1
00:00:19,600 --> 00:00:22,120
S1: Hey, what's up? So there's a common narrative right now

2
00:00:22,160 --> 00:00:24,840
S1: that you don't need to worry about AI. If you

3
00:00:24,880 --> 00:00:29,000
S1: are a knowledge worker, because humans are special and humans

4
00:00:29,000 --> 00:00:31,600
S1: are unique in the way they approach work. They are

5
00:00:31,600 --> 00:00:35,280
S1: better workers than AI. AI is just a text generator.

6
00:00:35,280 --> 00:00:38,080
S1: It's just an LLM. It has no understanding. It has

7
00:00:38,080 --> 00:00:41,760
S1: no knowledge. It doesn't really understand anything about what it's

8
00:00:41,760 --> 00:00:45,200
S1: trying to do. It can't possibly replace human workers. This

9
00:00:45,200 --> 00:00:48,720
S1: is a very popular narrative for multiple reasons. One, it

10
00:00:48,720 --> 00:00:50,960
S1: makes us feel better. It makes us feel more safe

11
00:00:50,960 --> 00:00:53,920
S1: because people are feeling afraid for good reason right now.

12
00:00:53,920 --> 00:00:58,960
S1: And it's just a very common thread right now, basically saying, relax,

13
00:00:58,960 --> 00:01:01,760
S1: you don't have to worry, right? AI is actually kind

14
00:01:01,760 --> 00:01:05,240
S1: of garbage and humans are basically awesome. You don't need

15
00:01:05,240 --> 00:01:08,000
S1: to worry. I think this is wrong. And more importantly,

16
00:01:08,000 --> 00:01:10,319
S1: I think it's dangerous. I think it's wrong for a

17
00:01:10,319 --> 00:01:12,480
S1: bunch of reasons, which I'm about to talk about. I

18
00:01:12,480 --> 00:01:15,640
S1: think it's dangerous because if you adopt this mindset, if

19
00:01:15,640 --> 00:01:19,150
S1: you accept this argument, you will not properly prepare to

20
00:01:19,190 --> 00:01:21,190
S1: get ready for what is coming, right? So this is

21
00:01:21,190 --> 00:01:24,150
S1: a polar opposite argument to the one that's being made.

22
00:01:24,150 --> 00:01:26,429
S1: And keep in mind, it's being made by some very

23
00:01:26,430 --> 00:01:28,710
S1: smart people. Right? I've got a friend we're going to

24
00:01:28,710 --> 00:01:31,749
S1: talk about Marcus Hutchins, very smart security guy, a friend

25
00:01:31,750 --> 00:01:34,390
S1: of mine, and we debated this once. Might do it

26
00:01:34,390 --> 00:01:37,350
S1: again in the future. He's very smart. He's very technical,

27
00:01:37,390 --> 00:01:41,350
S1: he's very knowledgeable. And he has this argument. He basically says,

28
00:01:41,350 --> 00:01:45,470
S1: don't worry, it's overinflated. AI doesn't understand anything. And it's

29
00:01:45,470 --> 00:01:48,070
S1: not just him. There's, you know, hundreds of people out

30
00:01:48,070 --> 00:01:51,550
S1: here saying this very loudly and more importantly to me

31
00:01:51,790 --> 00:01:54,550
S1: with a following, right? Marcus has a following. And there's

32
00:01:54,550 --> 00:01:57,350
S1: people who have much larger followings than Marcus and me

33
00:01:57,390 --> 00:01:59,829
S1: who are saying pretty much the same thing. And so

34
00:01:59,870 --> 00:02:02,950
S1: I want to really come at this argument directly, and

35
00:02:02,950 --> 00:02:06,190
S1: I want to start by talking about what work looks

36
00:02:06,190 --> 00:02:08,990
S1: like inside of companies and sort of just like break

37
00:02:09,030 --> 00:02:11,990
S1: down common things that I see all over the place

38
00:02:11,990 --> 00:02:14,429
S1: and have seen over the course of my career. All right.

39
00:02:14,430 --> 00:02:17,260
S1: So normally in these videos, I don't talk about my background.

40
00:02:17,300 --> 00:02:19,940
S1: It's because I just don't care about my background. I

41
00:02:19,940 --> 00:02:23,260
S1: don't care about other people's background. Usually. I mostly just

42
00:02:23,260 --> 00:02:24,700
S1: want to hear the content. I want to hear the

43
00:02:24,700 --> 00:02:27,419
S1: ideas that they're giving me, or if it's a tutorial

44
00:02:27,419 --> 00:02:29,980
S1: or whatever, I want the content, right? In this case,

45
00:02:29,980 --> 00:02:32,579
S1: I'm making a lot of claims about how businesses function,

46
00:02:32,579 --> 00:02:34,740
S1: how they work. And so I think it raises a

47
00:02:34,739 --> 00:02:37,940
S1: natural question which should be asked, which is, why should

48
00:02:37,940 --> 00:02:40,820
S1: I believe you? Why do I think you understand anything

49
00:02:40,820 --> 00:02:44,299
S1: about these companies? And the answer is, I've been working

50
00:02:44,299 --> 00:02:47,980
S1: inside of companies for like over 25 years. My background

51
00:02:48,019 --> 00:02:52,299
S1: is strictly information security, but I've held senior roles at Apple.

52
00:02:52,299 --> 00:02:55,139
S1: I've built a program and a product there which is

53
00:02:55,139 --> 00:02:58,380
S1: still running. I've been gone from there for a number

54
00:02:58,380 --> 00:03:01,740
S1: of years now. I've worked at Robinhood, HP. I've worked

55
00:03:01,739 --> 00:03:06,059
S1: in embedded in very large energy companies. Tons of experience actually,

56
00:03:06,060 --> 00:03:09,220
S1: in the fortune ten, let alone like the fortune 100.

57
00:03:09,220 --> 00:03:12,100
S1: And that's just the places that have actually been deeply

58
00:03:12,100 --> 00:03:15,600
S1: embedded in as a consultant, like basically a full time

59
00:03:15,600 --> 00:03:19,320
S1: worker there or working full time consulting though, which I

60
00:03:19,320 --> 00:03:22,239
S1: think is even more important for the context here. I've

61
00:03:22,239 --> 00:03:24,880
S1: actually worked for hundreds of companies in like the top

62
00:03:24,880 --> 00:03:28,880
S1: 10,000 companies in the world, like hundreds over these 25 years.

63
00:03:28,880 --> 00:03:31,719
S1: Inside of all these companies, I have seen these problems

64
00:03:31,720 --> 00:03:35,000
S1: that I'm about to describe over and over and over.

65
00:03:35,000 --> 00:03:37,400
S1: And in fact, when I describe them, I'm sure you

66
00:03:37,400 --> 00:03:40,000
S1: have also seen them. If you've worked in multiple companies

67
00:03:40,000 --> 00:03:41,800
S1: over the course of your career, all of these are

68
00:03:41,800 --> 00:03:44,760
S1: going to be extremely familiar. They're like not even controversial,

69
00:03:44,760 --> 00:03:47,240
S1: just a little bit of background. That's where I'm coming from.

70
00:03:47,240 --> 00:03:49,800
S1: And I just want to jump into this first part

71
00:03:49,840 --> 00:03:52,640
S1: here talking about the companies and their problems. So the

72
00:03:52,640 --> 00:03:56,280
S1: first thing to say about working in a, say, a

73
00:03:56,280 --> 00:03:59,520
S1: top 10,000 company or whatever, just an average company like

74
00:03:59,520 --> 00:04:03,200
S1: medium size, fairly large, we can talk about startups as well.

75
00:04:03,200 --> 00:04:06,600
S1: But in general, the natural state of thinking is like,

76
00:04:06,720 --> 00:04:09,320
S1: everything seems to be fine. Like, I don't even know

77
00:04:09,320 --> 00:04:12,440
S1: why we're talking about bringing AI into these companies. Isn't

78
00:04:12,440 --> 00:04:15,310
S1: everything going well? Like we have a good economy. Like, well,

79
00:04:15,310 --> 00:04:17,910
S1: we have had a good economy, right? We've built plenty

80
00:04:17,910 --> 00:04:20,950
S1: of companies. We've got plenty of market share. The Dow

81
00:04:20,990 --> 00:04:23,390
S1: was doing great or whatever for all these years, all

82
00:04:23,390 --> 00:04:26,310
S1: these decades. So what actually is the problem? The way

83
00:04:26,310 --> 00:04:29,910
S1: I answer that is most companies are actually a giant

84
00:04:29,910 --> 00:04:33,110
S1: soup sandwich or a football bat. These are military terms,

85
00:04:33,110 --> 00:04:37,950
S1: but essentially there is absolute chaos in most companies. Most companies,

86
00:04:37,950 --> 00:04:40,990
S1: it's actually difficult to find out what they're actually trying

87
00:04:40,990 --> 00:04:43,790
S1: to do, what they're actually trying to build, especially as

88
00:04:43,830 --> 00:04:45,990
S1: like a worker, right? It's very hard to get a

89
00:04:45,990 --> 00:04:48,910
S1: clear message of like, what exactly is the mission? This

90
00:04:48,910 --> 00:04:51,870
S1: is also the case for like fairly large startups as well.

91
00:04:51,870 --> 00:04:54,430
S1: It starts to become a problem fairly early, you know,

92
00:04:54,470 --> 00:04:56,910
S1: a few years into the startup, the vision is not

93
00:04:56,910 --> 00:04:59,510
S1: super clear. Okay, so the first problem I want to

94
00:04:59,510 --> 00:05:01,310
S1: talk about that I see in a lot of these

95
00:05:01,310 --> 00:05:03,790
S1: companies and that I'm sure you've seen as well, is

96
00:05:03,790 --> 00:05:07,670
S1: the vision isn't always very clear. In fact, sometimes it's non-existent,

97
00:05:07,670 --> 00:05:10,870
S1: sometimes it's not conveyed all the way through the company.

98
00:05:10,870 --> 00:05:13,620
S1: So people like aren't really sure exactly what they're working

99
00:05:13,620 --> 00:05:16,339
S1: on and why. And so that's a huge problem. Another

100
00:05:16,340 --> 00:05:19,140
S1: one is it's not clear. And this is kind of

101
00:05:19,180 --> 00:05:21,740
S1: like fighting for top here. It's hard to say. These

102
00:05:21,740 --> 00:05:24,540
S1: are really big issues. Hard to sort them. But another

103
00:05:24,540 --> 00:05:27,980
S1: really big one is there aren't really SOPs. There aren't

104
00:05:27,980 --> 00:05:31,300
S1: really standard operating procedures. That's what we call them in

105
00:05:31,300 --> 00:05:35,100
S1: the military is SOPs. Most companies just call them processes,

106
00:05:35,100 --> 00:05:37,180
S1: but like, how do you actually do things? And the

107
00:05:37,180 --> 00:05:40,260
S1: problem isn't that there isn't some document somewhere that says

108
00:05:40,260 --> 00:05:42,220
S1: how to do things. The problem is there might be

109
00:05:42,220 --> 00:05:44,980
S1: multiple of them and they change all the time, or

110
00:05:44,980 --> 00:05:47,980
S1: they decay over time, or people are referring to the

111
00:05:47,980 --> 00:05:51,779
S1: wrong one. Okay, so it's like things just keep getting

112
00:05:51,780 --> 00:05:55,180
S1: done incorrectly throughout the company. And the more layers of

113
00:05:55,180 --> 00:05:58,380
S1: management you have, the larger the company, the longer the

114
00:05:58,380 --> 00:06:01,500
S1: company's been around. The more employees you have, the worse

115
00:06:01,500 --> 00:06:04,860
S1: this problem gets. Now I want to caveat something. There

116
00:06:04,860 --> 00:06:07,660
S1: are certain companies, there are a few companies, very few

117
00:06:07,660 --> 00:06:10,410
S1: companies who do this really well. So if you are

118
00:06:10,450 --> 00:06:13,090
S1: one of these people who works at these companies or

119
00:06:13,090 --> 00:06:15,210
S1: you're a manager at one of these companies and you're like, well,

120
00:06:15,250 --> 00:06:17,850
S1: that's not true. I work at Google and it's pretty great.

121
00:06:17,850 --> 00:06:21,130
S1: Or I work at so-and-so Giant Bank, which has had

122
00:06:21,130 --> 00:06:23,650
S1: decades to figure this out and sort it out. Also,

123
00:06:23,650 --> 00:06:26,810
S1: in energy companies, they tend to be extremely risk averse,

124
00:06:26,810 --> 00:06:29,570
S1: and they have processes that they follow and they're really

125
00:06:29,570 --> 00:06:33,530
S1: good at processes. I'm talking about the other 99.9%. Most

126
00:06:33,530 --> 00:06:37,170
S1: companies it's not clear exactly how something is supposed to

127
00:06:37,170 --> 00:06:39,610
S1: be done. And what ends up happening when this is

128
00:06:39,610 --> 00:06:42,770
S1: the case is you have people doing work and the

129
00:06:42,770 --> 00:06:47,730
S1: output is massively inconsistent. Okay? Massively inconsistent, different people doing

130
00:06:47,730 --> 00:06:51,290
S1: that same job, even using the same document. They do

131
00:06:51,290 --> 00:06:54,610
S1: the job differently. You know, it's common knowledge within the company. Oh,

132
00:06:54,650 --> 00:06:57,410
S1: if Sarah does it well, yeah, sure. I love when

133
00:06:57,410 --> 00:07:00,930
S1: Sarah produces this report or produces this security assessment or

134
00:07:00,970 --> 00:07:03,370
S1: produces this artifact that we're going to send to the customer,

135
00:07:03,370 --> 00:07:06,050
S1: she does it really well. Well, guess what, Jim over here,

136
00:07:06,050 --> 00:07:08,960
S1: he looked at the same process. He got the same training.

137
00:07:08,960 --> 00:07:12,720
S1: He had the same onboarding. His output is absolute garbage. Okay.

138
00:07:12,760 --> 00:07:15,680
S1: That's even if there is a process in the company,

139
00:07:15,680 --> 00:07:19,280
S1: oftentimes there isn't. It was just like, you know, common onboarding, right?

140
00:07:19,320 --> 00:07:21,120
S1: You work at a company or whatever you come in,

141
00:07:21,160 --> 00:07:23,320
S1: it's like, hey, you know, here's where you put your keys.

142
00:07:23,320 --> 00:07:26,000
S1: We get lunch over here, you know, basically your day

143
00:07:26,000 --> 00:07:27,560
S1: to day is going to be, you've got to take

144
00:07:27,560 --> 00:07:29,720
S1: these emails, you've got to take it from here. Oh,

145
00:07:29,760 --> 00:07:32,400
S1: make sure you don't send it to Mr. Johnson like this.

146
00:07:32,400 --> 00:07:35,000
S1: He really hates that. Right. And so these are all

147
00:07:35,080 --> 00:07:37,720
S1: being sort of loosely like stored in your mind. Let's

148
00:07:37,720 --> 00:07:40,040
S1: say you're some sort of admin or something like that, right?

149
00:07:40,080 --> 00:07:43,080
S1: So this admin named Chris suddenly starts learning all these

150
00:07:43,080 --> 00:07:46,640
S1: things during onboarding. Yeah. Mr. Johnson doesn't like this. Make

151
00:07:46,680 --> 00:07:48,760
S1: sure the format, if you send it to the group,

152
00:07:48,800 --> 00:07:51,280
S1: make sure to include these people. Right. You're starting to

153
00:07:51,280 --> 00:07:53,720
S1: get all these little things that you're picking up, either

154
00:07:53,720 --> 00:07:57,000
S1: from the training or someone complaining about something and it

155
00:07:57,000 --> 00:07:59,840
S1: starts accruing knowledge, right? The problem is that's stuck inside

156
00:07:59,840 --> 00:08:03,160
S1: that one person's head. And based on how knowledgeable they are,

157
00:08:03,200 --> 00:08:05,640
S1: based on how good they are at listening, you know

158
00:08:05,680 --> 00:08:08,710
S1: how good they are at following processes like, did we

159
00:08:08,710 --> 00:08:10,630
S1: actually write this down or is it stuck in my

160
00:08:10,630 --> 00:08:13,590
S1: brain and me? Chris, can I just leave the company?

161
00:08:13,590 --> 00:08:16,270
S1: And now that whole process has to start all over again.

162
00:08:16,310 --> 00:08:19,110
S1: Very common stuff. You've seen this a million times. Another

163
00:08:19,110 --> 00:08:21,750
S1: thing you've seen related to this million times is there

164
00:08:21,750 --> 00:08:24,630
S1: are people in the company all throughout the company who

165
00:08:24,830 --> 00:08:28,430
S1: just never, ever got good at doing the job. They

166
00:08:28,430 --> 00:08:31,990
S1: never got good at following the process. They they don't

167
00:08:31,990 --> 00:08:35,309
S1: produce good output. It's really, really hard to hire people

168
00:08:35,309 --> 00:08:38,190
S1: in companies. It's also really, really hard to fire people

169
00:08:38,189 --> 00:08:41,390
S1: at companies. This person, Chris or Mike or Raj or

170
00:08:41,390 --> 00:08:44,750
S1: Sarah or whoever, can just be horrible at doing this

171
00:08:44,750 --> 00:08:48,510
S1: for whatever reason. Maybe it's nepotism, maybe it's, you know,

172
00:08:48,550 --> 00:08:51,990
S1: they're a nice person, maybe they're good at one little thing. And,

173
00:08:52,030 --> 00:08:54,590
S1: you know, it's just you would hate to give that up. So,

174
00:08:54,830 --> 00:08:57,429
S1: you know, they can stay on and keep breaking things

175
00:08:57,430 --> 00:09:00,430
S1: and messing things up when they execute work other places.

176
00:09:00,430 --> 00:09:04,350
S1: The point is this is endemic throughout companies. In a

177
00:09:04,350 --> 00:09:08,010
S1: large company, there are dozens or hundreds, maybe even thousands

178
00:09:08,010 --> 00:09:11,490
S1: of people who don't follow processes. They aren't good at

179
00:09:11,530 --> 00:09:14,570
S1: their job at all. Right. Now, keep in mind there

180
00:09:14,570 --> 00:09:18,650
S1: are also a players everywhere, right? There's a players everywhere

181
00:09:18,650 --> 00:09:20,810
S1: and they do a fantastic job. They don't even need

182
00:09:20,810 --> 00:09:23,770
S1: a process. They're awesome right. No one's disputing that there

183
00:09:23,770 --> 00:09:26,970
S1: are awesome humans and that they can do work. That's

184
00:09:26,970 --> 00:09:29,730
S1: not what this argument is about. This argument is about

185
00:09:29,730 --> 00:09:33,250
S1: overall work. Okay, just a pop quiz here. Guess how

186
00:09:33,250 --> 00:09:38,850
S1: much knowledge workers are paid every year? $50 trillion. That

187
00:09:38,850 --> 00:09:44,090
S1: is the compensation roughly worldwide for knowledge workers. Okay, that

188
00:09:44,090 --> 00:09:46,730
S1: includes the B players and C players and the D

189
00:09:46,730 --> 00:09:49,530
S1: players and the F players, right? Forget the A players.

190
00:09:49,530 --> 00:09:51,850
S1: We're not going to talk about them. Right. They're awesome.

191
00:09:51,850 --> 00:09:54,330
S1: They're going to continue to be awesome with or without AI.

192
00:09:54,370 --> 00:09:58,410
S1: Forget that we're talking about most work in most companies. Okay.

193
00:09:58,450 --> 00:10:02,050
S1: So you've got these people. They can't follow the process. Okay.

194
00:10:02,210 --> 00:10:04,880
S1: They produce bad work. Now you've got the fact that

195
00:10:04,880 --> 00:10:07,720
S1: the job isn't that rewarding. They're not exactly sure what

196
00:10:07,720 --> 00:10:10,520
S1: they're working on. Their bosses are changing all the time,

197
00:10:10,520 --> 00:10:13,240
S1: by the way, their work that they're supposed to be doing,

198
00:10:13,240 --> 00:10:16,319
S1: it's also changing all the time, right? Because the the

199
00:10:16,320 --> 00:10:18,760
S1: company goals, you know, that they were given a long

200
00:10:18,760 --> 00:10:21,280
S1: time ago, they change all the time. Guess what? That

201
00:10:21,280 --> 00:10:23,160
S1: has to trickle down. It has to go down to

202
00:10:23,200 --> 00:10:26,559
S1: a thing. Guess what that means. Meetings. Lots of meetings. Right?

203
00:10:26,600 --> 00:10:28,720
S1: So your developer, you're trying to code, you're trying to

204
00:10:28,720 --> 00:10:30,880
S1: do whatever. You're an engineer, you're trying to do a

205
00:10:30,880 --> 00:10:32,800
S1: bunch of technical stuff. Nope. You got to get on

206
00:10:32,800 --> 00:10:35,440
S1: this call. Why? Because so and so all these people

207
00:10:35,440 --> 00:10:37,760
S1: need to have this thing explained to them. You already

208
00:10:37,760 --> 00:10:40,800
S1: made a document. You already explained it to them 20 times.

209
00:10:40,800 --> 00:10:44,400
S1: You've already been in 40 meetings about this very same topic.

210
00:10:44,400 --> 00:10:47,320
S1: Nobody is listening. They keep doing the work. You keep

211
00:10:47,320 --> 00:10:50,280
S1: having to correct the work that they did because they

212
00:10:50,280 --> 00:10:54,040
S1: didn't follow the document, which you spent three weeks making, right?

213
00:10:54,080 --> 00:10:56,760
S1: This is not even special. This is not like a

214
00:10:56,760 --> 00:10:59,960
S1: surprise to you. If you've worked in any companies, most

215
00:10:59,960 --> 00:11:04,150
S1: companies across your career, this is endemic now. Add on

216
00:11:04,150 --> 00:11:07,590
S1: top of this, people are empire building. It's fucking Game

217
00:11:07,590 --> 00:11:11,070
S1: of Thrones is what's being played out in real time

218
00:11:11,110 --> 00:11:16,709
S1: across many companies for many employees, right? Managers are empire building.

219
00:11:16,710 --> 00:11:20,110
S1: They're picking their favorite people also. What about when you

220
00:11:20,110 --> 00:11:21,950
S1: have a good idea? How many times have you had

221
00:11:21,950 --> 00:11:24,910
S1: a good idea? You submit it up through management because

222
00:11:24,910 --> 00:11:28,270
S1: it actually would solve half the stuff that just ruins

223
00:11:28,270 --> 00:11:30,550
S1: your soul. It crushes your soul to have to do

224
00:11:30,550 --> 00:11:33,830
S1: this stupid work, which doesn't follow this process, which could

225
00:11:33,830 --> 00:11:37,230
S1: easily be fixed by these things that you're recommending, right?

226
00:11:37,230 --> 00:11:39,350
S1: You're like, look, if we just did this, it would

227
00:11:39,350 --> 00:11:41,710
S1: cut four steps out of the process. It would cut

228
00:11:41,710 --> 00:11:44,590
S1: four steps out of the process. Everything would be much better.

229
00:11:44,710 --> 00:11:47,990
S1: I spent three weeks, I put together this proposal. I

230
00:11:48,030 --> 00:11:50,670
S1: wrote all this documentation and they're like, yeah, yeah, I

231
00:11:50,670 --> 00:11:52,510
S1: don't think so. Oh, by the way, you need to

232
00:11:52,510 --> 00:11:55,110
S1: join this meeting. We're changing our goals again. Oh, by

233
00:11:55,110 --> 00:11:57,390
S1: the way, you're going to report to Sarah now instead

234
00:11:57,390 --> 00:11:59,950
S1: of Raj. And also we need you to redo all

235
00:11:59,950 --> 00:12:02,620
S1: that stuff that you just did. And it's like, okay, great.

236
00:12:02,620 --> 00:12:05,940
S1: Sounds awesome. Why do you think people dread Monday so much?

237
00:12:05,980 --> 00:12:08,340
S1: Why do you think people on Sunday night are just, like,

238
00:12:08,380 --> 00:12:11,140
S1: filled with this pit of despair because they get to

239
00:12:11,180 --> 00:12:13,500
S1: maybe get in a car and drive for an hour

240
00:12:13,500 --> 00:12:17,060
S1: to commute in, to go to another meeting, to redo

241
00:12:17,100 --> 00:12:20,660
S1: their plans, hopefully try to get a bonus, right? Meanwhile,

242
00:12:20,660 --> 00:12:23,380
S1: everyone around them is playing Game of Thrones and like

243
00:12:23,380 --> 00:12:27,220
S1: breaking everything. Now I'm being a little overdramatic because there

244
00:12:27,220 --> 00:12:29,819
S1: are some jobs where it's kind of chill and blah

245
00:12:29,820 --> 00:12:33,140
S1: blah blah. It's everything's fine. But in general, it has

246
00:12:33,140 --> 00:12:36,540
S1: been known for decades that this is a problem. It

247
00:12:36,540 --> 00:12:39,980
S1: has been known that this dysfunction has been happening for decades.

248
00:12:39,980 --> 00:12:43,140
S1: People were complaining about the same stuff. Look at office space.

249
00:12:43,140 --> 00:12:47,780
S1: We've been making shows about corporate drudgery and stupidity for

250
00:12:47,780 --> 00:12:53,100
S1: decade upon decade. We recently had severance soul killing work, right?

251
00:12:53,100 --> 00:12:55,060
S1: You're sitting in the queue, blah blah blah. The manager

252
00:12:55,060 --> 00:12:58,260
S1: comes by and smiles like nobody's having fun here. That

253
00:12:58,260 --> 00:13:01,650
S1: manager's not having fun. You're not having fun. Why? Because

254
00:13:01,650 --> 00:13:04,210
S1: your ideas are not getting into the system. You're not

255
00:13:04,210 --> 00:13:06,970
S1: even clear exactly what you're working on and why. There's

256
00:13:06,970 --> 00:13:10,569
S1: multiple layers separating between you and like whatever's happening at

257
00:13:10,570 --> 00:13:12,929
S1: the top, which seems to change all the time, right?

258
00:13:12,970 --> 00:13:17,210
S1: So this is not a good state. Okay. Human work

259
00:13:17,210 --> 00:13:22,610
S1: is not the ideal state. Workflows, processes. What confidence do

260
00:13:22,610 --> 00:13:25,970
S1: you have that if you have a nine step process,

261
00:13:26,010 --> 00:13:28,610
S1: each of which has an SOP for it, and you

262
00:13:28,610 --> 00:13:31,050
S1: give it to the new people that just came on

263
00:13:31,050 --> 00:13:33,850
S1: and got onboarded and got trained, that the people are

264
00:13:33,850 --> 00:13:35,730
S1: going to actually read the document and they're going to

265
00:13:35,730 --> 00:13:38,290
S1: execute those nine steps. And at the end of it,

266
00:13:38,290 --> 00:13:39,970
S1: you're going to have an artifact. You're going to have

267
00:13:39,970 --> 00:13:43,610
S1: an outcome, a document project, whatever it is, and it's

268
00:13:43,610 --> 00:13:45,850
S1: going to meet the standard. And what confidence do you

269
00:13:45,850 --> 00:13:48,810
S1: have that that will also be the same the next

270
00:13:48,810 --> 00:13:51,170
S1: time you run it. Because guess what, two of those

271
00:13:51,170 --> 00:13:53,730
S1: people are on vacation and they didn't share their knowledge

272
00:13:53,730 --> 00:13:56,730
S1: with anyone. Also, they didn't update the document because there

273
00:13:56,730 --> 00:14:00,550
S1: was a change. This whole thing is chaos and decay

274
00:14:00,550 --> 00:14:04,070
S1: and problems. Meanwhile, people are getting sick. People are changing

275
00:14:04,070 --> 00:14:06,870
S1: jobs like every 2 or 3 years or whatever, depending

276
00:14:06,870 --> 00:14:09,750
S1: on the industry. Then people are getting so like soul

277
00:14:09,750 --> 00:14:14,910
S1: crushed that they're doing the bare minimum, right? I think 23%

278
00:14:14,910 --> 00:14:18,310
S1: of total working time in a week is actually spent.

279
00:14:18,350 --> 00:14:21,990
S1: Like 40 hours is like 23% of a full week, right?

280
00:14:22,030 --> 00:14:24,710
S1: So people aren't working the full week. They're only working

281
00:14:24,710 --> 00:14:27,790
S1: this little slot at work. Many people are doing the

282
00:14:27,830 --> 00:14:31,750
S1: absolute bare minimum. They're making a fucking sport out of

283
00:14:31,790 --> 00:14:34,550
S1: doing the least amount of work possible, right? That was

284
00:14:34,550 --> 00:14:37,470
S1: also part of office space. Like this is extremely well

285
00:14:37,510 --> 00:14:41,950
S1: known stuff. We are holding this situation up as the

286
00:14:41,950 --> 00:14:46,710
S1: standard that AI can never achieve. AI can never, never

287
00:14:46,710 --> 00:14:50,390
S1: pierce this because humans are just so innovative, right? It's

288
00:14:50,430 --> 00:14:53,990
S1: absolutely insane to me. So another problem here is actually

289
00:14:54,030 --> 00:14:56,430
S1: because I've been talking mostly about the worker, right? Mostly

290
00:14:56,430 --> 00:14:59,660
S1: about the human doing this job, getting soul crushed. Let

291
00:14:59,660 --> 00:15:01,460
S1: me tell you about the other side of it as

292
00:15:01,460 --> 00:15:04,660
S1: a leader. Okay? Looking down because I've been here as well.

293
00:15:04,660 --> 00:15:07,580
S1: So you're looking down at this company and you're trying

294
00:15:07,580 --> 00:15:09,980
S1: to figure out you're looking with the CFO, you're looking

295
00:15:09,980 --> 00:15:12,780
S1: at your headcount spend and you're like, okay, well, I'm

296
00:15:12,780 --> 00:15:19,220
S1: paying $94 million on headcount across my entire staff, you know, engineers, admins,

297
00:15:19,260 --> 00:15:23,100
S1: like all these different people. I'm paying $94 million and

298
00:15:23,140 --> 00:15:25,020
S1: that's not even a lot of money. Let's just say

299
00:15:25,020 --> 00:15:28,980
S1: that's a total cost of human compensation. Okay, what exactly

300
00:15:28,980 --> 00:15:32,060
S1: are we getting? Can I see all of my projects?

301
00:15:32,100 --> 00:15:34,420
S1: Can I see all of the work that's being done?

302
00:15:34,460 --> 00:15:37,780
S1: Do I have a list of workflows that show all

303
00:15:37,780 --> 00:15:41,100
S1: the different processes in my company, what the steps are

304
00:15:41,180 --> 00:15:44,540
S1: and how they're actually being performed by the different teams?

305
00:15:44,540 --> 00:15:46,740
S1: Do I know how much that costs? Do I know

306
00:15:46,740 --> 00:15:50,220
S1: how long it takes to actually execute any of these things?

307
00:15:50,220 --> 00:15:53,740
S1: Do you know? I think you do. But rhetorically, do

308
00:15:53,740 --> 00:15:57,330
S1: you know what happens when a leader asks this question

309
00:15:57,330 --> 00:16:00,010
S1: to their staff, to their company. Hey, I need a

310
00:16:00,010 --> 00:16:02,330
S1: list of all work being done. I need a list

311
00:16:02,330 --> 00:16:04,970
S1: of what everyone is doing. I need a list of

312
00:16:04,970 --> 00:16:08,250
S1: how much this work is costing and how expensive it

313
00:16:08,250 --> 00:16:11,570
S1: is at the different stages. Oh, also, is it good work?

314
00:16:11,690 --> 00:16:14,210
S1: What's the average quality of an output of one of

315
00:16:14,210 --> 00:16:16,770
S1: these processes? Do you know what the answer is? The

316
00:16:16,770 --> 00:16:19,730
S1: answer is. Well, we could take a look. That's a

317
00:16:19,730 --> 00:16:22,650
S1: good question, ma'am. We've been thinking about that as well. Yeah.

318
00:16:22,650 --> 00:16:25,010
S1: If you would like, I could put together a project.

319
00:16:25,010 --> 00:16:27,050
S1: What we could do is we could put together a

320
00:16:27,050 --> 00:16:29,130
S1: list of all the stuff that we're doing and we

321
00:16:29,130 --> 00:16:33,290
S1: could figure out, you know, what exactly our work is

322
00:16:33,330 --> 00:16:35,290
S1: and how much it costs, and you know how good

323
00:16:35,290 --> 00:16:38,330
S1: of a job we're doing. That is a major endeavor. Oh,

324
00:16:38,330 --> 00:16:41,410
S1: by the way, it's separate from their work. This happens

325
00:16:41,410 --> 00:16:46,050
S1: constantly inside of companies. Constantly. I'm often brought in to

326
00:16:46,090 --> 00:16:48,650
S1: help them do this. I am often there when they

327
00:16:48,690 --> 00:16:51,490
S1: are doing it with someone else. So if they spin

328
00:16:51,490 --> 00:16:53,930
S1: up a project like this. Okay. Do you think this

329
00:16:53,930 --> 00:17:00,000
S1: takes hours. Nope. Days? Nope. Weeks? Nope. Months. This is multi-week, multi-month,

330
00:17:00,000 --> 00:17:03,640
S1: or multi-year. This is so work intensive. Keep in mind

331
00:17:03,640 --> 00:17:06,160
S1: what has to happen to actually gather all this stuff.

332
00:17:06,280 --> 00:17:08,919
S1: You actually have to call a bunch of meetings. Guess what?

333
00:17:08,920 --> 00:17:10,399
S1: You have to tell them. Hey, I want you to

334
00:17:10,440 --> 00:17:13,480
S1: put together the list of everything you're working on. I

335
00:17:13,480 --> 00:17:15,840
S1: want you to note all the different tasks that you

336
00:17:15,840 --> 00:17:18,359
S1: do during a day. This is our favorite. This is

337
00:17:18,359 --> 00:17:20,840
S1: our absolute favorite thing to do. Yeah. Let me tell

338
00:17:20,840 --> 00:17:23,080
S1: my manager exactly what I do. Why don't you look

339
00:17:23,080 --> 00:17:25,399
S1: at my output? Why don't you look at my schedule?

340
00:17:25,520 --> 00:17:30,440
S1: This is happening constantly while I'm speaking. Some managers, hundreds

341
00:17:30,439 --> 00:17:33,640
S1: of them, thousands of them are telling their people, I

342
00:17:33,639 --> 00:17:36,200
S1: need a detailed report of what you do and what

343
00:17:36,199 --> 00:17:38,199
S1: you work on, and how much it costs and how

344
00:17:38,199 --> 00:17:40,840
S1: long it takes, so I can put it up the chain,

345
00:17:40,840 --> 00:17:43,199
S1: that's all. So they can get this visibility. And you

346
00:17:43,199 --> 00:17:45,519
S1: know what's even crazier? If it's a large company, they

347
00:17:45,520 --> 00:17:48,239
S1: don't even ask their people that. They're like, hey, let's

348
00:17:48,240 --> 00:17:53,990
S1: bring in McKinsey or KPMG. Let's spend Thousand dollars to

349
00:17:54,030 --> 00:17:57,589
S1: have a team of smiling 22 year olds come in

350
00:17:57,590 --> 00:18:01,389
S1: here and do the audit for us. This is ridiculous.

351
00:18:01,429 --> 00:18:05,430
S1: This is what AI is competing with. Okay, $50 trillion,

352
00:18:05,510 --> 00:18:08,710
S1: by the way. It's around $10 trillion just for the US.

353
00:18:08,750 --> 00:18:13,550
S1: $50 trillion worldwide is spent. The CEO and the CFO

354
00:18:13,590 --> 00:18:16,949
S1: have very little idea what all is happening in the company.

355
00:18:16,949 --> 00:18:19,070
S1: They don't know where the money is going. They don't

356
00:18:19,070 --> 00:18:21,350
S1: know what the processes are. They don't know what the

357
00:18:21,389 --> 00:18:24,510
S1: workflows are. They don't know how well it's going. They

358
00:18:24,510 --> 00:18:27,189
S1: just have a rough idea based on, you know, performance

359
00:18:27,189 --> 00:18:29,669
S1: of the company or whatever. Like when they hear from

360
00:18:29,669 --> 00:18:32,669
S1: their managers and they see the stock doing whatever and

361
00:18:32,709 --> 00:18:35,429
S1: obviously they have control, they can say, go do this,

362
00:18:35,429 --> 00:18:38,549
S1: go do that. There is no visibility. They can't see shit.

363
00:18:38,590 --> 00:18:42,310
S1: This is the state that AI is competing with, which

364
00:18:42,310 --> 00:18:46,830
S1: is roughly chaos. Okay, so when you hear the argument,

365
00:18:46,830 --> 00:18:49,270
S1: I just want this to crystallize in your mind. When

366
00:18:49,270 --> 00:18:52,899
S1: you hear the argument, AI cannot compete with humans inside

367
00:18:52,899 --> 00:18:55,300
S1: of a company. This is what they have to compete with.

368
00:18:55,500 --> 00:18:58,020
S1: This is a bar that is not low. It is

369
00:18:58,020 --> 00:19:00,939
S1: on the floor. It. It burned a hole through the floor.

370
00:19:00,939 --> 00:19:04,020
S1: It is descending to the. The core of the earth.

371
00:19:04,020 --> 00:19:07,980
S1: This is a low bar. AI can follow instructions. We're

372
00:19:07,980 --> 00:19:09,739
S1: not even going to talk about AI yet, but the

373
00:19:09,740 --> 00:19:12,859
S1: ability to follow instructions over and over, doing the same

374
00:19:12,859 --> 00:19:17,580
S1: thing in a very uncreative, shitty way is better than

375
00:19:17,580 --> 00:19:20,179
S1: what is done in most companies most of the time. Like,

376
00:19:20,219 --> 00:19:23,899
S1: I cannot over express how bad the situation is and

377
00:19:23,899 --> 00:19:26,779
S1: how low this bar is. So the crazy thing about

378
00:19:26,780 --> 00:19:31,420
S1: this is everyone knew this before 2022. Everyone knew that

379
00:19:31,419 --> 00:19:36,020
S1: this was soul crushing, shitty corporate knowledge work. Everyone knew

380
00:19:36,060 --> 00:19:38,179
S1: that this was just like horrible. And I'm not talking

381
00:19:38,179 --> 00:19:40,580
S1: about like some people who love their jobs 100%. A

382
00:19:40,580 --> 00:19:43,219
S1: lot of people love their jobs. They're lucky. Maybe it

383
00:19:43,219 --> 00:19:45,819
S1: lasts for a while, maybe it's short lived. Whatever. There

384
00:19:45,820 --> 00:19:48,379
S1: are some people who love their jobs. Some companies are great,

385
00:19:48,379 --> 00:19:51,999
S1: they tend to be smaller, but whatever. In general, everyone

386
00:19:52,000 --> 00:19:55,760
S1: knew that this was soul crushing before 2022. They knew

387
00:19:55,840 --> 00:19:59,040
S1: it before 2022. They knew it in the 90s. They

388
00:19:59,040 --> 00:20:01,600
S1: knew it in the 80s. Like, that's why it's part

389
00:20:01,600 --> 00:20:04,639
S1: of culture. Everyone knew this was a problem and guess

390
00:20:04,639 --> 00:20:07,040
S1: what we were doing? Guess what everyone was doing throughout

391
00:20:07,040 --> 00:20:09,720
S1: the culture, they're trying to figure out a solution. What

392
00:20:09,719 --> 00:20:13,000
S1: is a better way? Because this is obviously not the

393
00:20:13,000 --> 00:20:15,840
S1: way that humans should live. That was the state before

394
00:20:15,879 --> 00:20:19,760
S1: 2022 when AI arrived. Okay, so another argument I'm going

395
00:20:19,800 --> 00:20:23,559
S1: to continue beating up on us humans here. One last piece.

396
00:20:23,600 --> 00:20:26,319
S1: I got positive stuff at the end to say about humans.

397
00:20:26,320 --> 00:20:28,760
S1: So don't worry about that. You can hang in for that.

398
00:20:28,760 --> 00:20:32,480
S1: This next piece is another negative piece about humans, and

399
00:20:32,480 --> 00:20:35,319
S1: it's actually something that a lot of people don't think about.

400
00:20:35,320 --> 00:20:38,720
S1: So and it relates to how AI thinks versus how

401
00:20:38,719 --> 00:20:42,000
S1: humans think. Another sort of part of this argument is

402
00:20:42,000 --> 00:20:45,800
S1: that humans just fundamentally think differently. And we just have

403
00:20:45,840 --> 00:20:49,709
S1: like a lot more integrity of thought. We're more cohesive.

404
00:20:49,750 --> 00:20:54,670
S1: We have understanding. We know ourselves. You know we can introspect.

405
00:20:54,709 --> 00:20:58,350
S1: We can be creative. We have intuition. We have all

406
00:20:58,350 --> 00:21:01,749
S1: these advantages over an AI which is like this, you know,

407
00:21:01,830 --> 00:21:04,789
S1: dead black box, basically. So I want to sort of

408
00:21:04,830 --> 00:21:08,069
S1: like dispel this as well. This is another angle to

409
00:21:08,109 --> 00:21:11,749
S1: this argument, which also makes absolutely no sense. The crux

410
00:21:11,750 --> 00:21:14,990
S1: of this argument is that humans are better thinking machines.

411
00:21:14,990 --> 00:21:18,550
S1: Humans are much better thinking machines than AI. If you've

412
00:21:18,550 --> 00:21:22,509
S1: ever studied or read any meditation documentation, or tried to

413
00:21:22,550 --> 00:21:24,669
S1: learn it, or tried to study it or whatever, what

414
00:21:24,669 --> 00:21:28,670
S1: it reveals about humans is absolutely mind blowing. Okay, so

415
00:21:28,669 --> 00:21:32,109
S1: the first thing that it reveals is that humans have

416
00:21:32,109 --> 00:21:35,510
S1: no idea what they're thinking at any given moment. Humans

417
00:21:35,510 --> 00:21:38,629
S1: are basically like, if you could just zoom in and

418
00:21:38,629 --> 00:21:42,149
S1: listen to the minds of people. Okay, sitting in their cars,

419
00:21:42,350 --> 00:21:45,310
S1: sitting in their office space or their cube at work

420
00:21:45,310 --> 00:21:48,179
S1: or wherever, you're just going to zoom in on these

421
00:21:48,179 --> 00:21:51,379
S1: conversations that are happening in their mind. It's like hamsters

422
00:21:51,379 --> 00:21:54,340
S1: running on a wheel. You'll just zoom into like Ravi here,

423
00:21:54,340 --> 00:21:56,660
S1: and Ravi is like, just can't believe he said that.

424
00:21:56,659 --> 00:21:58,740
S1: All that work I put in in the project, like,

425
00:21:58,740 --> 00:22:00,539
S1: I hope that guy gets fired. I just hope he

426
00:22:00,540 --> 00:22:02,500
S1: gets fired. Like, you know what I would love? Here's

427
00:22:02,500 --> 00:22:05,179
S1: what I would love. Just imagine like, okay, so a

428
00:22:05,179 --> 00:22:07,739
S1: new boss comes in and the new boss sees my

429
00:22:07,740 --> 00:22:10,139
S1: stuff and just absolutely loves it. And then he looks

430
00:22:10,139 --> 00:22:13,260
S1: at Chris's stuff. Can't you know, Chris, I can't believe

431
00:22:13,260 --> 00:22:15,859
S1: that happened. Yeah, yeah. So looks at Chris's stuff and

432
00:22:15,859 --> 00:22:18,220
S1: he's like, oh, this is garbage. You know, the one

433
00:22:18,219 --> 00:22:20,180
S1: I really like is I like this one. And then

434
00:22:20,179 --> 00:22:22,899
S1: that's me. That's mine. Like that would be so much better.

435
00:22:22,899 --> 00:22:25,300
S1: And then, oh, did I feed the dog? I think

436
00:22:25,300 --> 00:22:27,260
S1: I need to feed the dog. Hey, let me text someone.

437
00:22:27,260 --> 00:22:29,460
S1: I need to feed the dog. Oh, crap. It's almost lunch.

438
00:22:29,459 --> 00:22:31,580
S1: I got to get ready for lunch. That is the

439
00:22:31,580 --> 00:22:35,699
S1: mind of a person. That randomness is happening constantly. Now

440
00:22:35,699 --> 00:22:38,059
S1: check this out. When you try to meditate, what you

441
00:22:38,060 --> 00:22:40,580
S1: do is you try to observe thoughts. It is the

442
00:22:40,580 --> 00:22:42,659
S1: craziest thing to try to do if you try to

443
00:22:42,699 --> 00:22:44,859
S1: observe your thoughts. So you can pause this if you

444
00:22:44,859 --> 00:22:46,889
S1: want and try to observe your thoughts for a second.

445
00:22:46,929 --> 00:22:50,169
S1: It's a really fun and crazy exercise. It is like

446
00:22:50,169 --> 00:22:54,770
S1: this sliding window of chaos. Just scrolling through your brain

447
00:22:54,770 --> 00:22:56,929
S1: and it is literally the stuff I was just talking about.

448
00:22:56,969 --> 00:22:59,409
S1: Like you go into this, this woman over here, this

449
00:22:59,409 --> 00:23:02,250
S1: guy over here, this kid over here, the janitor over here.

450
00:23:02,250 --> 00:23:05,929
S1: It's just this rumination, this grinding of like, oh yeah,

451
00:23:05,969 --> 00:23:07,649
S1: I should, I'm going to watch that thing. I've got

452
00:23:07,689 --> 00:23:09,490
S1: to pick up those things. I don't know why he

453
00:23:09,490 --> 00:23:12,050
S1: said that. Yeah. You know, I should get a promotion

454
00:23:12,090 --> 00:23:14,210
S1: like these people don't respect me. Blah blah blah. It's

455
00:23:14,209 --> 00:23:18,090
S1: just this constant like droning. Where did those thoughts come from?

456
00:23:18,090 --> 00:23:21,489
S1: Who is making those thoughts and who is observing the thoughts?

457
00:23:21,490 --> 00:23:23,889
S1: The answer is you have no idea. I have no

458
00:23:23,889 --> 00:23:26,329
S1: idea what my thoughts are. I have no idea how

459
00:23:26,330 --> 00:23:28,290
S1: I'm going to finish this sentence. We'll talk about that

460
00:23:28,290 --> 00:23:31,290
S1: one in a second. Nobody has any idea where their

461
00:23:31,290 --> 00:23:33,530
S1: thoughts are coming from. Do you know what that reminds

462
00:23:33,530 --> 00:23:36,770
S1: me of? That reminds me of an LLM. When you

463
00:23:36,770 --> 00:23:39,689
S1: poke an LLM and you say, hey, say smart stuff

464
00:23:39,689 --> 00:23:44,170
S1: about logistic workflows and pipelines and you know how to

465
00:23:44,169 --> 00:23:47,119
S1: move heavy freight from the East Coast to the West

466
00:23:47,119 --> 00:23:51,039
S1: Coast on an 18 Wheeler. It just goes okay, and

467
00:23:51,320 --> 00:23:54,680
S1: it starts spewing, starts spewing words, right? And it turns

468
00:23:54,679 --> 00:23:57,600
S1: out they're really good and they're awesome. And the words

469
00:23:57,600 --> 00:24:00,600
S1: make sense. And it's actually factual unless it's not right.

470
00:24:00,639 --> 00:24:03,199
S1: And you could do deep research or whatever and correct that.

471
00:24:03,199 --> 00:24:05,640
S1: The point is, you could ping a human to do

472
00:24:05,639 --> 00:24:08,080
S1: that and they start spewing words, or you could ping

473
00:24:08,119 --> 00:24:10,800
S1: an LLM and start spewing words. Does the LLM have

474
00:24:10,800 --> 00:24:13,320
S1: any idea where its thoughts are coming from? No. Do

475
00:24:13,320 --> 00:24:16,159
S1: you have any idea where your thoughts are coming from? No.

476
00:24:16,159 --> 00:24:19,800
S1: It's a complete black box. Complete black box. We have

477
00:24:19,800 --> 00:24:23,480
S1: no idea where our thoughts are coming from, where our

478
00:24:23,480 --> 00:24:27,919
S1: dreams are coming from, where our concerns, our feelings, anything.

479
00:24:27,919 --> 00:24:31,119
S1: It's bubbling out of us from a black hole. Like

480
00:24:31,159 --> 00:24:33,480
S1: actually think about this. If you are in the middle

481
00:24:33,480 --> 00:24:36,999
S1: of a sentence like I have been multiple times here already,

482
00:24:37,000 --> 00:24:40,359
S1: I have no idea how I'm going to finish the sentence.

483
00:24:40,359 --> 00:24:42,840
S1: I'm not planning a sentence. First of all, I didn't

484
00:24:42,840 --> 00:24:45,139
S1: come up with the sentence at all. I didn't plan

485
00:24:45,139 --> 00:24:47,419
S1: that sentence. I just said, these are spewing out of

486
00:24:47,419 --> 00:24:50,260
S1: me just like an LLM. I have no idea how

487
00:24:50,260 --> 00:24:52,619
S1: the sentence is going to end. I am just as

488
00:24:52,619 --> 00:24:56,139
S1: surprised about anything I say as you are. Who is

489
00:24:56,139 --> 00:24:59,499
S1: receiving it right now? Guess what that sounds like also AI.

490
00:24:59,619 --> 00:25:02,499
S1: So going back to the work thing and now this,

491
00:25:02,500 --> 00:25:06,340
S1: there's this view that we are put together. We are cohesive.

492
00:25:06,540 --> 00:25:10,659
S1: We are self understanding. It's complete and utter crap. We

493
00:25:10,659 --> 00:25:12,859
S1: have no idea where our thoughts are coming from. We

494
00:25:12,859 --> 00:25:15,900
S1: have no idea how we're going to finish sentences. The

495
00:25:15,899 --> 00:25:19,899
S1: whole thing is like this random generator. So here's another example.

496
00:25:19,899 --> 00:25:22,019
S1: I'm going to ask you a question right now. What

497
00:25:22,020 --> 00:25:26,260
S1: are your top seven favorite restaurants in the entire world?

498
00:25:26,260 --> 00:25:29,899
S1: You've got 30s to answer. What is your brain doing

499
00:25:29,899 --> 00:25:33,379
S1: right now? You're thinking about these. Who are you receiving

500
00:25:33,379 --> 00:25:36,739
S1: content from? Where is that content coming from? Now you

501
00:25:36,740 --> 00:25:41,020
S1: might say, well, I have memories and those memories are me. Yeah, sure.

502
00:25:41,060 --> 00:25:45,010
S1: AI also has memories. AI has weights and AI has

503
00:25:45,010 --> 00:25:48,810
S1: context and it goes and retrieves those and brings back stuff,

504
00:25:48,810 --> 00:25:51,450
S1: but it brings it back in a different way each time. Right.

505
00:25:51,490 --> 00:25:54,770
S1: When it spews an answer to top seven restaurants, it's

506
00:25:54,770 --> 00:25:58,929
S1: reviewing content and spewing it in a way that is unpredictable. Right.

507
00:25:58,969 --> 00:26:02,010
S1: It's random, hallucinated. Some might say, well, guess what? When

508
00:26:02,010 --> 00:26:04,649
S1: I come up with my list of seven items, I

509
00:26:04,649 --> 00:26:06,409
S1: have no idea what they're going to be. If you

510
00:26:06,409 --> 00:26:09,409
S1: ask me multiple days or weeks apart or whatever, and

511
00:26:09,409 --> 00:26:12,369
S1: I'm not like prepared, the list will be different because

512
00:26:12,369 --> 00:26:15,449
S1: some I couldn't remember at the time. Actually, some my

513
00:26:15,449 --> 00:26:19,850
S1: preference might change in between. This is not a static, cohesive,

514
00:26:19,969 --> 00:26:23,569
S1: dependable process. Okay. And that takes us right into the

515
00:26:23,570 --> 00:26:27,290
S1: memory one. Okay. One of the most reproduced things in

516
00:26:27,490 --> 00:26:30,530
S1: psychology is the fact that memories are not what we

517
00:26:30,530 --> 00:26:33,209
S1: think they are in the human mind. When we store memories,

518
00:26:33,209 --> 00:26:36,850
S1: we're storing it with like emotion and bias and sort

519
00:26:36,850 --> 00:26:40,800
S1: of feelings. And when we recall things, it's extremely well

520
00:26:40,840 --> 00:26:42,959
S1: documented all through science, and I'm going to have a

521
00:26:42,959 --> 00:26:45,840
S1: whole bunch of science links in the blog that accompanies this.

522
00:26:45,840 --> 00:26:49,080
S1: So you can actually review this literature, but you actually

523
00:26:49,080 --> 00:26:51,960
S1: don't store. You're not a hard drive, you're not storing

524
00:26:52,000 --> 00:26:55,080
S1: unaltered ones and zeros. Okay. First of all, we don't

525
00:26:55,080 --> 00:26:57,680
S1: know how we're storing things in the brain. That's step one.

526
00:26:57,679 --> 00:27:01,680
S1: But we do know that when we recall events versus

527
00:27:01,679 --> 00:27:05,000
S1: what happened, they're massively different and it gets worse over

528
00:27:05,000 --> 00:27:07,399
S1: time and it gets worse the more you recall it.

529
00:27:07,399 --> 00:27:11,879
S1: So our memory system, even our memory system is completely flawed.

530
00:27:11,879 --> 00:27:13,840
S1: So we don't know where our thoughts are coming from.

531
00:27:13,840 --> 00:27:16,080
S1: We don't know how we're going to finish sentences. And

532
00:27:16,080 --> 00:27:19,479
S1: our memories are not static. They're not ones and zeros.

533
00:27:19,520 --> 00:27:23,080
S1: They're actually extremely fluid. And then there's the weird fact

534
00:27:23,080 --> 00:27:24,800
S1: that when you go to sleep and wake up, you're

535
00:27:24,840 --> 00:27:28,639
S1: actually like a different person. Your entire being, your entire identity,

536
00:27:28,679 --> 00:27:31,960
S1: like went offline and came back up, like spun up

537
00:27:31,959 --> 00:27:35,520
S1: with like mostly the same memories, like mostly the same feelings,

538
00:27:35,520 --> 00:27:38,159
S1: but you're kind of like a different person. So this

539
00:27:38,159 --> 00:27:42,310
S1: whole concept of companies are doing great work. They have

540
00:27:42,310 --> 00:27:46,709
S1: great workflows. People follow the processes. That's garbage. Then you

541
00:27:46,709 --> 00:27:49,149
S1: have the concept of humans are like these, you know,

542
00:27:49,189 --> 00:27:53,509
S1: bastions of like perfection. And we are the perfect thinking machines.

543
00:27:53,510 --> 00:27:56,670
S1: And AI can't possibly hold a candle to us. Also

544
00:27:56,669 --> 00:28:00,109
S1: highly flawed. The whole thing is built on just absolute madness.

545
00:28:00,109 --> 00:28:05,149
S1: So once again, as thinking machines, the bar is extremely low. Okay,

546
00:28:05,189 --> 00:28:06,790
S1: I want to show you this. I came up with

547
00:28:06,790 --> 00:28:09,990
S1: this capability stack, whatever. It's just an image. And by

548
00:28:09,990 --> 00:28:13,189
S1: the way, this is not like settled science. Like this

549
00:28:13,189 --> 00:28:15,310
S1: is all still up in the air. People are still

550
00:28:15,310 --> 00:28:17,270
S1: trying to figure all this out. I read a lot.

551
00:28:17,310 --> 00:28:20,350
S1: I've got pretty decent mental models. Just work with me

552
00:28:20,350 --> 00:28:23,990
S1: on this. Think of these blue pill boxes here, these

553
00:28:23,990 --> 00:28:28,710
S1: blue layers here as fundamentals of a knowledge worker. Okay?

554
00:28:28,909 --> 00:28:31,429
S1: Things that make us good at doing knowledge work. So

555
00:28:31,429 --> 00:28:36,749
S1: first of all, it's knowledge. So collecting, organizing and accessing facts, right?

556
00:28:36,790 --> 00:28:43,100
S1: Knowledge Facts. Easy understanding. Turning knowledge into concepts. Okay, so

557
00:28:43,140 --> 00:28:45,979
S1: it's the ability to actually like, let it seep in

558
00:28:45,980 --> 00:28:48,259
S1: and actually be able to say, okay, oh, that knowledge

559
00:28:48,260 --> 00:28:51,820
S1: is related to that knowledge. Oh, okay. That forms a concept, right?

560
00:28:51,860 --> 00:28:54,500
S1: So now we have concepts. So that's, that's what I'm

561
00:28:54,500 --> 00:28:58,100
S1: calling understanding here. Again, we could use lots of different definitions.

562
00:28:58,100 --> 00:29:02,019
S1: Just work with these knowledge understanding. Now on top of that,

563
00:29:02,020 --> 00:29:04,340
S1: in addition to having a whole bunch of facts stored,

564
00:29:04,340 --> 00:29:07,340
S1: that's not enough to be a good knowledge worker. Understanding. Okay,

565
00:29:07,380 --> 00:29:09,739
S1: now we're getting closer, right? You have to have understanding

566
00:29:09,740 --> 00:29:12,500
S1: of all these different concepts and everything and how they relate.

567
00:29:12,540 --> 00:29:15,940
S1: Intelligence is the next one above. This is the ability

568
00:29:15,940 --> 00:29:20,219
S1: to face obstacles, right? You're navigating obstacles. You're navigating the

569
00:29:20,219 --> 00:29:24,180
S1: world around you, and you have goals that you are pursuing. Okay,

570
00:29:24,219 --> 00:29:27,499
S1: so you're pursuing goals. So basically, if you're more intelligent,

571
00:29:27,500 --> 00:29:30,820
S1: you're able to navigate the world better and better achieve

572
00:29:30,820 --> 00:29:33,740
S1: your goals. So that's the definition of intelligence that we're

573
00:29:33,740 --> 00:29:37,930
S1: using here. And the final one is creativity. Now this one.

574
00:29:37,930 --> 00:29:40,769
S1: I would even argue not to get ahead of myself here,

575
00:29:40,770 --> 00:29:43,730
S1: but I would argue that this is truly something that

576
00:29:43,730 --> 00:29:47,050
S1: humans have that AI doesn't have right now. And who

577
00:29:47,050 --> 00:29:49,050
S1: knows when they're going to get it or how much

578
00:29:49,050 --> 00:29:51,690
S1: they're going to get. But even the first three are

579
00:29:51,690 --> 00:29:55,770
S1: really solid and most knowledge work, okay? Most knowledge work

580
00:29:55,770 --> 00:29:59,290
S1: that people are doing is actually the knowledge, the understanding

581
00:29:59,290 --> 00:30:03,530
S1: and the intelligence creativity. You can have some people have

582
00:30:03,650 --> 00:30:05,690
S1: some people have a lot of creativity. Some people have

583
00:30:05,690 --> 00:30:07,930
S1: a little, some people have none, but they're still really

584
00:30:07,930 --> 00:30:10,490
S1: good knowledge workers. I would argue you could even have

585
00:30:10,490 --> 00:30:14,650
S1: an A player who crushes knowledge, understanding and intelligence and

586
00:30:14,730 --> 00:30:16,890
S1: is still really great at their job. Then you have

587
00:30:16,890 --> 00:30:21,410
S1: creativity on top of that. So this is like intuition art,

588
00:30:21,450 --> 00:30:25,810
S1: you know, liberal arts thinking, philosophy, you know, logic, like

589
00:30:25,850 --> 00:30:28,370
S1: all the different things, the ability to come up with

590
00:30:28,370 --> 00:30:32,130
S1: novel things, net new things. That's ultimately what we're talking

591
00:30:32,130 --> 00:30:35,510
S1: about here. So taking just the left side, just the

592
00:30:35,510 --> 00:30:39,749
S1: blue layers. One of the arguments for why humans cannot

593
00:30:39,750 --> 00:30:42,670
S1: be replaced is that they have expertise. This is a

594
00:30:42,670 --> 00:30:45,350
S1: thing that my friend Marcus Hutchins is always talking about.

595
00:30:45,390 --> 00:30:48,510
S1: He's like, hey, look, there's like, expertise is the thing

596
00:30:48,510 --> 00:30:51,150
S1: that AI can't do. Like it can have facts and

597
00:30:51,150 --> 00:30:54,950
S1: it can regurgitate them. But like, that's not real useful.

598
00:30:54,950 --> 00:30:57,470
S1: That's not what work is. That's ultimately what we're trying

599
00:30:57,469 --> 00:31:00,030
S1: to answer here. What is work, right? And how does

600
00:31:00,030 --> 00:31:04,190
S1: it relate to human capabilities versus AI capabilities? So where

601
00:31:04,190 --> 00:31:07,350
S1: is expertise on this list? And maybe you think I

602
00:31:07,350 --> 00:31:09,510
S1: missed a layer. I don't think I did, but I'm

603
00:31:09,510 --> 00:31:13,070
S1: challenging you to tell me what I missed. Where is

604
00:31:13,070 --> 00:31:16,870
S1: the expertise layer here if it's not one of these four?

605
00:31:16,990 --> 00:31:21,550
S1: So my answer to that question is expertise is knowledge,

606
00:31:21,590 --> 00:31:25,870
S1: understanding and intelligence combined. Okay, so let's just go through

607
00:31:25,870 --> 00:31:29,390
S1: some different work tasks. Okay? You've got engineering, you've got

608
00:31:29,430 --> 00:31:33,820
S1: like civil engineering, you've got software developers, you've got site

609
00:31:33,860 --> 00:31:38,980
S1: reliability engineer. You've got product manager, you've got project manager,

610
00:31:38,980 --> 00:31:42,299
S1: you've got. Head of finance. You've got what do we

611
00:31:42,340 --> 00:31:45,580
S1: got lots of other jobs right. Million different jobs. Admin.

612
00:31:45,620 --> 00:31:49,500
S1: You're an admin for the company. You you handle you know, meals.

613
00:31:49,500 --> 00:31:52,339
S1: You handle emails, you handle events. You handle all that

614
00:31:52,340 --> 00:31:54,900
S1: kind of stuff. Maybe there's marketing as well. If I

615
00:31:54,900 --> 00:31:57,779
S1: didn't already mention that, tons of different roles, thousands of

616
00:31:57,780 --> 00:32:00,900
S1: different roles. Here's the question. When somebody is good at

617
00:32:00,940 --> 00:32:03,740
S1: that thing, which of these are they good at? Which

618
00:32:03,780 --> 00:32:06,579
S1: of these do they have in combination and which of

619
00:32:06,580 --> 00:32:10,700
S1: these are required? Obviously, some require more creativity than others.

620
00:32:10,700 --> 00:32:13,820
S1: Some don't require any creativity at all. I would argue

621
00:32:13,860 --> 00:32:17,020
S1: they pretty much all require intelligence. So this is really

622
00:32:17,020 --> 00:32:22,739
S1: interesting actually. Here's the question. Why hasn't existing automation previous automation,

623
00:32:22,740 --> 00:32:26,180
S1: why has it not killed off knowledge workers for all

624
00:32:26,180 --> 00:32:30,219
S1: these decades? We've had automation for decades. We've had automation

625
00:32:30,219 --> 00:32:33,930
S1: for hundreds of years, and we've had computers for decades

626
00:32:33,930 --> 00:32:38,730
S1: and decades, and we've had automation scripts, you know, programming.

627
00:32:38,730 --> 00:32:41,569
S1: We've had these things for decades as well with really,

628
00:32:41,570 --> 00:32:46,330
S1: really advanced automations, excel, all these different things. Why can't

629
00:32:46,370 --> 00:32:50,290
S1: they replace a knowledge worker? Why have they not replaced

630
00:32:50,290 --> 00:32:55,050
S1: a knowledge worker? My answer to that is understanding and intelligence. Okay.

631
00:32:55,090 --> 00:32:59,729
S1: Previous automation basically just churned it executed it followed a

632
00:32:59,770 --> 00:33:03,650
S1: list of steps deterministically, right? What it can't do is

633
00:33:03,650 --> 00:33:06,890
S1: the thing that everyone must do as a knowledge worker

634
00:33:06,890 --> 00:33:09,930
S1: inside of the chaos that I described earlier. Remember I

635
00:33:09,930 --> 00:33:13,210
S1: was describing how everything is chaos. Everything's changing all the time. Oh,

636
00:33:13,250 --> 00:33:15,290
S1: the goals are different. This is different when you come

637
00:33:15,290 --> 00:33:17,890
S1: into work on an average day and an average company

638
00:33:17,890 --> 00:33:19,850
S1: as a knowledge worker, you don't know what you're going

639
00:33:19,890 --> 00:33:22,690
S1: to get. They might be like, hey, we're moving offices. Hey,

640
00:33:22,730 --> 00:33:25,850
S1: our remit is completely different. All of our jobs are different.

641
00:33:25,850 --> 00:33:28,450
S1: We now report to a different person. They have different goals.

642
00:33:28,450 --> 00:33:31,320
S1: Maybe the goals are the same, but they like everything

643
00:33:31,360 --> 00:33:35,079
S1: to be done differently. They like blue templates. So you

644
00:33:35,080 --> 00:33:37,800
S1: need to put all your stuff in brand new templates. Well,

645
00:33:37,800 --> 00:33:41,279
S1: if it's base automation, okay, it didn't learn how to

646
00:33:41,280 --> 00:33:44,400
S1: use blue templates. It didn't learn how to have new goals.

647
00:33:44,400 --> 00:33:47,440
S1: All that stuff needs to be rewritten. Intelligence is the

648
00:33:47,440 --> 00:33:52,080
S1: ability to adapt, to change, to navigate, to work around

649
00:33:52,080 --> 00:33:55,040
S1: and basically say, okay, well, here are my actual goals,

650
00:33:55,040 --> 00:33:57,000
S1: here's what I know and I'm supposed to do. Here's

651
00:33:57,000 --> 00:34:00,160
S1: my knowledge, here's my understanding of what is meant by

652
00:34:00,160 --> 00:34:02,520
S1: all of this. So here are the changes I'm going

653
00:34:02,560 --> 00:34:04,959
S1: to make to the way I do my work based

654
00:34:04,959 --> 00:34:09,840
S1: on these new requirements that intelligence piece previous automation didn't have. Okay,

655
00:34:09,880 --> 00:34:12,600
S1: now you add creativity. On top of that, the ability

656
00:34:12,600 --> 00:34:15,799
S1: to come up with new ideas and innovate and stuff

657
00:34:15,800 --> 00:34:19,480
S1: like that, that's way beyond automation. And I would say

658
00:34:19,680 --> 00:34:24,280
S1: arguably today, AI, it's also way beyond today's AI. Now,

659
00:34:24,280 --> 00:34:26,919
S1: there are some edge cases where it's starting to seep in,

660
00:34:26,920 --> 00:34:30,310
S1: but let's just set those aside. So this is what

661
00:34:30,310 --> 00:34:34,670
S1: makes a human valuable as a knowledge worker. Expertise is

662
00:34:34,670 --> 00:34:37,670
S1: all for. But I would say mostly it's knowledge, understanding

663
00:34:37,670 --> 00:34:40,749
S1: and intelligence. So give me an example. Try to think

664
00:34:40,750 --> 00:34:43,230
S1: of an example. And this is sort of me arguing

665
00:34:43,230 --> 00:34:46,430
S1: with Marcus here. Give me an example of expertise that

666
00:34:46,430 --> 00:34:50,750
S1: isn't this. Think about what expertise is. It's lots of experience.

667
00:34:50,750 --> 00:34:53,910
S1: It is the knowledge, right? It's the understanding of the concepts.

668
00:34:53,910 --> 00:34:57,190
S1: It's the ability to adapt with the intelligence and navigate

669
00:34:57,190 --> 00:35:00,830
S1: and change and everything. The combination of those combined with

670
00:35:00,830 --> 00:35:05,910
S1: experience is ultimately what gives you your expertise right now.

671
00:35:05,950 --> 00:35:10,310
S1: Maybe you would argue, well, what about experience? Maybe experience

672
00:35:10,310 --> 00:35:13,510
S1: should be on the list, but experience is memory, right?

673
00:35:13,549 --> 00:35:19,189
S1: Experience is recall experiences, accrued iterations on seeing different problems.

674
00:35:19,190 --> 00:35:22,390
S1: And ultimately this is why I have the model. This way.

675
00:35:22,390 --> 00:35:26,469
S1: It is actually improving your knowledge or your understanding. Okay,

676
00:35:26,509 --> 00:35:29,450
S1: your intelligence is not really improving. That's just a thing

677
00:35:29,450 --> 00:35:31,290
S1: that you have. It's a tool that you can use,

678
00:35:31,290 --> 00:35:35,690
S1: but your knowledge improves with the experience. An experienced person

679
00:35:35,690 --> 00:35:39,689
S1: has more knowledge. An experienced person has more understanding. They

680
00:35:39,690 --> 00:35:43,969
S1: have better understanding, okay? Which means they have linked more

681
00:35:43,969 --> 00:35:47,209
S1: and more facts and more and more situations and more

682
00:35:47,210 --> 00:35:51,170
S1: and more patterns to interlock, to connect to each other

683
00:35:51,170 --> 00:35:53,129
S1: so they could just see patterns better. And this is

684
00:35:53,130 --> 00:35:57,569
S1: related to intelligence as well. But somebody with 25 years experience,

685
00:35:57,569 --> 00:36:02,930
S1: 30 years experience, 50 years experience of doing financial audits,

686
00:36:02,930 --> 00:36:06,329
S1: they just immediately see things. They see patterns. Okay, here's

687
00:36:06,330 --> 00:36:09,450
S1: the question. What part of that does AI not have?

688
00:36:09,489 --> 00:36:13,569
S1: First of all, how is AI a knowledge? Okay, insane. Okay.

689
00:36:13,610 --> 00:36:16,249
S1: It's read every book. It's a deep expert. I don't

690
00:36:16,250 --> 00:36:20,010
S1: want to lead the witness here. It's deeply knowledgeable about

691
00:36:20,009 --> 00:36:23,770
S1: many fields, tens of thousands of fields and subdomains. It

692
00:36:23,770 --> 00:36:27,760
S1: can rattle off all the details, explain everything to you

693
00:36:27,799 --> 00:36:30,039
S1: whatever way you want to receive it. It could write

694
00:36:30,080 --> 00:36:32,960
S1: a full book on the topic by itself. It could

695
00:36:32,960 --> 00:36:36,919
S1: make flashcards for you about this very specific thing that

696
00:36:36,920 --> 00:36:39,519
S1: is like very few people on earth actually know about,

697
00:36:39,520 --> 00:36:41,960
S1: but the AI knows about it. It's also read most

698
00:36:41,960 --> 00:36:48,160
S1: every book effectively. Okay, so knowledge, forget about it. Okay. Understanding. Yes. Okay.

699
00:36:48,319 --> 00:36:51,759
S1: You could argue AI's don't have understanding. There is a

700
00:36:51,759 --> 00:36:55,840
S1: type of understanding that AI's don't have and that is experience. Okay.

701
00:36:55,880 --> 00:37:00,839
S1: If you argue that understanding requires subjective experience, well, then

702
00:37:00,839 --> 00:37:03,560
S1: they don't have it right because they don't experience anything.

703
00:37:03,600 --> 00:37:09,800
S1: My definition of understanding is connecting concepts, finding patterns across domains,

704
00:37:09,799 --> 00:37:14,239
S1: and AI is unbelievably good at this. It is mind

705
00:37:14,279 --> 00:37:19,560
S1: blowingly good at connecting things, cross linking patterns, identifying a

706
00:37:19,560 --> 00:37:22,840
S1: pattern in a piece of content. Oh, this relates to

707
00:37:22,880 --> 00:37:26,710
S1: the 1930s Russia in the following way. Oh well, how

708
00:37:26,710 --> 00:37:31,669
S1: does that relate to chicken feed farming practices from Idaho?

709
00:37:31,710 --> 00:37:34,430
S1: It's like, oh, well good question. Here's how it relates.

710
00:37:34,430 --> 00:37:37,390
S1: It can cross-reference anything. It holds it all in its

711
00:37:37,390 --> 00:37:40,029
S1: mind all at once and can reference this. It can

712
00:37:40,029 --> 00:37:43,470
S1: also do that supplemented by tools to go and research

713
00:37:43,469 --> 00:37:45,989
S1: anything and add to it. It could also do that

714
00:37:45,989 --> 00:37:49,310
S1: by looking at Rag or at a giant composite list

715
00:37:49,310 --> 00:37:52,149
S1: of additional knowledge that you give it. Right? So it

716
00:37:52,150 --> 00:37:54,629
S1: doesn't have to actually be in the model weights. It

717
00:37:54,630 --> 00:37:58,270
S1: could actually be part of a unified system, which includes

718
00:37:58,270 --> 00:38:01,430
S1: going to get new information and cross-referencing to make sure

719
00:38:01,430 --> 00:38:06,270
S1: it's real. Right. So knowledge and understanding, crushing it, absolutely

720
00:38:06,270 --> 00:38:09,989
S1: crushing it. Let's go to intelligence. Can AI navigate the

721
00:38:09,989 --> 00:38:15,350
S1: world and achieve goals? Yes, absolutely. And this is fairly modern, right?

722
00:38:15,390 --> 00:38:17,149
S1: It's been going on for a couple of years. But

723
00:38:17,150 --> 00:38:20,190
S1: the last year, the last six months. Are you kidding me?

724
00:38:20,190 --> 00:38:23,700
S1: An agent infrastructure that has access to all these tools

725
00:38:23,700 --> 00:38:27,580
S1: and all your company's knowledge, all this context and a really,

726
00:38:27,580 --> 00:38:30,900
S1: really smart model and a system of agents that actually

727
00:38:30,900 --> 00:38:33,700
S1: can use that context. It actually has a list of

728
00:38:33,700 --> 00:38:36,859
S1: rules to follow. And okay, so it's proceeding down a

729
00:38:36,860 --> 00:38:40,219
S1: track and you're like, okay, well, now imagine that the

730
00:38:40,219 --> 00:38:43,100
S1: customer says, we actually can't do the deal in Japan.

731
00:38:43,100 --> 00:38:46,460
S1: We're actually need to do it in the UK. And actually,

732
00:38:46,500 --> 00:38:49,180
S1: here's another restraint. So the size of the project has

733
00:38:49,180 --> 00:38:51,579
S1: to be larger than this. By the way, you need

734
00:38:51,580 --> 00:38:56,300
S1: to employ only people who have residency in Singapore or

735
00:38:56,299 --> 00:38:59,980
S1: the UK. Also, we need some people who have understanding

736
00:39:00,020 --> 00:39:03,580
S1: of Scotland and also everyone should be over the age

737
00:39:03,580 --> 00:39:07,580
S1: of 35, right? And also make sure the report format

738
00:39:07,580 --> 00:39:10,660
S1: has to follow this particular template. And also we have

739
00:39:10,660 --> 00:39:14,299
S1: to take into account these other 90 things which I'm

740
00:39:14,299 --> 00:39:17,259
S1: about to give you. Those are new requirements. That is

741
00:39:17,259 --> 00:39:21,980
S1: dynamic changing requirements on the ground. Here's another requirement. Also

742
00:39:22,049 --> 00:39:25,809
S1: go and research all things related to this. Find all

743
00:39:25,810 --> 00:39:29,610
S1: scientific documents related to this and do thorough research for

744
00:39:29,610 --> 00:39:32,169
S1: a day and a half using all your resources. Bring

745
00:39:32,170 --> 00:39:35,129
S1: in all the sources to support the document that you're

746
00:39:35,170 --> 00:39:38,089
S1: actually going to produce for this customer. That's going to

747
00:39:38,089 --> 00:39:44,010
S1: cost $300,000. Those are dynamic, changing situations. Everyone who's used

748
00:39:44,049 --> 00:39:48,569
S1: AI recently, who is a decent user, knows that you

749
00:39:48,569 --> 00:39:51,410
S1: could throw anything at these eyes. You could throw anything

750
00:39:51,450 --> 00:39:53,370
S1: at them. And it's like, oh, well, in that case,

751
00:39:53,370 --> 00:39:55,770
S1: you should do this. And you're like, yeah, but here's

752
00:39:55,770 --> 00:39:58,890
S1: another requirement. Oh, well, in that case, we should adjust.

753
00:39:58,890 --> 00:40:02,890
S1: So can it do intelligence? Do modern AIS have intelligence

754
00:40:02,890 --> 00:40:09,129
S1: according to this definition, 100% expertise. Like I said, expertise, knowledge,

755
00:40:09,130 --> 00:40:14,209
S1: understanding and intelligence, mostly knowledge and understanding. Okay. Intelligence, I

756
00:40:14,210 --> 00:40:17,930
S1: would say definitely. Sure. Let's include that creativity. Again, I'm

757
00:40:17,969 --> 00:40:20,609
S1: not sure you need it for expertise. You can have

758
00:40:20,719 --> 00:40:23,799
S1: somebody who is just the craziest bookworm and is just

759
00:40:23,799 --> 00:40:26,799
S1: like locked in. You give them this task. They just

760
00:40:26,799 --> 00:40:28,719
S1: crush it. You give them a different task. They just

761
00:40:28,759 --> 00:40:32,200
S1: absolutely crush it. Are they inventing anything new? Not really.

762
00:40:32,200 --> 00:40:34,440
S1: Are they coming up with a new way to do it? Nope.

763
00:40:34,480 --> 00:40:36,479
S1: The way that they were trained or the way they

764
00:40:36,480 --> 00:40:38,600
S1: decided they wanted to do it, that's the way they

765
00:40:38,600 --> 00:40:42,239
S1: execute every single time doesn't require creativity. Okay, so now

766
00:40:42,239 --> 00:40:46,640
S1: let's slide over here. Let's look at the AI one intelligence, understanding,

767
00:40:46,640 --> 00:40:49,880
S1: knowledge creativity. Let's just take it off the list. Let's

768
00:40:49,880 --> 00:40:53,719
S1: just pretend it doesn't have it. AI has the expertise.

769
00:40:53,839 --> 00:40:56,880
S1: So this argument that a bunch of people are making

770
00:40:56,880 --> 00:41:01,120
S1: very smart people, very knowledgeable, very technical, they're saying that

771
00:41:01,120 --> 00:41:06,480
S1: you can't possibly replicate human expertise, human thinking, the quality

772
00:41:06,480 --> 00:41:09,680
S1: of human thought, the quality of human thinking, the quality

773
00:41:09,680 --> 00:41:13,120
S1: of human execution of work, everything that we've been talking

774
00:41:13,120 --> 00:41:16,400
S1: about tells us that is just not true. In fact,

775
00:41:16,400 --> 00:41:20,180
S1: the polar opposite is actually more true that even decent

776
00:41:20,219 --> 00:41:24,459
S1: AI today, given these types of tasks, will just crush

777
00:41:24,500 --> 00:41:28,580
S1: an average knowledge worker. Just absolutely crush them on tasks

778
00:41:28,580 --> 00:41:31,979
S1: that are there defined. Okay, what are we doing all day?

779
00:41:32,020 --> 00:41:36,220
S1: We're accepting emails. We're responding to emails, we're producing reports,

780
00:41:36,219 --> 00:41:41,140
S1: we're writing code. These basic commodity tasks make up. I'm

781
00:41:41,140 --> 00:41:44,700
S1: making up a number 95% of all knowledge work. It's

782
00:41:44,700 --> 00:41:49,300
S1: common stuff. The only thing that stops somebody from walking

783
00:41:49,299 --> 00:41:52,580
S1: in off the street who has the requisite, like education

784
00:41:52,580 --> 00:41:55,380
S1: or training in this field, which again is just knowledge

785
00:41:55,380 --> 00:41:59,219
S1: and understanding, mostly knowledge. Someone walking in brand new to

786
00:41:59,259 --> 00:42:01,060
S1: do that in the company, they just don't have the

787
00:42:01,060 --> 00:42:04,700
S1: company knowledge. Well guess what? They accrue it over time.

788
00:42:04,700 --> 00:42:07,419
S1: They go through onboarding, they read the documentation, they do

789
00:42:07,420 --> 00:42:09,219
S1: all that well guess what? AI can do that in

790
00:42:09,219 --> 00:42:12,820
S1: a few seconds. Here's a core point about all of this.

791
00:42:12,860 --> 00:42:16,380
S1: A big reason that people think there's a gap between

792
00:42:16,380 --> 00:42:18,930
S1: what the humans have and what A's have. A big

793
00:42:18,930 --> 00:42:21,810
S1: reason they think there's a gap is because the humans

794
00:42:21,810 --> 00:42:26,009
S1: have not articulated the stuff. It's not written down. Think

795
00:42:26,009 --> 00:42:28,850
S1: about this. The way that AI has gotten so good

796
00:42:28,850 --> 00:42:31,770
S1: in the last year is largely because of cloud code.

797
00:42:31,810 --> 00:42:34,689
S1: What did cloud Code do? Did it have a dramatically

798
00:42:34,690 --> 00:42:38,050
S1: better model than the other models? No. What it did

799
00:42:38,049 --> 00:42:42,249
S1: was scaffolding a harness around the model. What it did

800
00:42:42,250 --> 00:42:47,530
S1: is it built context and context management relations between documents,

801
00:42:47,529 --> 00:42:52,289
S1: allowing it to have knowledge, understanding and intelligence. It linked

802
00:42:52,290 --> 00:42:55,489
S1: together everything that the model needs to be able to

803
00:42:55,529 --> 00:42:59,609
S1: perform to actually do intelligence tasks, to actually do work.

804
00:42:59,650 --> 00:43:01,930
S1: That is what this thing did. That is what cloud

805
00:43:01,930 --> 00:43:05,290
S1: code did. Okay. That stuff being written down. Okay, so

806
00:43:05,330 --> 00:43:08,410
S1: another thing related to this that's extremely hot right now.

807
00:43:08,410 --> 00:43:11,610
S1: It's been hot for a number of months is skills.

808
00:43:11,610 --> 00:43:17,840
S1: What are skills? Skills are a combination of resources, tools, Context. Knowledge.

809
00:43:18,040 --> 00:43:20,200
S1: It's a whole bunch of stuff written down. This is

810
00:43:20,200 --> 00:43:24,279
S1: the center piece. It's essentially the core benefit of a

811
00:43:24,319 --> 00:43:27,160
S1: harness versus a model. A model is just like a

812
00:43:27,160 --> 00:43:31,040
S1: whole bunch of knowledge, right? Once you realize how powerful

813
00:43:31,080 --> 00:43:34,200
S1: these skills are, once you realize how powerful it is

814
00:43:34,200 --> 00:43:36,959
S1: to have all this context in one place, that sort

815
00:43:36,960 --> 00:43:40,080
S1: of wraps an AI model, you start to realize that

816
00:43:40,080 --> 00:43:43,720
S1: the problem is not the AI. The gap between human

817
00:43:43,719 --> 00:43:48,200
S1: expertise and AI expertise is not the model, okay? It's

818
00:43:48,200 --> 00:43:52,000
S1: not the fact that humans have expertise and AI doesn't.

819
00:43:52,040 --> 00:43:56,600
S1: It's the fact that we have not articulated our expertise

820
00:43:56,719 --> 00:44:01,680
S1: down into files, into tools, into contexts. I've been doing

821
00:44:01,680 --> 00:44:04,440
S1: this for over a year now. I've built a system

822
00:44:04,440 --> 00:44:08,000
S1: called Pi. It is a context management system. It is extraordinary.

823
00:44:08,000 --> 00:44:10,080
S1: It's open source. You can check it out. Everyone is

824
00:44:10,080 --> 00:44:12,719
S1: building with skills now. It is becoming the most powerful

825
00:44:12,719 --> 00:44:16,230
S1: concept in all of AI is skills. Why is that?

826
00:44:16,230 --> 00:44:20,990
S1: Because it is literally expertise packaged for this very smart

827
00:44:20,989 --> 00:44:24,629
S1: model to then be able to use the problem inside companies.

828
00:44:24,870 --> 00:44:29,310
S1: That makes people think there's a disconnect between human capability

829
00:44:29,310 --> 00:44:32,390
S1: and AI capability. Nobody has any of this written down.

830
00:44:32,390 --> 00:44:35,430
S1: All the stuff that is in Robbie's mind and Susie's

831
00:44:35,430 --> 00:44:38,749
S1: mind and Kristy's mind and Chris's mind, all the stuff

832
00:44:38,750 --> 00:44:41,710
S1: that's in their minds, it's not written down. Their expertise

833
00:44:41,750 --> 00:44:44,630
S1: is not captured the way this is supposed to be done,

834
00:44:44,630 --> 00:44:47,790
S1: this particular task, this sending of the emails, the creation

835
00:44:47,790 --> 00:44:50,189
S1: of the report, the writing of the code. It's not

836
00:44:50,190 --> 00:44:54,190
S1: written down anywhere. It's passed from human brain to human brain.

837
00:44:54,190 --> 00:44:56,710
S1: And they go on vacation, they go on maternity leave,

838
00:44:56,870 --> 00:45:00,709
S1: they retire, they jump companies every couple of years. That

839
00:45:00,710 --> 00:45:04,230
S1: knowledge just dies. What's starting to happen now is knowledge

840
00:45:04,230 --> 00:45:10,350
S1: is dispersing from brains, from very smart, high expertise, people

841
00:45:10,350 --> 00:45:13,510
S1: into skills. This is why there was a giant stock

842
00:45:13,620 --> 00:45:16,939
S1: drop when anthropic came out with a whole bunch of

843
00:45:16,940 --> 00:45:21,100
S1: legal stuff, right? Those are just processes. Those are expertise

844
00:45:21,100 --> 00:45:25,180
S1: captured in documents, which AI can then follow. So the

845
00:45:25,180 --> 00:45:29,980
S1: expertise gap between humans and AI is actually the failure

846
00:45:30,020 --> 00:45:34,459
S1: so far of us to articulate all the different chaos

847
00:45:34,500 --> 00:45:37,899
S1: things that we talked about in the beginning, all that chaos,

848
00:45:37,940 --> 00:45:41,499
S1: all those random pieces of knowledge inside of people's brains.

849
00:45:41,660 --> 00:45:44,980
S1: Like there's some super old guy who's 62 years old,

850
00:45:45,020 --> 00:45:48,100
S1: you know, his name's Cliff. He's the one you call

851
00:45:48,219 --> 00:45:50,980
S1: when the things go down and there's no docs, you

852
00:45:50,980 --> 00:45:52,540
S1: can't find a thing. You don't know how to get

853
00:45:52,540 --> 00:45:54,700
S1: the server back up. People don't even know what servers

854
00:45:54,700 --> 00:45:57,620
S1: are anymore. You call Cliff. Cliff has never written the

855
00:45:57,620 --> 00:46:00,700
S1: stuff down. Cliff has never created a skill. He's never

856
00:46:00,700 --> 00:46:04,419
S1: done a full debrief interview to capture his knowledge. Now

857
00:46:04,420 --> 00:46:05,660
S1: you're not going to get all of it. It's not

858
00:46:05,660 --> 00:46:08,820
S1: going to be perfect. But if Cliff could document that,

859
00:46:08,819 --> 00:46:10,859
S1: and by the way, lots of people are. Lots of

860
00:46:10,860 --> 00:46:15,200
S1: people are documenting what they know about Cliff's topic and

861
00:46:15,200 --> 00:46:18,360
S1: putting it into skills and their open source. This stuff,

862
00:46:18,360 --> 00:46:21,279
S1: this is P that's seeping into the pool and it's

863
00:46:21,279 --> 00:46:23,279
S1: not going to come out again, but inside of a

864
00:46:23,319 --> 00:46:26,320
S1: given company, if that stuff were to be captured and

865
00:46:26,319 --> 00:46:29,400
S1: brought into this context system like I have with Pi,

866
00:46:29,440 --> 00:46:32,479
S1: like lots of other people have like are captured in

867
00:46:32,480 --> 00:46:36,759
S1: skills that's just absolutely flooding the internet right now. Flooding,

868
00:46:36,799 --> 00:46:40,360
S1: especially the AI space. That is what bridges the gap

869
00:46:40,360 --> 00:46:46,200
S1: between human expertise and AI expertise. And this is happening fast.

870
00:46:46,200 --> 00:46:49,080
S1: It's happening extremely fast. All right. So, so far I've

871
00:46:49,080 --> 00:46:51,640
S1: been talking about the problem. I've been talking about what

872
00:46:51,640 --> 00:46:53,520
S1: I believe to be the problem with the way people

873
00:46:53,520 --> 00:46:57,400
S1: are thinking about this, right? Comparing humans to AI, thinking

874
00:46:57,400 --> 00:47:01,120
S1: that work is like easily done by humans and it's like,

875
00:47:01,120 --> 00:47:03,759
S1: it's working fine. Like what's even the problem now I

876
00:47:03,759 --> 00:47:06,719
S1: want to transition into and we also talked about the

877
00:47:06,719 --> 00:47:09,879
S1: whole expertise thing and like what that really means and

878
00:47:09,880 --> 00:47:11,709
S1: what I think the gap is there. What I want

879
00:47:11,750 --> 00:47:14,629
S1: to pivot to now is essentially talking about what I

880
00:47:14,630 --> 00:47:18,590
S1: think the solution is and what I think is inevitable. Okay,

881
00:47:18,630 --> 00:47:21,110
S1: so this is kind of getting to why I think

882
00:47:21,110 --> 00:47:23,590
S1: this is going to happen and why I think people

883
00:47:23,589 --> 00:47:25,989
S1: need to prepare and get ready. So all of that

884
00:47:25,989 --> 00:47:28,390
S1: was sort of a setup to what I think is

885
00:47:28,390 --> 00:47:31,229
S1: about to happen. And this is an architecture that I've

886
00:47:31,230 --> 00:47:33,830
S1: come up with. I call it the lattice architecture. I've

887
00:47:33,830 --> 00:47:36,230
S1: been working on this for quite a while. I'm going

888
00:47:36,230 --> 00:47:39,190
S1: to release it as an open source project as well.

889
00:47:39,190 --> 00:47:41,870
S1: It's the same system that I use with customers to

890
00:47:41,910 --> 00:47:46,109
S1: basically build their back end systems to help them solve

891
00:47:46,110 --> 00:47:48,350
S1: the stuff that we talked about on the company side.

892
00:47:48,350 --> 00:47:50,910
S1: But the system itself, I'm actually going to release, just

893
00:47:50,910 --> 00:47:53,150
S1: like I did with Pi, I'm going to release this

894
00:47:53,150 --> 00:47:56,870
S1: as kind of a sister project for actually helping companies,

895
00:47:56,910 --> 00:48:01,150
S1: actually any type of entity basically solve that chaos mess.

896
00:48:01,190 --> 00:48:06,710
S1: So essentially, the concept is addressing this transparency problem, addressing

897
00:48:06,710 --> 00:48:10,019
S1: the fact that all these different groups are opaque. The

898
00:48:10,060 --> 00:48:14,460
S1: work is opaque. The processes, the SOPs are not clear.

899
00:48:14,460 --> 00:48:18,100
S1: So what Latisse does is that it actually has a

900
00:48:18,100 --> 00:48:22,019
S1: single unified daemon that everything reports up into. And what

901
00:48:22,020 --> 00:48:24,819
S1: you have are the different tiers of organization. You have

902
00:48:24,819 --> 00:48:28,300
S1: the company level, you have like a unified context of

903
00:48:28,299 --> 00:48:31,700
S1: documents and processes and SOPs and everything, all the different

904
00:48:31,739 --> 00:48:35,060
S1: knowledge for a particular company in a unified place. And

905
00:48:35,060 --> 00:48:38,020
S1: then you have a set of things for each company.

906
00:48:38,020 --> 00:48:41,460
S1: So you have SOPs, you have metrics, you have work,

907
00:48:41,500 --> 00:48:45,060
S1: you have goals, you have budget, you have all these

908
00:48:45,060 --> 00:48:48,779
S1: different components of a particular entity all the way from

909
00:48:48,779 --> 00:48:51,580
S1: the company level. But you also have that same exact

910
00:48:51,580 --> 00:48:55,140
S1: thing for the department. You also have it for a team,

911
00:48:55,140 --> 00:48:58,860
S1: and you also have it for individual people working inside

912
00:48:58,860 --> 00:49:01,779
S1: the company as a worker. So what this does is

913
00:49:01,779 --> 00:49:04,779
S1: it takes a system like pie and you can have

914
00:49:04,779 --> 00:49:07,499
S1: your own version. People will have their own versions. This

915
00:49:07,500 --> 00:49:09,489
S1: is the one that I use. And again, it's open

916
00:49:09,489 --> 00:49:11,850
S1: source project. You can go and check out. And what

917
00:49:11,850 --> 00:49:15,410
S1: it is, is as an individual, you will have your

918
00:49:15,529 --> 00:49:18,530
S1: personal SOPs the way that you like to get things

919
00:49:18,529 --> 00:49:22,370
S1: done that you follow. You will have your personal budget

920
00:49:22,410 --> 00:49:24,689
S1: of what you spend things on. You will have your

921
00:49:24,690 --> 00:49:28,050
S1: own personal metrics, which maybe you got from management or whatever. Actually,

922
00:49:28,049 --> 00:49:29,850
S1: you're about to see where you got it from. You

923
00:49:29,850 --> 00:49:33,729
S1: have all these different components. You have your knowledge in skills,

924
00:49:33,730 --> 00:49:36,969
S1: you have all these capabilities that you have when you

925
00:49:37,009 --> 00:49:39,489
S1: learn something new or whatever, you capture it into the

926
00:49:39,489 --> 00:49:45,450
S1: system and most importantly, your individual unit. You are broadcasting APIs.

927
00:49:45,730 --> 00:49:49,689
S1: You as a person, you, Sarah, you, Ravi, you, Chris,

928
00:49:49,730 --> 00:49:53,129
S1: you are broadcasting what your SOPs are. You are broadcasting

929
00:49:53,130 --> 00:49:56,290
S1: what your metrics are. You are broadcasting what your budget is.

930
00:49:56,290 --> 00:50:00,610
S1: You're broadcasting your work items. They are available for any

931
00:50:00,650 --> 00:50:04,490
S1: agent or any other team member or the agent for

932
00:50:04,489 --> 00:50:07,200
S1: the team or the agent for the department or the

933
00:50:07,200 --> 00:50:10,920
S1: agent for the company can ping these things. Can query

934
00:50:10,960 --> 00:50:14,360
S1: you and actually pull the stuff in real time. Most importantly,

935
00:50:14,360 --> 00:50:17,919
S1: this allows the CEO and the CFO actually to be

936
00:50:17,960 --> 00:50:20,959
S1: able to look down and say, okay, this is the

937
00:50:20,960 --> 00:50:22,919
S1: work that we're doing. Okay, so you're going to have

938
00:50:22,920 --> 00:50:26,919
S1: workflows also documented in this exact same way. I want

939
00:50:26,960 --> 00:50:29,160
S1: to show you some other diagrams here. Okay, so here's

940
00:50:29,160 --> 00:50:33,600
S1: another way of looking at it. So again, goals, plans, SOPs, metrics,

941
00:50:33,600 --> 00:50:36,200
S1: work cost quality. The thing I want to express to

942
00:50:36,200 --> 00:50:38,600
S1: you here, like we talked about already, so you probably

943
00:50:38,600 --> 00:50:40,880
S1: already have this point. So I won't belabor it too much.

944
00:50:40,880 --> 00:50:46,160
S1: These things goals, plans, SOPs, metrics, work, cost quality, these things,

945
00:50:46,160 --> 00:50:50,000
S1: the workflows for the company. This stuff is extremely opaque

946
00:50:50,040 --> 00:50:52,680
S1: to the leaders of a company. This is extremely hard

947
00:50:52,680 --> 00:50:56,400
S1: to gather and this is going to turn it into seconds.

948
00:50:56,400 --> 00:50:59,960
S1: It's going to turn it into less than a second oftentimes, right?

949
00:51:00,000 --> 00:51:03,479
S1: It's going to turn it into minutes at the absolute maximum.

950
00:51:03,480 --> 00:51:05,910
S1: Like if you have a super large company, right? And

951
00:51:05,910 --> 00:51:08,029
S1: it depends on the implementation. Like I want to give

952
00:51:08,029 --> 00:51:10,709
S1: actual numbers or anything. That's just ridiculous. It depends on

953
00:51:10,710 --> 00:51:14,629
S1: the architecture that you use. But the point is transparency.

954
00:51:14,670 --> 00:51:17,110
S1: First of all, and the other core point is that

955
00:51:17,110 --> 00:51:21,789
S1: each layer from the worker to the team to the department,

956
00:51:21,790 --> 00:51:26,470
S1: to the company is independently authoritative over its stuff. So

957
00:51:26,630 --> 00:51:29,390
S1: one of the problems that I see tons with customers

958
00:51:29,509 --> 00:51:33,070
S1: is they have no idea what their other departments are doing, right.

959
00:51:33,110 --> 00:51:36,469
S1: They have no visibility into how they're interacting with the customer,

960
00:51:36,469 --> 00:51:39,070
S1: what they've said to the customer, what's been done with them.

961
00:51:39,150 --> 00:51:42,629
S1: You can also have context silos like this for each

962
00:51:42,630 --> 00:51:46,109
S1: customer inside of a customer container, right? And you can

963
00:51:46,110 --> 00:51:49,029
S1: have other agents because I'm a security person, there's going

964
00:51:49,029 --> 00:51:52,310
S1: to be another agent flow here, basically a gateway that

965
00:51:52,310 --> 00:51:55,870
S1: basically says, okay, which agents can ask which questions about

966
00:51:55,870 --> 00:51:59,390
S1: what context zones, right? So there's going to be permissions.

967
00:51:59,390 --> 00:52:01,270
S1: You've got identity stuff, you've got a whole bunch of

968
00:52:01,270 --> 00:52:03,890
S1: stuff that's all being worked on. But the core concept

969
00:52:03,890 --> 00:52:08,330
S1: of lattice is visibility up and down and across. I

970
00:52:08,330 --> 00:52:10,410
S1: could just query, first of all, I'm just I'm not

971
00:52:10,410 --> 00:52:13,410
S1: going to be actually running any APIs myself. I just

972
00:52:13,410 --> 00:52:16,650
S1: talk to my agent system as an individual worker on

973
00:52:16,650 --> 00:52:18,850
S1: the front end team over here. And I say, hey,

974
00:52:18,850 --> 00:52:21,330
S1: what are my coworkers working on? I'm thinking about working

975
00:52:21,330 --> 00:52:23,770
S1: on this. And they're like, oh, yeah, you know, Julie's

976
00:52:23,770 --> 00:52:26,089
S1: already working on that. You should talk to, you know,

977
00:52:26,130 --> 00:52:28,850
S1: Modi or whatever. Yeah, hit them up. They're already working

978
00:52:28,850 --> 00:52:30,450
S1: on it. Maybe you can do a collab. Do you

979
00:52:30,450 --> 00:52:33,370
S1: want me to set up a meeting? Boom. Okay, cool. Well,

980
00:52:33,410 --> 00:52:35,970
S1: now you just posted that you're in a meeting now

981
00:52:35,969 --> 00:52:38,969
S1: it just posted that you're going to collaborate about this

982
00:52:38,969 --> 00:52:41,930
S1: particular feature that could go into your API. It's now

983
00:52:41,930 --> 00:52:45,289
S1: available when someone asks, hey, is anyone working on this? Well,

984
00:52:45,290 --> 00:52:48,410
S1: now it includes you that you're actually also working on this.

985
00:52:48,450 --> 00:52:52,049
S1: It's now available to other teams, other departments, all the

986
00:52:52,049 --> 00:52:54,769
S1: way up to the company level. And this unified sort

987
00:52:54,770 --> 00:52:58,569
S1: of context layer here, the lattice daemon is where you

988
00:52:58,569 --> 00:53:01,049
S1: can query and get it for everyone. Now I want

989
00:53:01,089 --> 00:53:03,600
S1: to talk about this whole concept of the SOPs and

990
00:53:03,600 --> 00:53:07,000
S1: the goals. Think of what happens right now in a company.

991
00:53:07,000 --> 00:53:08,719
S1: I didn't talk about this too much in the sort

992
00:53:08,719 --> 00:53:12,039
S1: of negative state. When you go to create goals and

993
00:53:12,040 --> 00:53:14,319
S1: change your goals, how many people are triggered by the

994
00:53:14,319 --> 00:53:16,920
S1: word OKR? For a good reason. The reason is because

995
00:53:16,920 --> 00:53:19,439
S1: they change all the time. People have different systems for

996
00:53:19,440 --> 00:53:23,479
S1: doing them. Imagine a giant all hands meeting where the

997
00:53:23,759 --> 00:53:27,040
S1: CEO gets together with everyone and says, hey, check it out.

998
00:53:27,040 --> 00:53:29,320
S1: These are the new goals. This is what we're doing.

999
00:53:29,319 --> 00:53:31,560
S1: Here are our new metrics that go with the goals.

1000
00:53:31,560 --> 00:53:34,400
S1: By the way, we've got some new SOPs we just published. Boom.

1001
00:53:34,440 --> 00:53:36,999
S1: Guess what they say. They have now been published. All

1002
00:53:37,000 --> 00:53:40,600
S1: documentation is now updated, goes all the way down. Updates

1003
00:53:40,600 --> 00:53:44,560
S1: the documentation, updates the goals. Now inside of the goals

1004
00:53:44,560 --> 00:53:47,759
S1: of each group. That content is now updated. We're talking

1005
00:53:47,759 --> 00:53:50,560
S1: about text files here. This is not like magic stuff.

1006
00:53:50,560 --> 00:53:52,839
S1: This is not difficult stuff. We are doing much more

1007
00:53:52,839 --> 00:53:55,799
S1: difficult stuff with AI already. This is just not a

1008
00:53:55,799 --> 00:53:58,279
S1: system that has been built yet and it will take

1009
00:53:58,319 --> 00:54:00,520
S1: time to roll out. Some companies will be able to

1010
00:54:00,520 --> 00:54:03,830
S1: adopt this very quickly. Some companies will take, you know,

1011
00:54:03,870 --> 00:54:05,910
S1: five years or ten years or more, or they'll just

1012
00:54:05,910 --> 00:54:08,150
S1: die off because they couldn't get there fast enough. And

1013
00:54:08,150 --> 00:54:10,669
S1: to be clear, this is one architecture. This is the

1014
00:54:10,670 --> 00:54:15,109
S1: architecture that I'm using. But the concept is actually more important.

1015
00:54:15,150 --> 00:54:19,190
S1: The functions that this architecture allows is the most important piece.

1016
00:54:19,190 --> 00:54:23,229
S1: And I think the broadcasting of the demon, the broadcasting

1017
00:54:23,230 --> 00:54:27,550
S1: of the APIs for each tier level, for each object

1018
00:54:27,589 --> 00:54:31,390
S1: type within the company. That is what makes this all work, right.

1019
00:54:31,430 --> 00:54:34,149
S1: So the protocols that we use, who knows? Like it's

1020
00:54:34,150 --> 00:54:36,269
S1: hard to predict exactly how it's going to play out.

1021
00:54:36,310 --> 00:54:39,029
S1: Maybe it's going to be something better than HTTP. Maybe

1022
00:54:39,029 --> 00:54:41,549
S1: it's going to be something better than rest. It doesn't matter.

1023
00:54:41,549 --> 00:54:44,469
S1: The concept is what I want you to think about here. Okay,

1024
00:54:44,509 --> 00:54:46,870
S1: so here's another view of this. And just another way

1025
00:54:46,870 --> 00:54:48,469
S1: to think about it. I've got a whole bunch of

1026
00:54:48,469 --> 00:54:51,549
S1: these diagrams, by the way, that I put together over

1027
00:54:51,589 --> 00:54:53,589
S1: quite a decent amount of time, and I'm going to

1028
00:54:53,589 --> 00:54:55,749
S1: make them all available there in the blog post. They're

1029
00:54:55,750 --> 00:54:58,310
S1: also in, I think the link is in the description

1030
00:54:58,310 --> 00:55:00,379
S1: to the video as well. But look at this. All

1031
00:55:00,380 --> 00:55:03,540
S1: these different queries can go up to the latest statement.

1032
00:55:03,540 --> 00:55:06,739
S1: You basically post your status. It could be push or pull, right?

1033
00:55:06,779 --> 00:55:09,780
S1: A combination of the both. But again you have the tiers.

1034
00:55:09,779 --> 00:55:13,459
S1: You have goals, budget, plans for the department container, the

1035
00:55:13,460 --> 00:55:16,819
S1: team container, the individual node, right? And they all are

1036
00:55:16,819 --> 00:55:20,380
S1: speaking to one central place, but they're all able to

1037
00:55:20,580 --> 00:55:23,540
S1: post to each other and or receive from each other.

1038
00:55:23,540 --> 00:55:26,499
S1: And the other thing that's part of different diagrams is

1039
00:55:26,500 --> 00:55:30,060
S1: essentially a big unified context as well. Like kind of

1040
00:55:30,100 --> 00:55:32,500
S1: like a context data lake, which I've talked about in

1041
00:55:32,500 --> 00:55:35,379
S1: previous videos. Okay, so here's another view. And this is

1042
00:55:35,380 --> 00:55:38,500
S1: absolutely critical. This is huge. A big part of the

1043
00:55:38,500 --> 00:55:41,500
S1: problem that we have is not being able to see

1044
00:55:41,660 --> 00:55:45,100
S1: the work that we're doing. We don't understand the workflows.

1045
00:55:45,100 --> 00:55:48,100
S1: We don't understand who's doing which part. So I've got

1046
00:55:48,100 --> 00:55:51,100
S1: another video where I talked about companies are just graphs

1047
00:55:51,100 --> 00:55:54,859
S1: of algorithms. Okay? This is a really core concept. The

1048
00:55:54,860 --> 00:55:58,330
S1: idea that someone's doing like insurance claims or whatever. Well,

1049
00:55:58,330 --> 00:56:01,009
S1: it's not wizardry. They don't make it up each time.

1050
00:56:01,009 --> 00:56:04,130
S1: They should be following a set of steps, right? But

1051
00:56:04,130 --> 00:56:06,290
S1: it's not clear like we talked about in the beginning.

1052
00:56:06,290 --> 00:56:08,570
S1: It's not clear what those steps are. It's not clear

1053
00:56:08,569 --> 00:56:11,330
S1: if they're being followed correctly. It's not clear exactly what

1054
00:56:11,330 --> 00:56:13,689
S1: is happening inside the company. There's not even a list

1055
00:56:13,690 --> 00:56:15,969
S1: of them. And like I said, it's actually really difficult

1056
00:56:15,969 --> 00:56:18,890
S1: to go and collect a list. So look at this here.

1057
00:56:18,890 --> 00:56:21,169
S1: This is the type of view that somebody is going

1058
00:56:21,210 --> 00:56:23,610
S1: to have. They're going to have the ability, anyone, whoever

1059
00:56:23,610 --> 00:56:26,489
S1: has access, right, with the security permissions to be able

1060
00:56:26,489 --> 00:56:27,930
S1: to look at this stuff and they're going to be

1061
00:56:27,930 --> 00:56:31,450
S1: able to say, okay, well, we still got human components here.

1062
00:56:31,450 --> 00:56:35,049
S1: We got automated components here. Okay, we got review over here. Okay,

1063
00:56:35,089 --> 00:56:37,649
S1: we've got audit over here. We've got a quality check

1064
00:56:37,650 --> 00:56:40,850
S1: over here. This is the type of view that CEOs,

1065
00:56:40,890 --> 00:56:44,890
S1: CFOs like senior management, really any manager really wants to

1066
00:56:44,890 --> 00:56:47,009
S1: be able to see because you can't optimize what you

1067
00:56:47,009 --> 00:56:50,090
S1: don't understand. You can't optimize what you don't see. Right.

1068
00:56:50,130 --> 00:56:53,129
S1: All right. So this is actually another really cool aspect

1069
00:56:53,130 --> 00:56:56,689
S1: of once you have those visuals, once you have those workflows,

1070
00:56:56,830 --> 00:57:00,270
S1: things just fundamentally change. You know how like someone comes up?

1071
00:57:00,310 --> 00:57:02,189
S1: You know, they take you to a steak dinner, they

1072
00:57:02,190 --> 00:57:04,189
S1: meet you for some meeting or whatever, and they're like, hey,

1073
00:57:04,230 --> 00:57:08,350
S1: you know, we have the best image background removal process.

1074
00:57:08,350 --> 00:57:10,350
S1: I don't know what you're using for that today, but

1075
00:57:10,350 --> 00:57:12,629
S1: ours is way better. And so it's like, okay, well,

1076
00:57:12,670 --> 00:57:14,510
S1: I guess we'll do a pilot. You know, I really

1077
00:57:14,509 --> 00:57:16,870
S1: liked the steak. So I guess we'll do a pilot

1078
00:57:16,870 --> 00:57:18,830
S1: and I'll tell part of my team, you know, you're

1079
00:57:18,830 --> 00:57:21,550
S1: going to spend six weeks coming up with test criteria

1080
00:57:21,550 --> 00:57:23,230
S1: and we're going to test this thing and a B

1081
00:57:23,270 --> 00:57:25,669
S1: test or whatever. And if they like it, you know,

1082
00:57:25,710 --> 00:57:27,829
S1: I guess we'll take a look at it that goes

1083
00:57:27,830 --> 00:57:31,430
S1: away when you actually have a lattice over here of

1084
00:57:31,470 --> 00:57:34,950
S1: like all the different pieces that make up your company.

1085
00:57:34,950 --> 00:57:37,870
S1: These are the different processes, these are the different components.

1086
00:57:37,870 --> 00:57:40,230
S1: This is the tool, this is the decision. This is

1087
00:57:40,230 --> 00:57:43,830
S1: the human making the decision. This is an AI making decision.

1088
00:57:43,830 --> 00:57:46,870
S1: This is an AI processing work. This is the graph

1089
00:57:46,870 --> 00:57:51,510
S1: of algorithms that makes up the company. One particular node,

1090
00:57:51,510 --> 00:57:54,710
S1: this little node right here, one of hundreds or thousands

1091
00:57:54,710 --> 00:57:57,820
S1: or tens of thousands or millions of nodes inside your company.

1092
00:57:57,820 --> 00:58:03,540
S1: Is the process the algorithm for doing image background removal. Today,

1093
00:58:03,540 --> 00:58:06,300
S1: we don't have metrics on that. Really, most people don't.

1094
00:58:06,340 --> 00:58:08,180
S1: What do we have? We have like, yeah, it works

1095
00:58:08,180 --> 00:58:10,499
S1: pretty good. I like it, yeah, it's pretty good. Maybe

1096
00:58:10,500 --> 00:58:12,260
S1: we could pull some kind of metrics, but it's not

1097
00:58:12,260 --> 00:58:15,019
S1: like really good what we're going to have, what we're

1098
00:58:15,020 --> 00:58:18,620
S1: moving towards and most importantly, what people want to get to,

1099
00:58:18,660 --> 00:58:21,820
S1: what leadership wants to get to is all of these

1100
00:58:21,820 --> 00:58:25,860
S1: have metrics for them. They're all clearly defined. So now

1101
00:58:26,100 --> 00:58:30,660
S1: a vendor doesn't come with a steak dinner. They cannot anymore. Okay.

1102
00:58:30,700 --> 00:58:34,499
S1: Because we're going from wizardry to excel, okay? We show

1103
00:58:34,500 --> 00:58:36,820
S1: them our metrics. We show them, hey, look, this is

1104
00:58:36,820 --> 00:58:40,420
S1: how we do things. This is our current lattice structure.

1105
00:58:40,420 --> 00:58:44,580
S1: This is our current flow structure. This is image background removal.

1106
00:58:44,580 --> 00:58:46,499
S1: This is how much it costs us. This is how

1107
00:58:46,500 --> 00:58:49,260
S1: good it is. This is our quality ratings according to

1108
00:58:49,300 --> 00:58:51,980
S1: so and so standard. What are your ratings? What are

1109
00:58:51,980 --> 00:58:55,450
S1: your metrics? What are your cost numbers for this? If

1110
00:58:55,450 --> 00:58:57,450
S1: they can't bring that and all they brought was the

1111
00:58:57,490 --> 00:59:02,050
S1: steak dinner, that's not a conversation anymore. Once this transition happens, right?

1112
00:59:02,090 --> 00:59:06,890
S1: This is absolutely essential. Absolutely essential. Essentially, it's turning the

1113
00:59:06,890 --> 00:59:11,530
S1: company from these sort of black box, sort of vibey

1114
00:59:11,650 --> 00:59:14,690
S1: things that are happening that hopefully work out pretty well

1115
00:59:14,690 --> 00:59:21,290
S1: into actual pipelines, actual workflows with metrics and reliability and

1116
00:59:21,290 --> 00:59:23,890
S1: testing and all this sort of stuff as part of

1117
00:59:23,890 --> 00:59:25,930
S1: the system. All right. Real quick, this is just a

1118
00:59:25,930 --> 00:59:28,210
S1: little teaser. I'm not going to talk too much about this.

1119
00:59:28,210 --> 00:59:30,810
S1: This is a screenshot that I took earlier, I think

1120
00:59:30,850 --> 00:59:34,090
S1: last week. And this is actually just an example of

1121
00:59:34,090 --> 00:59:37,930
S1: the lattice system working. And it's actually one agent system

1122
00:59:37,930 --> 00:59:41,930
S1: reporting up into the lattice daemon basically describing its work.

1123
00:59:41,930 --> 00:59:44,970
S1: So this is a particular system called the algorithm. And

1124
00:59:44,970 --> 00:59:47,209
S1: it's running, I guess if you know, you know, but

1125
00:59:47,210 --> 00:59:50,050
S1: it's part of Pi and it runs work. This is

1126
00:59:50,050 --> 00:59:52,209
S1: the type of thing I'm just trying to show. This

1127
00:59:52,210 --> 00:59:55,200
S1: is not a theoretical thing. It is quite easy. And

1128
00:59:55,200 --> 00:59:57,600
S1: if you're an experienced user of AI, you already know

1129
00:59:57,600 --> 00:59:59,280
S1: you could do this type of stuff. This is the

1130
00:59:59,280 --> 01:00:02,200
S1: type of dashboard that line managers want to be able

1131
01:00:02,200 --> 01:00:04,600
S1: to see, that engineering managers want to be able to

1132
01:00:04,600 --> 01:00:07,240
S1: see that anybody should be able to drill into and

1133
01:00:07,240 --> 01:00:10,320
S1: actually see the work that's proceeding through the system. And

1134
01:00:10,320 --> 01:00:14,120
S1: this is making simple Rest calls and pushing this content

1135
01:00:14,120 --> 01:00:17,559
S1: up into the system and giving you a dashboard like this.

1136
01:00:17,600 --> 01:00:19,680
S1: All right. Essentially, what I want you to have up

1137
01:00:19,680 --> 01:00:22,120
S1: to this point is sort of dispelling some of the

1138
01:00:22,120 --> 01:00:25,519
S1: myths about how perfect humans are. We are not perfect.

1139
01:00:25,520 --> 01:00:28,360
S1: I'm not trying to talk crap about anybody. Some of

1140
01:00:28,360 --> 01:00:31,439
S1: my best friends are actually humans. So like, I'm not

1141
01:00:31,440 --> 01:00:34,600
S1: trying to throw stones here, right? I'm trying to describe

1142
01:00:34,720 --> 01:00:38,080
S1: the reality that's actually happening according to my view. Like

1143
01:00:38,120 --> 01:00:40,480
S1: you could watch all this and still think I'm wrong,

1144
01:00:40,480 --> 01:00:43,040
S1: but hopefully you won't. So where is this taking us?

1145
01:00:43,040 --> 01:00:45,800
S1: It seems like this is just horribly bad news. So

1146
01:00:45,800 --> 01:00:47,880
S1: what I want you to do have had so far

1147
01:00:48,000 --> 01:00:51,309
S1: is essentially the current state of work. The current state

1148
01:00:51,310 --> 01:00:54,150
S1: of human work, the current state of companies is not good.

1149
01:00:54,150 --> 01:00:58,189
S1: The perfect human thinking machine is not correct, right? The

1150
01:00:58,190 --> 01:01:01,830
S1: way that we actually process information, store information, it is

1151
01:01:01,830 --> 01:01:05,070
S1: much more like an AI than is actually comfortable to

1152
01:01:05,110 --> 01:01:08,749
S1: think about. And all that combined basically means the bar

1153
01:01:08,750 --> 01:01:11,310
S1: is actually extremely low. And then when you add on

1154
01:01:11,310 --> 01:01:14,550
S1: top of that, this lattice system, which is just one

1155
01:01:14,550 --> 01:01:17,910
S1: implementation that one person came up with, I'm telling you

1156
01:01:18,110 --> 01:01:20,870
S1: that is the direction that things are going. This is

1157
01:01:20,870 --> 01:01:24,789
S1: the architecture that companies are moving to that they are

1158
01:01:24,790 --> 01:01:27,350
S1: going to have. A lot of them haven't figured it

1159
01:01:27,350 --> 01:01:30,350
S1: out yet. Like this is still very early stuff, but

1160
01:01:30,350 --> 01:01:32,910
S1: people will catch on to this very soon and they

1161
01:01:32,910 --> 01:01:36,229
S1: will start building this type of architecture. And it's going

1162
01:01:36,270 --> 01:01:39,230
S1: to happen fast. It's going to happen extremely fast. Now,

1163
01:01:39,470 --> 01:01:41,990
S1: one other point I want to mention about this is

1164
01:01:42,150 --> 01:01:45,070
S1: a way to think about automation in a completely different way.

1165
01:01:45,070 --> 01:01:47,310
S1: If you think about it in terms of like, okay,

1166
01:01:47,350 --> 01:01:50,980
S1: we have the humans. Now let's think about adding AI.

1167
01:01:51,180 --> 01:01:54,420
S1: Oh my God, that's crazy. I have the same instinct. Okay,

1168
01:01:54,460 --> 01:01:58,860
S1: it feels icky because they're coming into our processes. Here's

1169
01:01:58,860 --> 01:02:01,340
S1: another way to think about this. Imagine you have a

1170
01:02:01,340 --> 01:02:06,060
S1: factory production line that produces iPhones. So it's soldering two

1171
01:02:06,300 --> 01:02:11,180
S1: nanometer transistors for iPhones. And this thing is amazing. It's

1172
01:02:11,180 --> 01:02:13,900
S1: producing whatever, $1 million a minute. Who knows what the

1173
01:02:13,900 --> 01:02:16,100
S1: numbers are, but it's just fully automated. It's a whole

1174
01:02:16,100 --> 01:02:19,500
S1: bunch of robots and it's just flowing perfectly. Who would say,

1175
01:02:19,540 --> 01:02:21,580
S1: you know what this thing really needs? It lacks a

1176
01:02:21,580 --> 01:02:24,060
S1: human touch. It lacks a human touch. We need to

1177
01:02:24,060 --> 01:02:26,740
S1: get people need jobs. People need jobs. We need to

1178
01:02:26,740 --> 01:02:30,499
S1: get some humans in here because AI's don't have the

1179
01:02:30,500 --> 01:02:33,260
S1: right thinking. They're just not as good as humans at

1180
01:02:33,260 --> 01:02:36,220
S1: doing work. Do you have any idea what would happen

1181
01:02:36,780 --> 01:02:40,260
S1: if you tried to ask a human to solder two

1182
01:02:40,300 --> 01:02:44,300
S1: nanometer transistors onto circuit boards? It would not go well.

1183
01:02:44,340 --> 01:02:47,900
S1: Also known as impossible. Okay. Another one which I think

1184
01:02:47,940 --> 01:02:51,160
S1: can't remember who mentioned this. Somebody mentioned an example of

1185
01:02:51,160 --> 01:02:54,560
S1: a amazing Excel sheet. We've all seen one of these before.

1186
01:02:54,600 --> 01:02:57,520
S1: It's got a million tabs. Looks like an actual application.

1187
01:02:57,520 --> 01:03:00,320
S1: The thing is just fantastic. Maybe it handles the entire

1188
01:03:00,320 --> 01:03:04,480
S1: company's finances and just produces the best results. It never

1189
01:03:04,480 --> 01:03:07,600
S1: messes up. It's just awesome. It's like, you know, 36

1190
01:03:07,600 --> 01:03:10,919
S1: months of work goes into this crown jewel of an

1191
01:03:10,920 --> 01:03:14,400
S1: Excel spreadsheet, and it's just processing millions and millions of

1192
01:03:14,400 --> 01:03:18,880
S1: lines for the entire enterprise globally. Where on the spreadsheet

1193
01:03:19,000 --> 01:03:22,360
S1: should we add a human to help the spreadsheet? Where

1194
01:03:22,360 --> 01:03:26,520
S1: do we add it when automation works, when AI workflows work,

1195
01:03:26,520 --> 01:03:29,600
S1: when the lattice system that I was showing you works,

1196
01:03:29,600 --> 01:03:33,240
S1: the answer will not be to add humans to the thing.

1197
01:03:33,240 --> 01:03:37,960
S1: It becomes extremely obvious that adding humans actually makes it worse. Okay,

1198
01:03:38,000 --> 01:03:40,360
S1: with an exception, which we're going to talk about. So

1199
01:03:40,520 --> 01:03:44,480
S1: think about that inversion when you're considering the uncomfortable feeling

1200
01:03:44,480 --> 01:03:47,590
S1: that you're getting when you're thinking about adding automation, think

1201
01:03:47,590 --> 01:03:51,310
S1: of the opposite. The default is opposite automation first. And

1202
01:03:51,310 --> 01:03:53,709
S1: there's another reason to think that one saying that I

1203
01:03:53,710 --> 01:03:56,150
S1: have that I think is very, very true is the

1204
01:03:56,150 --> 01:03:59,830
S1: ideal number of human employees for a company is zero. Now,

1205
01:03:59,830 --> 01:04:02,229
S1: I'm not talking about small boutique startups. I'm not talking

1206
01:04:02,230 --> 01:04:05,310
S1: about collaborations with your friends. I'm not talking about like

1207
01:04:05,350 --> 01:04:07,630
S1: sole projects that you just love to do. You want

1208
01:04:07,630 --> 01:04:10,229
S1: to work with other humans? I do too, that's always

1209
01:04:10,230 --> 01:04:13,070
S1: going to be there. I'm talking about for large companies,

1210
01:04:13,070 --> 01:04:16,550
S1: medium sized companies, companies that need to do things at scale.

1211
01:04:16,550 --> 01:04:21,630
S1: These companies, again, they are paying $50 trillion in knowledge

1212
01:04:21,630 --> 01:04:25,750
S1: work compensation. They want to do that work internally. Okay.

1213
01:04:25,990 --> 01:04:28,870
S1: If I had an ice cream truck business and all

1214
01:04:28,870 --> 01:04:30,990
S1: I had was my ice cream truck, and I had

1215
01:04:30,990 --> 01:04:33,750
S1: a machine that could actually make the perfect ice cream cone.

1216
01:04:33,750 --> 01:04:37,430
S1: And maybe I have three trucks and I'm the only employee.

1217
01:04:37,430 --> 01:04:40,070
S1: And the trucks, they have X number of machines. They

1218
01:04:40,070 --> 01:04:42,070
S1: can pump out the things. I can hand them off

1219
01:04:42,070 --> 01:04:44,990
S1: to the kids with a lollipop or whatever, and I'm

1220
01:04:44,990 --> 01:04:48,060
S1: making X amount of money and I'm happy. I'm just happy.

1221
01:04:48,060 --> 01:04:50,260
S1: This is a great amount of money. That is the

1222
01:04:50,260 --> 01:04:52,620
S1: state that companies want to be in. Of course they

1223
01:04:52,620 --> 01:04:54,979
S1: want to grow, right? So they'll want more and more

1224
01:04:54,980 --> 01:04:57,780
S1: bigger trucks, you know, more stuff like that. But they

1225
01:04:57,780 --> 01:05:01,180
S1: won't say, I need more humans to add to this.

1226
01:05:01,220 --> 01:05:03,459
S1: If I'm sitting here with my ice cream truck and

1227
01:05:03,460 --> 01:05:05,939
S1: I've got my ice cream cone makers, and I am

1228
01:05:05,940 --> 01:05:08,460
S1: perfectly happy with the amount of money that I'm making,

1229
01:05:08,460 --> 01:05:10,459
S1: if a group of people come and they pick at

1230
01:05:10,460 --> 01:05:13,900
S1: me and they're like, you're against humans. Humans need jobs,

1231
01:05:13,900 --> 01:05:16,580
S1: humans need jobs. I'm like, what are you talking about?

1232
01:05:16,580 --> 01:05:19,220
S1: I got my truck. I've got my ice cream cone makers.

1233
01:05:19,220 --> 01:05:23,460
S1: It's just me. This is literally me. Those machines are me.

1234
01:05:23,460 --> 01:05:25,939
S1: And this ice cream truck is me. This, as a

1235
01:05:25,940 --> 01:05:30,140
S1: unit is me by myself doing the work. Well, guess what?

1236
01:05:30,140 --> 01:05:32,499
S1: That diagram I showed you of the lattice system of

1237
01:05:32,500 --> 01:05:35,540
S1: the entire company. Let's just say nothing in there was human,

1238
01:05:35,540 --> 01:05:38,220
S1: which that'll be forever before that happens. Or a long

1239
01:05:38,220 --> 01:05:41,140
S1: time anyway. Let's just say it's fully automated. Everything in

1240
01:05:41,140 --> 01:05:43,700
S1: the whole lattice structure, it's all automated, or let's say

1241
01:05:43,700 --> 01:05:47,330
S1: it's a small company or whatever. Fully automated. Right? Fully automated.

1242
01:05:47,330 --> 01:05:51,050
S1: It's working great. That is their company. They consider that

1243
01:05:51,050 --> 01:05:54,249
S1: to be them doing the work. All that infrastructure, all

1244
01:05:54,290 --> 01:05:57,770
S1: that automation, all the machines, all the tech. It is

1245
01:05:57,770 --> 01:06:00,450
S1: part of the company now. It's like the walls and

1246
01:06:00,450 --> 01:06:03,610
S1: the floors and the ceilings and the chairs and the desks.

1247
01:06:03,610 --> 01:06:06,330
S1: It's literally just part of the core of the company

1248
01:06:06,330 --> 01:06:09,730
S1: that is the preferred mode. They are currently paying $50

1249
01:06:09,730 --> 01:06:14,209
S1: trillion for outside help. They want that help to be inside.

1250
01:06:14,210 --> 01:06:17,209
S1: They want it to be all internal, okay? That is

1251
01:06:17,210 --> 01:06:20,250
S1: the desired state. And going back to the whole setup

1252
01:06:20,250 --> 01:06:23,410
S1: with like the factory and adding humans, one argument I've

1253
01:06:23,410 --> 01:06:26,410
S1: heard is like, well, yeah, but if you want to scale, right?

1254
01:06:26,450 --> 01:06:29,249
S1: That's there's no problem. You'll just hire more humans. Like

1255
01:06:29,250 --> 01:06:32,450
S1: why wouldn't you hire more humans? Because humans can also

1256
01:06:32,450 --> 01:06:35,090
S1: do the job. Well, no, they can't do the job

1257
01:06:35,090 --> 01:06:38,410
S1: as good as it's already being done by the automation. Right?

1258
01:06:38,450 --> 01:06:41,570
S1: So that's where that argument breaks down. You don't add

1259
01:06:41,570 --> 01:06:45,040
S1: humans to a working system that doesn't have humans in it,

1260
01:06:45,040 --> 01:06:47,920
S1: because the working system without humans in it is likely

1261
01:06:47,920 --> 01:06:51,200
S1: to be many times better and many times faster for

1262
01:06:51,200 --> 01:06:53,999
S1: a lot of tasks, maybe most tasks. So just keep

1263
01:06:54,000 --> 01:06:56,200
S1: that in mind. All right, so here we are. I

1264
01:06:56,200 --> 01:06:58,560
S1: promise some good news and this is where we get

1265
01:06:58,560 --> 01:07:00,600
S1: to it. So we've talked about the bad news. We've

1266
01:07:00,640 --> 01:07:03,440
S1: talked about what the problems are and what I think

1267
01:07:03,440 --> 01:07:07,240
S1: the inevitable solution is. And I do believe that is true.

1268
01:07:07,240 --> 01:07:09,999
S1: Now let's talk about, okay, what are humans supposed to do?

1269
01:07:10,000 --> 01:07:12,680
S1: What is the good news? This sounds like horribly bad

1270
01:07:12,680 --> 01:07:16,280
S1: news with no possible upside. All the humans are screwed.

1271
01:07:16,280 --> 01:07:18,320
S1: What are we going to do? And for that, I

1272
01:07:18,320 --> 01:07:21,360
S1: want to bring you over to this content here. All right,

1273
01:07:21,360 --> 01:07:23,200
S1: so for that I want to bring you over to

1274
01:07:23,240 --> 01:07:26,280
S1: this content here. So remember when we talked about the

1275
01:07:26,280 --> 01:07:31,999
S1: four layers and we said, okay, knowledge, understanding, intelligence and

1276
01:07:32,000 --> 01:07:35,480
S1: creativity and we said, look, this is the stuff that

1277
01:07:35,520 --> 01:07:39,240
S1: humans can do. But there were two layers underneath. And

1278
01:07:39,240 --> 01:07:42,660
S1: these are the layers that eyes don't have. Okay. They

1279
01:07:42,660 --> 01:07:48,380
S1: don't have subjective experience. They don't have desires to do anything. Okay,

1280
01:07:48,420 --> 01:07:53,020
S1: so imagine that whole workflow again, fully automated. It's amazing.

1281
01:07:53,020 --> 01:07:56,100
S1: And it doesn't need any humans because it's just better

1282
01:07:56,100 --> 01:08:00,020
S1: and faster and more consistent, higher quality, all of that

1283
01:08:00,020 --> 01:08:03,700
S1: cheaper way, way cheaper. Fine. Who decides what to make?

1284
01:08:03,740 --> 01:08:06,380
S1: Who decides what company to make in the first place?

1285
01:08:06,380 --> 01:08:09,100
S1: Here's what AI doesn't have, and this should be a

1286
01:08:09,100 --> 01:08:12,300
S1: source of hope. Hopefully what I've done here is show

1287
01:08:12,300 --> 01:08:15,740
S1: you how bad one part of the situation is, and

1288
01:08:15,940 --> 01:08:19,140
S1: it should be equally clear. Hopefully, after what I'm about

1289
01:08:19,140 --> 01:08:22,179
S1: to show you that the opposite is also true. What

1290
01:08:22,180 --> 01:08:25,460
S1: this unlocks for humans is actually what we should have

1291
01:08:25,460 --> 01:08:29,420
S1: been doing all along. We should not be the row

1292
01:08:29,420 --> 01:08:32,380
S1: in the Excel sheet. We should not be the section

1293
01:08:32,380 --> 01:08:36,100
S1: of the robot army that you know solders the transistor.

1294
01:08:36,100 --> 01:08:39,140
S1: We should not be the one in the car dreading

1295
01:08:39,180 --> 01:08:42,370
S1: going into the office Monday morning that all should never

1296
01:08:42,370 --> 01:08:46,290
S1: have happened. Which, like I said, we knew that before 2022.

1297
01:08:46,330 --> 01:08:49,570
S1: It's totally understandable to freak out when the jobs are

1298
01:08:49,570 --> 01:08:52,130
S1: going away and AI is taking them. The natural tendency

1299
01:08:52,130 --> 01:08:54,450
S1: of someone yanx something out of your arms is to

1300
01:08:54,490 --> 01:08:58,090
S1: pull it closer, right? Understandable. Like not mad at anyone

1301
01:08:58,090 --> 01:09:00,370
S1: for that. I'm just trying to help you sort of

1302
01:09:00,410 --> 01:09:04,650
S1: think through this. That work was garbage. It is garbage.

1303
01:09:04,650 --> 01:09:07,930
S1: It was garbage for decades. It still is garbage. And

1304
01:09:07,970 --> 01:09:10,370
S1: AI's are coming to take it. Fine. The place that

1305
01:09:10,370 --> 01:09:12,810
S1: we're going, the place that we need to get to,

1306
01:09:12,850 --> 01:09:16,770
S1: is where humans actually have ideas. People are actually broadcasting

1307
01:09:16,770 --> 01:09:19,610
S1: their own capabilities and their own plans and their own

1308
01:09:19,610 --> 01:09:22,770
S1: creativity and their own beliefs and their own ideas and

1309
01:09:22,770 --> 01:09:28,250
S1: their own creations. Nurturing people, helping people grow, creating new art,

1310
01:09:28,290 --> 01:09:31,650
S1: new experiences for other people. And guess what? Coming up

1311
01:09:31,650 --> 01:09:34,850
S1: with companies to do this right. How will the company

1312
01:09:34,850 --> 01:09:37,930
S1: be implemented? How will all the work tasks get done

1313
01:09:37,970 --> 01:09:40,800
S1: using a system similar to Latisse. It's going to be

1314
01:09:40,800 --> 01:09:44,080
S1: an all AI system doing the actual crunching of the work.

1315
01:09:44,080 --> 01:09:46,920
S1: But what can the AI not do? It doesn't have

1316
01:09:46,920 --> 01:09:50,200
S1: a desire. It doesn't have any goals. It doesn't have ambitions.

1317
01:09:50,240 --> 01:09:53,000
S1: AI doesn't see a problem in the world and go, yeah,

1318
01:09:53,000 --> 01:09:54,920
S1: I just don't like that. I just don't like that.

1319
01:09:54,960 --> 01:09:57,000
S1: You know, that's not the way they should be treated.

1320
01:09:57,040 --> 01:09:59,760
S1: You know, this is a horrible experience. I don't like

1321
01:09:59,760 --> 01:10:01,960
S1: this product. We should make a better one. AI is

1322
01:10:01,960 --> 01:10:04,720
S1: not doing that. Why? Because it doesn't experience anything. It

1323
01:10:04,720 --> 01:10:08,160
S1: doesn't experience problems and feel bad. It doesn't come up

1324
01:10:08,160 --> 01:10:10,440
S1: with new ideas. If you ask it to come up

1325
01:10:10,440 --> 01:10:12,680
S1: with a new idea, it can. But it's not sitting

1326
01:10:12,680 --> 01:10:16,960
S1: around dormant in like unhappy. We are unhappy. We have

1327
01:10:16,960 --> 01:10:21,960
S1: these desires and we have this subjective experience for another reason, right?

1328
01:10:22,000 --> 01:10:25,160
S1: It's because we are powered by evolution. This is kind

1329
01:10:25,160 --> 01:10:28,040
S1: of a big idea as well. Might be slightly controversial,

1330
01:10:28,040 --> 01:10:30,040
S1: I think if you're watching the channel, might not be

1331
01:10:30,040 --> 01:10:33,760
S1: as controversial as it is to most people, but evolution

1332
01:10:33,760 --> 01:10:36,320
S1: is the thing that gave us our desires. When we

1333
01:10:36,360 --> 01:10:39,990
S1: sit around ruminating, when that whole freight train of ideas

1334
01:10:39,990 --> 01:10:42,910
S1: is just scrolling across the things that's in our mind

1335
01:10:42,950 --> 01:10:47,310
S1: is generated by evolution. The reason we have these desires

1336
01:10:47,310 --> 01:10:51,710
S1: and goals and ambitions is because evolution is what designed us.

1337
01:10:51,710 --> 01:10:55,070
S1: We are basically mech suits, we're giant mechs. We're like

1338
01:10:55,070 --> 01:10:59,470
S1: giant transformer mech suits. Around evolution, which is trying to

1339
01:10:59,470 --> 01:11:02,350
S1: do things and it's trying to do things through us.

1340
01:11:02,390 --> 01:11:08,390
S1: Our entire pleasure system, our chemicals, dopamine, serotonin, cortisol, all

1341
01:11:08,390 --> 01:11:11,430
S1: these things are designed. The whole purpose that they're there

1342
01:11:11,430 --> 01:11:15,350
S1: is to steer us, to steer us towards desires. Do

1343
01:11:15,350 --> 01:11:17,910
S1: we have any control over our desires? Can you say

1344
01:11:17,910 --> 01:11:21,310
S1: I suddenly like broccoli when you didn't a second ago?

1345
01:11:21,350 --> 01:11:23,630
S1: Can you decide to do that? No you can't. You

1346
01:11:23,630 --> 01:11:27,510
S1: also can't really control what you desire. You can't really

1347
01:11:27,510 --> 01:11:30,470
S1: control your ambitions. You kind of have them. You could

1348
01:11:30,470 --> 01:11:32,990
S1: try to use your intellect and steer them, I'll grant

1349
01:11:32,990 --> 01:11:35,780
S1: you that. But in general, we are being powered. We

1350
01:11:35,820 --> 01:11:39,700
S1: are being powered by a very ancient, powerful thing, which

1351
01:11:39,700 --> 01:11:45,340
S1: is our upbringing, our evolution, our hardware, our mammal minds, right?

1352
01:11:45,380 --> 01:11:51,700
S1: Our brains, our minds, our identity. It is driving forward, survive, reproduce,

1353
01:11:51,700 --> 01:11:58,340
S1: become attractive, become powerful, become respected. That drive makes us fundamentally,

1354
01:11:58,340 --> 01:12:01,660
S1: I would say better, definitely different, but I would say better.

1355
01:12:01,660 --> 01:12:05,540
S1: It gives us an advantage over eyes. Eyes are dormant.

1356
01:12:05,540 --> 01:12:08,500
S1: They just sit there. They just execute things that we want.

1357
01:12:08,540 --> 01:12:11,300
S1: When they act like they feel things, when they act

1358
01:12:11,300 --> 01:12:13,700
S1: like they have goals and they act like, hey, I'm

1359
01:12:13,700 --> 01:12:16,340
S1: going to generate ideas that is not coming from their

1360
01:12:16,340 --> 01:12:19,100
S1: true center of self the way that it does for humans.

1361
01:12:19,100 --> 01:12:21,860
S1: So what does this mean? What this means is we

1362
01:12:21,860 --> 01:12:25,020
S1: are still the centerpiece of this whole thing. Everything we've

1363
01:12:25,020 --> 01:12:28,460
S1: been talking about in this video, everything around why AI

1364
01:12:28,460 --> 01:12:32,260
S1: is going to replace us doing work is around the execution.

1365
01:12:32,260 --> 01:12:36,370
S1: It's all around the implementation of someone's idea of a

1366
01:12:36,370 --> 01:12:40,090
S1: founder's idea of a company builders idea. Well guess what?

1367
01:12:40,090 --> 01:12:42,810
S1: There's a trillion more ideas out there that need to

1368
01:12:42,810 --> 01:12:45,730
S1: be made, and AI makes it so that it's easy

1369
01:12:45,730 --> 01:12:48,530
S1: to go and actually implement them, because you could just

1370
01:12:48,530 --> 01:12:51,410
S1: implement a system like lattice and stand up your company

1371
01:12:51,410 --> 01:12:53,650
S1: and you have some humans there, you have your buddies there,

1372
01:12:53,650 --> 01:12:56,210
S1: you have a bunch of A's and a bunch of agents.

1373
01:12:56,210 --> 01:12:59,689
S1: You have transparency. You say, okay, I want to execute this.

1374
01:12:59,690 --> 01:13:01,890
S1: I want to do this. Where are humans in this

1375
01:13:01,890 --> 01:13:05,650
S1: process of like starting companies, managing companies, doing all this stuff?

1376
01:13:05,690 --> 01:13:08,010
S1: They're at the top. We are at the top. Do

1377
01:13:08,010 --> 01:13:10,690
S1: you remember how you saw all those workflows and all

1378
01:13:10,690 --> 01:13:13,930
S1: those diagrams, all those interconnected pieces? What is the role

1379
01:13:13,930 --> 01:13:17,130
S1: of the human when the AI does all that work better?

1380
01:13:17,130 --> 01:13:19,370
S1: The role the human is to say, I don't want

1381
01:13:19,370 --> 01:13:22,530
S1: you to do that task anymore, okay, I can't improve

1382
01:13:22,530 --> 01:13:25,690
S1: how you do the task. Your AI, you're great. Do

1383
01:13:25,690 --> 01:13:29,290
S1: the workflow. You got it faster, cheaper, great. Who determines

1384
01:13:29,290 --> 01:13:32,050
S1: what the company should be doing, why they're doing it,

1385
01:13:32,090 --> 01:13:35,630
S1: what problem they're solving that is only human, right? This

1386
01:13:35,630 --> 01:13:40,710
S1: is incredibly positive. Incredibly optimistic. This is what we do

1387
01:13:40,750 --> 01:13:44,230
S1: that they can't. Okay. And this is where my recommendation

1388
01:13:44,230 --> 01:13:46,550
S1: and everything I talk about in all my videos, all

1389
01:13:46,550 --> 01:13:49,710
S1: my open source projects, my recommendation is to make the

1390
01:13:49,710 --> 01:13:54,190
S1: switch from focusing on the execution and the implementation, which

1391
01:13:54,190 --> 01:13:56,310
S1: is all the first part of this video. It is

1392
01:13:56,310 --> 01:13:59,230
S1: all this debate about can AI do it? Can AI

1393
01:13:59,310 --> 01:14:02,430
S1: not do it? It's all it's all a distraction. AI

1394
01:14:02,430 --> 01:14:05,710
S1: is going to crush that stuff. Those jobs are gone.

1395
01:14:05,710 --> 01:14:09,550
S1: That's why I consider this threat so dangerous. That's why

1396
01:14:09,550 --> 01:14:12,750
S1: I made this video to address this narrative that our

1397
01:14:12,750 --> 01:14:15,870
S1: jobs are not in danger from AI. It's because if

1398
01:14:15,910 --> 01:14:18,910
S1: you don't realize this, then you will keep focusing your

1399
01:14:18,910 --> 01:14:22,030
S1: energy on trying to maintain this job that is going away.

1400
01:14:22,030 --> 01:14:25,510
S1: Instead of transitioning all your effort to getting to the

1401
01:14:25,510 --> 01:14:28,150
S1: other side. This thing I call human 3.0, it is

1402
01:14:28,150 --> 01:14:33,420
S1: basically becoming your full self, Magnifying your true self, understanding

1403
01:14:33,420 --> 01:14:37,540
S1: your true self, broadcasting it and interconnecting that with other

1404
01:14:37,540 --> 01:14:39,780
S1: humans who are doing the same. So we can have

1405
01:14:39,780 --> 01:14:42,900
S1: value exchange from human to human without the need for

1406
01:14:42,900 --> 01:14:46,460
S1: this corporate hierarchy, all these corporations, all that kind of stuff.

1407
01:14:46,460 --> 01:14:49,100
S1: It's all, it's all going away. First of all, but

1408
01:14:49,100 --> 01:14:51,100
S1: it's not good in the first place. We don't need

1409
01:14:51,100 --> 01:14:52,740
S1: to hold on to it. We need to get to

1410
01:14:52,780 --> 01:14:55,020
S1: the other side where we can look down at a

1411
01:14:55,020 --> 01:14:58,140
S1: lattice like system where all the AI is executing, but

1412
01:14:58,140 --> 01:15:00,860
S1: we got to decide what to build. We built that company,

1413
01:15:00,860 --> 01:15:03,500
S1: we started that business. We can have 20 of them,

1414
01:15:03,500 --> 01:15:05,460
S1: and we have a dashboard that looks down at all

1415
01:15:05,460 --> 01:15:07,500
S1: 20 of them, and they're all doing awesome stuff. We

1416
01:15:07,500 --> 01:15:09,140
S1: could drill in and we could say, hey, we're not

1417
01:15:09,140 --> 01:15:11,460
S1: doing this anymore. We're doing it in a different way. Hey,

1418
01:15:11,460 --> 01:15:13,100
S1: I want to optimize this. I want to do it

1419
01:15:13,100 --> 01:15:16,340
S1: that way. And the AI goes and executes. This is

1420
01:15:16,340 --> 01:15:19,140
S1: why I care so much about this. This is why

1421
01:15:19,420 --> 01:15:24,100
S1: I counter this argument that AI will not replace knowledge work.

1422
01:15:24,100 --> 01:15:26,220
S1: It is going to replace it. And we need to

1423
01:15:26,220 --> 01:15:29,260
S1: get ready. And it's not bad news. It's actually really

1424
01:15:29,260 --> 01:15:30,820
S1: good news. We'll see you in the next one.