:16:22,730
我意思是这些都相互契合

423
00:16:22,730 --> 00:16:26,669
《华尔街日报》曾报道马斯克曾考虑造手机

424
00:16:26,669 --> 00:16:32,460
因不满苹果对X应用的分发控制

425
00:16:32,860 --> 00:16:35,860
造手机的想法让我想死

426
00:16:35,860 --> 00:16:37,460
马斯克在十月表示

427
00:16:37,460 --> 00:16:39,480
但如果我们必须造手机

428
00:16:39,480 --> 00:16:41,679
我们将在二月推出

429
00:16:41,679 --> 00:16:44,240
马斯克否认公司开发手机

430
00:16:44,240 --> 00:16:47,240
驳斥路透社关于SpaceX造手机的报道

431
00:16:47,240 --> 00:16:51,269
并将直接连接SpaceX卫星网络

432
00:16:51,269 --> 00:16:53,169
我们不开发手机

433
00:16:53,169 --> 00:16:53,889
他说道

434
00:16:53,889 --> 00:16:58,580
但马斯克以留有余地著称

435
00:16:58,580 --> 00:16:59,100
是啊

436
00:16:59,100 --> 00:17:00,399
我们不开发手机

437
00:17:00,399 --> 00:17:02,700
我们在开发AI设备

438
00:17:02,700 --> 00:17:04,319
这确实是真的

439
00:17:04,880 --> 00:17:11,119
所以基于马斯克否认过往的事实

440
00:17:11,119 --> 00:17:13,318
而最终结果却成真

441
00:17:13,318 --> 00:17:15,759
大致类似

442
00:17:15,759 --> 00:17:17,019
当你知道时

443
00:17:17,019 --> 00:17:21,140
有传言称SpaceX正在筹集150亿美元

444
00:17:21,140 --> 00:17:23,920
或是X公司正在筹集150亿美元

445
00:17:23,920 --> 00:17:26,029
然后他写完全错误

446
00:17:26,029 --> 00:17:30,839
一周后发现实际筹集了200亿美元

447
00:17:30,839 --> 00:17:31,359
好吧

448
00:17:31,359 --> 00:17:33,480
所以确实在融资是真的

449
00:17:33,480 --> 00:17:36,180
但150亿的说法是假的

450
00:17:36,180 --> 00:17:39,880
你讨论的内容远不止这些有趣的部分

451
00:17:39,880 --> 00:17:42,420
不过把这些全部结合起来

452
00:17:42,420 --> 00:17:44,279
确实变得有点有趣了

453
00:17:44,279 --> 00:17:51,150
因为这里甚至有一个詹森谈论基站架构的视频

454
00:17:51,150 --> 00:17:52,359
与美满电子科技合作

455
00:17:52,359 --> 00:17:54,279
这是一段三个月前的视频

456
00:17:54,279 --> 00:17:57,749
但仔细听这段关于云端数据中心

457
00:17:57,749 --> 00:18:03,109
以及全球电信AI基站网络的内容

458
00:18:03,109 --> 00:18:05,430
未来的电信网络

459
00:18:05,430 --> 00:18:08,950
今天的电信网络本质上是哑管道

460
00:18:08,950 --> 00:18:11,449
未来的无线哑管道

461
00:18:11,449 --> 00:18:13,548
就像我们做云计算那样

462
00:18:13,548 --> 00:18:18,099
我们将把AI基站变成

463
00:18:18,460 --> 00:18:20,099
成为AI基础设施的一部分

464
00:18:20,099 --> 00:18:24,640
因此我们能在基站内部运行AI模型

465
00:18:24,640 --> 00:18:25,638
在未来

466
00:18:25,638 --> 00:18:27,459
电信与基站

467
00:18:27,459 --> 00:18:34,039
是美满电子及马特多年来构建的核心业务

468
00:18:34,039 --> 00:18:36,949
我们现在要将MV链路连接到这里

469
00:18:36,949 --> 00:18:41,289
从而将NVIDIA生态延伸至基站

470
00:18:41,289 --> 00:18:42,210
好的

471
00:18:42,210 --> 00:18:45,009
让我试着解释一下，这里有很多内容

472
00:18:45,009 --> 00:18:45,529
好的

473
00:18:45,529 --> 00:18:46,849
首先

474
00:18:47,089 --> 00:18:48,509
当你想到基站

475
00:18:48,509 --> 00:18:49,609
只需想象

476
00:18:49,609 --> 00:18:50,630
你知道的

477
00:18:50,630 --> 00:18:51,170
一座塔

478
00:18:51,170 --> 00:18:52,730
不一定非要是塔

479
00:18:52,730 --> 00:18:56,609
可能只是屋顶上的小天线

480
00:18:56,609 --> 00:18:58,430
这就是大城市获得5G的方式

481
00:18:58,430 --> 00:19:01,309
他们在屋顶或商铺安装大量天线

482
00:19:01,309 --> 00:19:01,829
或者其他地方

483
00:19:01,829 --> 00:19:04,079
从而在城市内实现5G服务

484
00:19:04,079 --> 00:19:06,079
越靠近那个天线

485
00:19:06,079 --> 00:19:06,980
信号越好

486
00:19:06,980 --> 00:19:08,339
这很正常

487
00:19:08,339 --> 00:19:10,099
但为了讨论方便

488
00:19:10,099 --> 00:19:12,039
假设这是一个巨型天线

489
00:19:12,039 --> 00:19:13,259
巨型塔

490
00:19:13,259 --> 00:19:13,779
好的

491
00:19:13,779 --> 00:19:16,859
这就是基站

492
00:19:17,309 --> 00:19:25,839
英伟达认为存在一个AI或无线接入网络

493
00:19:25,839 --> 00:19:28,369
可以被创建

494
00:19:28,369 --> 00:19:33,750
基站因为需要位于设备与数据中心之间

495
00:19:33,750 --> 00:19:35,789
无论是手机还是其他设备

496
00:19:35,789 --> 00:19:39,159
和远方的数据中心

497
00:19:39,159 --> 00:19:42,519
基站需要大量计算能力

498
00:19:42,519 --> 00:19:48,309
不仅要在5G和未来6G间传输数据

499
00:19:48,309 --> 00:19:53,210
甚至可能进行预计算或完全计算

500
00:19:53,210 --> 00:19:58,880
基站的位置本身可以加速与大型数据中心的连接

501
00:19:58,880 --> 00:20:05,779
这就是Marvell拥有的名为Octagon基带处理器的产品

502
00:20:05,779 --> 00:20:08,319
这些处理器支持5G和6G网络

503
00:20:08,319 --> 00:20:12,940
英伟达也有名为AI Aerial的产品

504
00:20:12,940 --> 00:20:14,119
我马上调出来

505
00:20:14,119 --> 00:20:14,740
很快

506
00:20:14,740 --> 00:20:15,279
呃

507
00:20:15,279 --> 00:20:16,200
不过

508
00:20:16,200 --> 00:20:21,868
这就是我所说的Marvell的长远布局

509
00:20:21,868 --> 00:20:28,240
长远来看最终会实现分布式AI

510
00:20:28,240 --> 00:20:29,980
在未来五到十年

511
00:20:29,980 --> 00:20:35,369
分布式AI需要更冷静的设备

512
00:20:35,369 --> 00:20:37,029
更快更好

513
00:20:37,029 --> 00:20:38,349
更低延迟

514
00:20:38,349 --> 00:20:42,730
这样就能实现埃隆想要的目标

515
00:20:42,730 --> 00:20:45,829
Nvlink将所有组件连接起来

516
00:20:45,829 --> 00:20:47,630
这也是 Jensen 所强调的

517
00:20:47,630 --> 00:20:50,839
如果把这些整合在一起

518
00:20:50,839 --> 00:20:54,759
希望能鼓励大家订阅和观看这类视频

519
00:20:54,759 --> 00:20:56,559
如果把这些整合在一起

520
00:20:56,559 --> 00:20:59,640
你会发现

521
00:21:00,039 --> 00:21:04,480
英伟达正在打造原生AI无线网络

522
00:21:05,839 --> 00:21:11,690
称为AI无线接入网络及5G/6G解决方案

523
00:21:12,890 --> 00:21:16,490
Marvell正是其中一环

524
00:21:16,490 --> 00:21:23,880
这也是他们通过连接分布式AI站点合作的原因

525
00:21:23,880 --> 00:21:25,700
无论是你的设备

526
00:21:25,700 --> 00:21:28,660
基站还是数据中心

527
00:21:28,660 --> 00:21:35,160
埃隆·马斯克有星链进行实际蜂窝通信

528
00:21:35,160 --> 00:21:36,920
所以啊

529
00:21:36,920 --> 00:21:41,259
手机可以连接这些基站或其他网络

530
00:21:41,259 --> 00:21:45,039
当超出基站覆盖范围时

531
00:21:45,200 --> 00:21:51,059
埃隆·马斯克可能在设计类似

532
00:21:51,059 --> 00:21:57,339
OpenAI正在尝试打造的AI替代设备或全功能设备

533
00:21:57,339 --> 00:21:58,619
谁知道呢

534
00:21:58,619 --> 00:22:00,619
现在已有大量设备

535
00:22:00,619 --> 00:22:03,420
随时随地记录你的活动并生成每日摘要

536
00:22:03,420 --> 00:22:05,920
我个人认为有些设备有点奇怪

537
00:22:05,920 --> 00:22:08,019
主要是因为太令人震惊了

538
00:22:08,019 --> 00:22:11,409
周围人会记录你说的每一句话并上传云端

539
00:22:11,409 --> 00:22:12,969
听起来法律风险很大

540
00:22:12,969 --> 00:22:14,909
对双方都有风险

541
00:22:14,909 --> 00:22:18,108
也显得有些侵犯隐私

542
00:22:18,108 --> 00:22:19,750
不过随你吧

543
00:22:19,750 --> 00:22:21,710
但把这些整合起来

544
00:22:21,710 --> 00:22:23,470
你可以看到

545
00:22:23,470 --> 00:22:26,259
当埃隆说这是完全错误的

546
00:22:26,259 --> 00:22:27,859
可能不是

547
00:22:27,859 --> 00:22:30,099
很可能完全正确

548
00:22:30,099 --> 00:22:32,839
说实话我挺支持的

549
00:22:32,839 --> 00:22:34,400
听起来还挺有意思的

550
00:22:34,400 --> 00:22:38,099
听起来很有趣，我很期待看到接下来的发展

551
00:22:38,099 --> 00:22:40,210
我非常期待继续投资漫威

552
00:22:40,210 --> 00:22:43,869
我认为英伟达长期来看仍然是令人兴奋的公司

553
00:22:43,869 --> 00:22:46,210
我现在更看好美满电子科技而非英伟达

554
00:22:46,210 --> 00:22:47,190
这只是个决定

555
00:22:47,190 --> 00:22:51,028
嗯，我当初卖掉了很多英伟达股票，平均成本价

556
00:22:51,028 --> 00:22:54,409
或平均成交价约为195美元

557
00:22:54,409 --> 00:22:56,328
然后向所有课程成员发送了提醒

558
00:22:56,328 --> 00:22:57,400
当我这样做时

559
00:22:57,400 --> 00:22:59,740
随后投资了漫威

560
00:22:59,740 --> 00:23:00,559
嗯

561
00:23:00,559 --> 00:23:00,900
你知道的

562
00:23:00,900 --> 00:23:03,190
从那时起漫威涨了不少

563
00:23:03,190 --> 00:23:04,890
显然有些股票上涨了

564
00:23:04,890 --> 00:23:05,710
有些股票下跌了

565
00:23:05,710 --> 00:23:07,740
但漫威的表现简直惊人

566
00:23:07,740 --> 00:23:09,359
所以我们进入漫威大概

567
00:23:09,359 --> 00:23:09,619
哦

568
00:23:09,619 --> 00:23:10,539
五个月前

569
00:23:10,539 --> 00:23:11,559
但这取决于什么

570
00:23:11,559 --> 00:23:12,500
呃，其实

571
00:23:12,500 --> 00:23:15,009
现在涨50%就是翻倍了

572
00:23:15,009 --> 00:23:17,690
但取决于每日价格波动

573
00:23:17,690 --> 00:23:21,509
不过重点在于，如今硬件领域变化很多

574
00:23:21,509 --> 00:23:23,609
但重点是如果你想加入会员

575
00:23:23,609 --> 00:23:26,839
并查看支撑这些观点的实际基本面分析

576
00:23:26,839 --> 00:23:29,079
请加入我们mekevin.com

577
00:23:29,079 --> 00:23:29,640
你可以加入会议

578
00:23:29,640 --> 00:23:30,319
凯文会员

579
00:23:30,319 --> 00:23:31,099
全部九门课程

580
00:23:31,099 --> 00:23:32,359
或许可以税前抵扣

581
00:23:32,359 --> 00:23:32,880
你的税务

582
00:23:32,880 --> 00:23:34,140
咨询你的税务专业人士

583
00:23:34,140 --> 00:23:35,099
给我们发送交易提醒

584
00:23:35,099 --> 00:23:35,940
每个私人直播

585
00:23:35,940 --> 00:23:37,569
每份阿尔法报告和朋友们

586
00:23:37,569 --> 00:23:38,789
我们下次再见

587
00:23:38,789 --> 00:23:39,529
再见，祝你好运

588
00:23:39,529 --> 00:23:41,650
为什么不在这里宣传这些内容呢

589
00:23:41,650 --> 00:23:43,390
我觉得没人知道这件事

590
00:23:43,390 --> 00:23:45,599
我们试试小范围广告看看效果

591
00:23:45,599 --> 00:23:46,380
恭喜

592
00:23:46,380 --> 00:23:46,640
兄弟

593
00:23:46,640 --> 00:23:47,420
你做了这么多

594
00:23:47,420 --> 00:23:48,240
大家都喜欢你

595
00:23:48,240 --> 00:23:49,180
大家都敬佩你

596
00:23:49,180 --> 00:23:50,220
凯文·帕弗拉特在此

597
00:23:50,220 --> 00:23:52,000
金融分析师兼YouTuber

598
00:23:52,000 --> 00:23:52,680
认识凯文

599
00:23:52,680 --> 00:23:54,059
很高兴听到你的见解

共 2 条，按点赞数排序。

采集完成，该视频无弹幕