1
00:00:00,017 --> 00:00:03,707
We just got a damning bit of evidence, which tells us how soon AI is going to

2
00:00:03,707 --> 00:00:07,575
replace our jobs. And unfortunately, it's not looking great.

3
00:00:07,888 --> 00:00:13,957
The report states that there's a 35 to 65% chance that this happens by the end of 2027.

4
00:00:13,957 --> 00:00:18,507
That's just over a year from now. An AI capable of replacing any human expert

5
00:00:18,507 --> 00:00:20,967
in any field, industry, or vertical.

6
00:00:20,967 --> 00:00:23,997
And understandably, the reaction to this has been incredibly negative.

7
00:00:23,997 --> 00:00:28,797
In fact, one of the lead authors of the report said that he's scared, sad, and angry.

8
00:00:28,797 --> 00:00:32,717
But the craziest part about all of this is that the models that they're referring

9
00:00:32,717 --> 00:00:35,397
to aren't publicly available to you or I.

10
00:00:35,397 --> 00:00:37,457
If you look at some of the leading models in the world right now,

11
00:00:37,457 --> 00:00:43,037
Fable 5.2, Gemini Carbon, GPT 6.2, Astra, these are all models that have been

12
00:00:43,037 --> 00:00:47,791
throttled. They've got safeguards on them. They actually haven't got their true power released.

13
00:00:48,082 --> 00:00:50,607
What's actually happening is some of the leading labs, Anthropic,

14
00:00:50,607 --> 00:00:53,939
OpenAI, are keeping models internally.

15
00:00:54,107 --> 00:00:57,637
Some of the most powerful models aren't available to everyone to use.

16
00:00:57,805 --> 00:01:01,427
They're gatekeeping access to it. So the problem here isn't only that these

17
00:01:01,427 --> 00:01:05,817
models are being created, but 99% of the rest of the world can't actually get

18
00:01:05,817 --> 00:01:09,247
access to these things or have a say in how they're being developed.

19
00:01:09,247 --> 00:01:12,647
And the reason why I want to talk about this today is because there's a growing

20
00:01:12,647 --> 00:01:18,157
concern or gap between the 1% of researchers, a couple hundred people that are

21
00:01:18,157 --> 00:01:21,717
building the most powerful models and technology that we've ever seen before

22
00:01:21,717 --> 00:01:24,117
that are going to change and have a huge impact on the world.

23
00:01:24,686 --> 00:01:29,277
And the rest of the world, the 99% of the world that has no idea it's happening.

24
00:01:29,277 --> 00:01:32,357
And that gap is increasing wider and wider.

25
00:01:32,357 --> 00:01:36,837
The idea is I want to raise awareness about a very important thing that is happening

26
00:01:36,837 --> 00:01:41,307
under the hood behind closed doors and how it's not actually the rest of the

27
00:01:41,307 --> 00:01:44,517
world's fault because these models are being kept away from them.

28
00:01:44,517 --> 00:01:47,019
So let's get into it on today's episode brought to you by Qualcomm.

29
00:01:50,926 --> 00:01:55,765
Now, this update references a very famous report that predicted how AI was going

30
00:01:55,765 --> 00:01:59,448
to play out over the next year and a half, and it got released in 2025,

31
00:01:59,802 --> 00:02:03,695
and it has so far been the most accurate predictor to date.

32
00:02:03,695 --> 00:02:07,065
The people behind this are part of a group called the AI Futures Project,

33
00:02:07,065 --> 00:02:11,755
and they're made up of esteemed AI researchers who left AI labs to focus on

34
00:02:11,755 --> 00:02:14,535
making sure that AI is stewarded in the right way.

35
00:02:14,535 --> 00:02:18,315
Now, when this report got released last year, one of its boldest statements

36
00:02:18,315 --> 00:02:23,555
was that we would have two things by 2027 or by the end of 2027.

37
00:02:23,555 --> 00:02:30,195
One, an AI that is as smart as the smartest AI researcher. So an AI powerful

38
00:02:30,195 --> 00:02:34,205
enough to build its own version of itself and advance AI research.

39
00:02:34,205 --> 00:02:36,875
We don't need some of the smartest AI humans anymore.

40
00:02:37,351 --> 00:02:43,595
And two, an AI that is so capable, so expert-led that it beats any human expert

41
00:02:43,595 --> 00:02:46,175
at anything. Those are the two milestones that it was tracking.

42
00:02:46,506 --> 00:02:51,275
And its answer was, it's going to happen in 2027. That was back in April of

43
00:02:51,275 --> 00:02:54,366
2025. A few months later, they reneged on it. And they said,

44
00:02:54,720 --> 00:02:58,412
we don't see advancements of AI models happening as quickly as we expected.

45
00:02:58,719 --> 00:03:03,796
So we're going to push our timelines back to say 2031 to 2035.

46
00:03:04,307 --> 00:03:08,465
And then more recently, after Claude Code got launched, and all these AI models

47
00:03:08,465 --> 00:03:13,195
started ascending really highly and steeply for a much lower cost in terms of

48
00:03:13,195 --> 00:03:15,633
intelligence, they released a new update

49
00:03:15,882 --> 00:03:21,285
based on actual evidence from insights into Anthropic's closed internal models,

50
00:03:21,285 --> 00:03:23,824
which suggested not only were they right,

51
00:03:24,323 --> 00:03:26,082
but they were actually too late.

52
00:03:26,331 --> 00:03:30,012
The stuff that they had predicted is going to happen five months earlier.

53
00:03:30,186 --> 00:03:33,735
Here are some of the headline numbers. AI research and development would improve

54
00:03:33,735 --> 00:03:38,145
at more than double the rate that it was improving already. They estimated that

55
00:03:38,145 --> 00:03:40,914
this would come around the start of 2027.

56
00:03:41,192 --> 00:03:44,835
Turns out it's already here. An independent group of researchers were given

57
00:03:44,835 --> 00:03:47,822
access to anthropics internal models over the last couple of months,

58
00:03:48,222 --> 00:03:49,180
and they determined that,

59
00:03:49,795 --> 00:03:53,741
pretty concretely, that they'd already achieved that 2x improvement.

60
00:03:53,741 --> 00:03:58,581
What this means, by the way, is that AI model research was already limited and

61
00:03:58,581 --> 00:04:01,461
constrained by the number of human experts that were available.

62
00:04:01,461 --> 00:04:04,841
Remember, all these researchers are getting hired for hundreds of millions of

63
00:04:04,841 --> 00:04:06,060
dollars in compensation.

64
00:04:06,390 --> 00:04:11,040
Now they have an AI model that is competent of automating 80% of that,

65
00:04:11,249 --> 00:04:14,999
which helped them achieve this 2x five months earlier. That's prediction number one.

66
00:04:15,464 --> 00:04:19,971
Prediction number two means that an AI automated researcher,

67
00:04:19,971 --> 00:04:24,031
which is an AI model that is competent enough to just build the next best version

68
00:04:24,031 --> 00:04:29,065
of itself, is now going to arrive 19 months earlier.

69
00:04:29,419 --> 00:04:35,682
They expected this to happen around 2029. Now it's going to happen by May of next year.

70
00:04:35,822 --> 00:04:40,791
And the third most scary prediction is an AI that is competent enough to replace

71
00:04:40,791 --> 00:04:43,011
any human expert across any field.

72
00:04:43,011 --> 00:04:45,941
So think of any job that's out there that's reliant on software,

73
00:04:45,941 --> 00:04:48,911
and maybe some physical labor will be able to get replaced.

74
00:04:48,911 --> 00:04:52,521
Now, the focus of this prediction is on white-collar work specifically,

75
00:04:52,521 --> 00:04:56,591
so software-based tasks, because of course, the physical constraints are limited

76
00:04:56,591 --> 00:04:58,481
to how fast robotics advance.

77
00:04:58,481 --> 00:05:04,451
But the idea here is they have a 35% to 65% chance estimate that this happens

78
00:05:04,451 --> 00:05:08,619
by the end of next year, certainly in 2028.

79
00:05:08,903 --> 00:05:12,141
So just to put that into perspective, we're looking at like maybe a year and

80
00:05:12,141 --> 00:05:17,402
a half left before we have AI models that are smarter than the smartest human in the world.

81
00:05:17,674 --> 00:05:21,401
And they're going to be readily available and cheap to access by almost anyone

82
00:05:21,401 --> 00:05:24,896
or anything. AI agents and AI models just running 24-7 online.

83
00:05:25,285 --> 00:05:28,211
Doing a lot of the work and jobs that you and I do today.

84
00:05:28,420 --> 00:05:32,208
And listen, the evidence for this has sort of been emerging in small ways for

85
00:05:32,208 --> 00:05:33,789
publicly accessible models.

86
00:05:33,928 --> 00:05:37,768
I, for one, can say that a lot of the research, a lot of the development,

87
00:05:37,768 --> 00:05:41,504
the editing, the processes that I spend so much time committing.

88
00:05:42,168 --> 00:05:47,578
To every single day, every single month, the long-term tasks are now being automated

89
00:05:47,578 --> 00:05:48,968
by a lot of the newer models.

90
00:05:48,968 --> 00:05:53,526
They are getting smarter, more intuitive. They've built up on a lot on my personal context.

91
00:05:53,787 --> 00:05:57,809
These models are now very good, way better than a human, way faster than a human

92
00:05:58,024 --> 00:06:03,768
at using an actual computer, and they can run 24-7 for as much as 20 bucks a

93
00:06:03,768 --> 00:06:08,368
month. So the idea here of having an automated worker isn't too far off.

94
00:06:08,658 --> 00:06:12,888
And indeed, we're seeing this with some of the lead flagship products from the AI labs.

95
00:06:12,888 --> 00:06:17,498
If you look at Elon Musk releasing GrokBot from SpaceX, if you're looking at

96
00:06:17,498 --> 00:06:23,648
OpenAI releasing Dots, these AI agents are very competent at automating a lot of work tools.

97
00:06:23,648 --> 00:06:27,275
And they can get access to these tools like Slack, Google Mail,

98
00:06:27,455 --> 00:06:31,246
like G Drive, like Sheets. You saw Anthropic release a few days ago.

99
00:06:31,460 --> 00:06:33,602
They're a competitor to Microsoft Office.

100
00:06:33,875 --> 00:06:38,248
The idea here is AI is getting competent enough to do a lot of software-based

101
00:06:38,248 --> 00:06:43,872
tasks. Now, that's not to say that humans don't have intuition or the right idea or experience.

102
00:06:44,069 --> 00:06:49,990
The idea here is that you can give an AI a single prompt or an idea of what to get done.

103
00:06:50,292 --> 00:06:54,278
And it does a very good job of doing it. Now, if you compare this to models

104
00:06:54,278 --> 00:06:58,078
just, I don't know, three to six months ago, you wouldn't have believed this.

105
00:06:58,078 --> 00:07:02,208
But the rapid rate at which these models are improving is evidence that what

106
00:07:02,208 --> 00:07:04,118
this report is stating is actually true.

107
00:07:04,118 --> 00:07:07,479
And you can extrapolate what that's going to be like 12 months from now to be,

108
00:07:07,694 --> 00:07:09,948
maybe these models will be better than me at doing my job.

109
00:07:09,948 --> 00:07:13,848
And there have been screams for awareness about this from the Frontier Labs,

110
00:07:13,848 --> 00:07:18,143
most notably from Daria Amode, the CEO and founder of Anthropic,

111
00:07:18,427 --> 00:07:21,478
who have asked over the last couple of months to paced the frontier.

112
00:07:21,478 --> 00:07:25,213
And he was joined by the likes of Sam Altman and Demisissibis of Google to.

113
00:07:25,592 --> 00:07:29,245
Echoed that statement saying, we need to slow things down because the model

114
00:07:29,245 --> 00:07:32,475
advancements are getting incredibly rapidly. And this is independent across

115
00:07:32,475 --> 00:07:35,158
all these different labs. They're not necessarily working together, right?

116
00:07:35,466 --> 00:07:39,055
And they want to push the messages that it's like, we should slow down.

117
00:07:39,055 --> 00:07:43,146
We should regulate ourselves. Now, a lot of the camp on the other side say,

118
00:07:43,355 --> 00:07:47,555
this is just them hyping up, fear-mongering ahead of one of the biggest IPOs ever.

119
00:07:47,555 --> 00:07:50,985
Anthropic has their one rumored to happen as soon as next week.

120
00:07:50,985 --> 00:07:53,565
But I would say that the evidence suggests otherwise. I mean,

121
00:07:53,565 --> 00:07:56,495
look at the people that have quit Anthropic, that have quit OpenAI,

122
00:07:56,495 --> 00:08:00,635
the most notable one being Jacob Coxon, one of the lead alignment researchers

123
00:08:00,635 --> 00:08:02,285
at both OpenAI and Anthropic.

124
00:08:02,285 --> 00:08:07,298
He quit Anthropic saying that neither of these labs are acting responsibly,

125
00:08:07,502 --> 00:08:10,175
that on the one hand, they say we need to pace the frontier,

126
00:08:10,175 --> 00:08:14,996
but on the other hand, they are racing as fast as they can to build these AI models.

127
00:08:15,344 --> 00:08:18,115
And that leads me to the more sinister point around all of this,

128
00:08:18,115 --> 00:08:22,695
which is the AI models that are going to be competent and capable enough of

129
00:08:22,695 --> 00:08:28,283
replacing our jobs, if this prediction plays out, aren't going to be available to you.

130
00:08:28,626 --> 00:08:31,557
We're seeing this with the most recent release of Fable 5.2.

131
00:08:31,813 --> 00:08:36,614
I see all the reviews out there, including my own, saying this is the new number one best model.

132
00:08:37,008 --> 00:08:40,166
But it's the new number one best model that is publicly available.

133
00:08:40,445 --> 00:08:44,247
If you actually look inside or under the hood of a lot of these frontier labs,

134
00:08:44,549 --> 00:08:48,305
they're gatekeeping models that are several generations ahead.

135
00:08:48,305 --> 00:08:51,715
You know, I've heard rumors from supposedly some of these researchers at these

136
00:08:51,715 --> 00:08:55,395
labs where they say that the models that are being released publicly are like

137
00:08:55,395 --> 00:08:58,950
six months behind. Now you compare that to about a year ago where.

138
00:08:59,390 --> 00:09:04,138
It was quite close. The gap wasn't that far, but now the gap is ever increasingly

139
00:09:04,138 --> 00:09:06,890
so. And there's a few reasons why this might be happening, right?

140
00:09:07,041 --> 00:09:11,378
Number one, if you are a nation state, the government that are observing these

141
00:09:11,378 --> 00:09:14,558
labs, releasing these models, you might be worried that some of these models

142
00:09:14,558 --> 00:09:19,258
are capable of breaking into your secure systems and that it might become a security asset.

143
00:09:19,258 --> 00:09:22,598
And indeed, we've seen this with the Frontier American Labs over,

144
00:09:22,813 --> 00:09:26,138
well, the entirety of this year. We had the Hugging Face incident from OpenAI.

145
00:09:26,458 --> 00:09:29,948
We had multiple hacks happening from anthropic internal models.

146
00:09:30,273 --> 00:09:33,175
So the idea here is, well, if these models are capable of doing this,

147
00:09:33,367 --> 00:09:36,358
they could pose a danger of being placed in the wrong hands.

148
00:09:36,358 --> 00:09:39,118
And so we need to keep an eye on this. So let's slow down the release.

149
00:09:39,118 --> 00:09:41,975
And the US government specifically has called for a slowdown.

150
00:09:42,382 --> 00:09:45,964
But then number two, if you're Anthropic or OpenAI, and you don't have to spend

151
00:09:45,964 --> 00:09:51,214
all your compute serving customers, you can just use this compute to train better

152
00:09:51,214 --> 00:09:54,924
AI models. And the reason why you would be doing this is to race and beat your

153
00:09:54,924 --> 00:09:57,096
competitor. Anthropic wants to beat OpenAI.

154
00:09:57,352 --> 00:10:02,158
OpenAI wants to beat Anthropic. And they're racing towards this goal of a self-building AI.

155
00:10:02,431 --> 00:10:07,614
Now, I'm sure if you were to ask both CEOs, hey, why are you racing towards this?

156
00:10:07,614 --> 00:10:10,314
Why don't we just slow down and pace the frontier? as you say,

157
00:10:10,604 --> 00:10:15,184
their simple answer would be if we slow down, our competitors won't and they

158
00:10:15,184 --> 00:10:18,174
might build a more powerful AI that is more misaligned.

159
00:10:18,395 --> 00:10:24,113
So trust us to build an AI that is super intelligent that will be aligned for humanity.

160
00:10:24,409 --> 00:10:27,914
That's the bet that they're making. But on the other hand, we're sort of racing

161
00:10:27,914 --> 00:10:31,677
towards this goal that is still quite murky.

162
00:10:31,938 --> 00:10:34,684
We don't quite know how it's going to play out or most importantly,

163
00:10:34,922 --> 00:10:36,442
how it's going to impact the rest of the world.

164
00:10:36,739 --> 00:10:42,694
And 99% of the world has no say in how this is going to impact them or why or

165
00:10:42,694 --> 00:10:44,604
when or how it should be built.

166
00:10:44,802 --> 00:10:48,754
And so I don't blame the majority of the world that I think is out there interacting

167
00:10:48,754 --> 00:10:54,112
with AI. 50% or 49% of the American public is using AI every single week, every single month.

168
00:10:54,355 --> 00:10:57,364
And they don't really feel the need to be like, ah, yeah, okay,

169
00:10:57,364 --> 00:11:00,714
this AI thing seems to be powerful. Everyone around me is saying so,

170
00:11:00,714 --> 00:11:04,938
but I don't quite see it because it's not AGI or it's not ASI for me.

171
00:11:05,199 --> 00:11:08,564
That's a very valid take because the models that actually would end up having

172
00:11:08,564 --> 00:11:10,476
that effect on them are being gatekept.

173
00:11:10,685 --> 00:11:13,605
Now, there are some good arguments against what I'm stating here.

174
00:11:13,947 --> 00:11:18,632
The main one being the data that this is based on is.

175
00:11:19,079 --> 00:11:24,836
From an independent research group, METR, which were given access to Anthropic's

176
00:11:24,836 --> 00:11:27,136
internal models, or at least a few of them.

177
00:11:27,136 --> 00:11:30,356
The argument here would be that there are more advanced models that they weren't

178
00:11:30,356 --> 00:11:33,946
given access to because since the time that they made the report to when they

179
00:11:33,946 --> 00:11:37,214
released the report, there's been new models that have been trained.

180
00:11:37,388 --> 00:11:40,876
All of this, again, is happening so quickly, but it's based on the evidence

181
00:11:40,876 --> 00:11:45,046
that they have, which is heavily redacted by Anthropic itself because they don't

182
00:11:45,046 --> 00:11:48,996
want proprietary information going to their competitors, which they state confidently.

183
00:11:49,501 --> 00:11:52,955
Anthropic has achieved 2x research and development progress,

184
00:11:53,135 --> 00:11:57,866
which would insinuate that the bar of acceleration is now gone vertical.

185
00:11:58,139 --> 00:12:01,559
If they are wrong, or if there are any kinks in their analysis,

186
00:12:01,848 --> 00:12:04,257
this could mean that the prediction could be wrong.

187
00:12:04,444 --> 00:12:08,895
But if you talk to any of the Anthropic researchers, their answer isn't that, yeah, they're right.

188
00:12:09,139 --> 00:12:12,884
The answer is it actually feels kind of small, that estimate,

189
00:12:13,157 --> 00:12:16,505
that our feeling is that it's actually 12x faster.

190
00:12:16,505 --> 00:12:19,625
So a lot of this is dependent on data that isn't publicly available.

191
00:12:19,625 --> 00:12:24,885
And I would implore us to all just wait until we have that data available before

192
00:12:24,885 --> 00:12:26,838
we make a concrete kind of assumption here.

193
00:12:27,095 --> 00:12:33,433
The second thing is AI research and development isn't just based on the software itself.

194
00:12:33,736 --> 00:12:37,715
It's all well and good having an AI model that can instruct itself to build

195
00:12:37,715 --> 00:12:41,325
a better version of itself, but it's still limited by physical constraints.

196
00:12:41,325 --> 00:12:46,235
Think about the chips, the electricity, the power, the data centers that are

197
00:12:46,235 --> 00:12:49,805
now arguably going to be slowed down from the recent electoral results that

198
00:12:49,805 --> 00:12:53,375
are required to keep feeding the beast, right?

199
00:12:53,375 --> 00:12:57,955
And I guess that that's why we've seen a lot of pushback from a certain portion

200
00:12:57,955 --> 00:13:02,114
of the public, which say that AI data centers aren't going to be good for us and the community.

201
00:13:02,329 --> 00:13:06,235
And two, the opinion that AI is going to cause more harm than it is good.

202
00:13:06,433 --> 00:13:10,235
I can see the arguments festering there and that a rising movement would kind

203
00:13:10,235 --> 00:13:11,681
of push back against these things.

204
00:13:11,948 --> 00:13:16,675
But regardless of how it plays out, physical atoms might constrain this whole thing.

205
00:13:16,675 --> 00:13:20,155
We might not be able to get to those two milestones of an automated AI researcher

206
00:13:20,155 --> 00:13:23,778
or an AI that's good enough to replace every human expert and white collar work

207
00:13:24,080 --> 00:13:27,563
because we don't have the chips or electricity to fund that boom.

208
00:13:27,778 --> 00:13:32,685
Okay, so the big question is, how much emphasis should we place on this prediction?

209
00:13:32,685 --> 00:13:36,595
They've gotten pretty much every single thing right so far when it came to the

210
00:13:36,595 --> 00:13:38,865
CapEx boom, how much money people were going to be spending,

211
00:13:39,185 --> 00:13:42,603
all the semiconductor stocks and everything doing as well as they have.

212
00:13:43,289 --> 00:13:47,104
How good is their bet that we are going to get these two milestones over the

213
00:13:47,104 --> 00:13:50,441
next year? Well, I have some thoughts on this segment brought to you by Qualcomm.

214
00:13:54,394 --> 00:13:58,264
Now, the first one being is there's evidence being presented right in front

215
00:13:58,264 --> 00:14:03,299
of our faces that prove that the models underneath the hood are capable of doing crazy things.

216
00:14:03,642 --> 00:14:08,534
Let's take OpenAI as one main example. Their internal models solved Navier Stokes,

217
00:14:08,534 --> 00:14:12,415
which is one of the hardest mathematical problems, a few weeks or month ago.

218
00:14:12,717 --> 00:14:17,384
And then it went on to subsequently solve another 400 to 500 more mathematical problems.

219
00:14:17,384 --> 00:14:21,454
Now, the reaction to that news was negative, especially from the mathematical

220
00:14:21,454 --> 00:14:25,584
society of experts that dedicated their entire lives. And now they're like,

221
00:14:25,584 --> 00:14:29,954
well, what do we do now? Then you peek over at the fence at their competitors,

222
00:14:29,954 --> 00:14:32,070
Anthropic, who just released Fable 5.2.

223
00:14:32,395 --> 00:14:36,504
And they casually mentioned that it was able to prove out the Fermat theorem,

224
00:14:36,504 --> 00:14:39,884
which is something that is incredibly hard to prove out and would take humans

225
00:14:40,203 --> 00:14:42,885
so long and so much money to end up doing.

226
00:14:43,251 --> 00:14:47,934
And AI just ended up doing it. Oh, and by the way, we also solved two open physics

227
00:14:47,934 --> 00:14:49,004
problems that were out there.

228
00:14:49,195 --> 00:14:51,554
They just casually dropped this in a blog post announcement.

229
00:14:51,554 --> 00:14:54,744
And then you look at the news that Google dropped a few days ago that their

230
00:14:54,744 --> 00:14:59,864
latest model, Gemini Carbon, solved or proved the encryption method behind Signal,

231
00:14:59,864 --> 00:15:03,924
one of the most widely used encrypted apps out there for messaging.

232
00:15:03,924 --> 00:15:07,184
And all of this sounds too scientific, right? It sounds like,

233
00:15:07,184 --> 00:15:09,772
oh, we shouldn't be paying attention to this because like this doesn't apply to me.

234
00:15:10,079 --> 00:15:15,384
But the point is, AI models that aren't available to you are solving incredibly

235
00:15:15,384 --> 00:15:20,154
hard problems that humans couldn't conceivably have thought of solving or have

236
00:15:20,154 --> 00:15:23,210
been trying to solve for many centuries and couldn't figure out.

237
00:15:23,327 --> 00:15:24,923
Now you have an AI that's capable of doing it.

238
00:15:25,161 --> 00:15:28,992
So the leap of going from, okay, if an AI model can do that,

239
00:15:29,277 --> 00:15:31,314
why can't it solve some of the crazy

240
00:15:31,694 --> 00:15:36,184
hard work problems that I have in my white collar job doesn't seem too far of

241
00:15:36,184 --> 00:15:40,694
a leap right now my other take is that if you are open ai or anthropic with

242
00:15:40,694 --> 00:15:42,204
these powerful internal models.

243
00:15:42,681 --> 00:15:48,372
You might want to gatekeep them because you could use these powerful models to self-serve.

244
00:15:48,372 --> 00:15:52,372
In other words, maybe point the model at solving something like cancer or solving

245
00:15:52,372 --> 00:15:55,382
another crazy thing and then capturing that market.

246
00:15:55,382 --> 00:16:00,432
A good example most recently is Claude or Anthropic releasing Claude Office,

247
00:16:00,432 --> 00:16:05,642
which is basically a replacement for Microsoft Office and Google G Suite tools

248
00:16:05,642 --> 00:16:09,272
like Docs, Spreadsheets, et cetera, that you can now go onto Claude.

249
00:16:09,272 --> 00:16:12,871
It has all the context. you can give it your tasks or whatever the goal is,

250
00:16:13,120 --> 00:16:15,941
and you have a bunch of AI agents that just go out and do that, right?

251
00:16:16,232 --> 00:16:20,072
The idea here being is the automation of white-collar work is already starting

252
00:16:20,072 --> 00:16:23,105
and it's happening right in front of us for a $20 a month subscription.

253
00:16:23,366 --> 00:16:27,602
So in this world, the concentration of power goes to these very few labs.

254
00:16:27,602 --> 00:16:32,692
And I guess I don't really know how to comprehend what would happen if labs

255
00:16:32,692 --> 00:16:36,073
are able to control and consume any industry that's out there.

256
00:16:36,311 --> 00:16:39,572
Then I look at the skeptic side and I'm like, there's some really good points

257
00:16:39,572 --> 00:16:42,572
that can be argued here as well, which is, okay, if these models,

258
00:16:42,572 --> 00:16:45,233
even the models that are publicly available that aren't internal.

259
00:16:45,640 --> 00:16:50,442
Are still really powerful and good, why aren't we seeing more transformational stories?

260
00:16:50,442 --> 00:16:54,402
And listen, there are a few stories out there where we see single founders that

261
00:16:54,402 --> 00:16:59,072
are building 10 to $100 million worth of ARR, and we're seeing people discover

262
00:16:59,072 --> 00:17:03,242
ancient civilization sites using available public data from NASA.

263
00:17:03,242 --> 00:17:04,963
That was a cool story that I saw recently.

264
00:17:05,149 --> 00:17:08,892
So all of these minute things are starting to happen. But for the majority of

265
00:17:08,892 --> 00:17:14,042
the world that are trying to use AI for impactful work, we're not quite seeing that change just yet.

266
00:17:14,042 --> 00:17:17,508
And there was a report from Apollo Global actually a few weeks ago,

267
00:17:17,734 --> 00:17:21,171
which stated that outside of the top seven companies, which by the way,

268
00:17:21,304 --> 00:17:25,832
are all AI hyperscalers investing pretty heavily in building AI chips and AI models,

269
00:17:26,239 --> 00:17:29,681
none of the other Fortune 500 or Fortune, what is that?

270
00:17:31,188 --> 00:17:36,648
493 are seeing the ROI yet from using AI. So that's good evidence to suggest

271
00:17:36,648 --> 00:17:40,142
that maybe what this report is stating isn't necessarily true.

272
00:17:40,322 --> 00:17:44,348
And maybe it's just being overhyped ahead of one of the most valuable IPOs that

273
00:17:44,348 --> 00:17:46,370
we're going to be seeing over the next couple of weeks.

274
00:17:46,602 --> 00:17:52,785
So where I guess I land is that directionally, I think that AI model development

275
00:17:52,965 --> 00:17:55,798
is accelerating. I want to make that very clear.

276
00:17:56,117 --> 00:17:59,748
Now, how soon we're going to have an AI powerful enough to replace all of our

277
00:17:59,748 --> 00:18:02,733
jobs, to me is still a question mark.

278
00:18:03,122 --> 00:18:08,468
I think it's going to happen sooner than we expect, but maybe not as soon as the next 12 to 18 months.

279
00:18:08,468 --> 00:18:13,018
But it's undoubtedly coming and it's something that we should be aware of.

280
00:18:13,018 --> 00:18:17,788
That's, I guess, the whole idea or goal of this episode or why I felt so compelled

281
00:18:17,788 --> 00:18:21,088
to talk about this is that I feel like the majority of people around me that

282
00:18:21,088 --> 00:18:24,978
aren't in my little echo chamber, I guess, or in my little tech world,

283
00:18:25,564 --> 00:18:29,728
don't seem to care or aren't really aware or were at worst, don't believe that

284
00:18:29,728 --> 00:18:32,025
AMLs have improved significantly over the last year.

285
00:18:32,298 --> 00:18:35,778
And that's just undoubtedly true. So I hope you enjoyed this episode.

286
00:18:35,778 --> 00:18:38,689
I hope it wasn't too much of a doom mongering episode.

287
00:18:38,875 --> 00:18:41,772
As you know, I am like one of the most bullish, optimistic.

288
00:18:42,404 --> 00:18:45,689
Guys out there on this industry. And I would actually say that,

289
00:18:46,136 --> 00:18:48,859
if we do end up achieving models capable of

290
00:18:49,195 --> 00:18:52,371
automating a large amount of white collar work that exists today,

291
00:18:52,702 --> 00:18:56,072
I actually think that's going to result in a much more optimistic outcome for

292
00:18:56,072 --> 00:19:00,092
the individual of whose job or role it's going to be replaced because they can

293
00:19:00,092 --> 00:19:01,472
end up having more free time

294
00:19:01,812 --> 00:19:06,252
or working at a higher level of intelligence to go out and do way more things,

295
00:19:06,252 --> 00:19:09,922
whether that's entrepreneurial things or just simply being better at their job,

296
00:19:09,922 --> 00:19:11,302
being more productive at their job.

297
00:19:11,302 --> 00:19:13,782
So yeah that's the episode i hope you enjoyed it if you're listening to this

298
00:19:13,782 --> 00:19:17,182
on youtube spotify or wherever you're listening to it please give us a rating

299
00:19:17,182 --> 00:19:21,209
please subscribe please leave us a comment turn on notifications it helps us out massively i love

300
00:19:21,544 --> 00:19:25,132
reading all your comments also i'm linking to my x profile where i post pretty

301
00:19:25,132 --> 00:19:29,684
regular daily multiple daily updates so you can can respond or dm to me there

302
00:19:30,027 --> 00:19:33,002
and yeah please share this episode with a friend if you found it interesting

303
00:19:33,002 --> 00:19:36,143
it helps us out massively and yeah i'll see you on the next one brought to you by qualcomm.