1
00:00:00,000 --> 00:00:02,717
the Europeans, I mean, they have
a really different mindset for all this

2
00:00:02,800 --> 00:00:07,277
stuff. When when they talk about AI
leadership, what they mean is that

3
00:00:07,360 --> 00:00:10,957
they're taking the lead in defining the
regulations. You know, they get together

4
00:00:11,040 --> 00:00:13,917
in Brussels and figure out what all
the rules should be. And that's what they

5
00:00:14,000 --> 00:00:15,757
call leadership.
>> It's almost like a game show or

6
00:00:15,840 --> 00:00:18,157
something. They do everything they can
to strangle them in their crib and then

7
00:00:18,240 --> 00:00:21,437
if they if they make it if they make
[laughter] it through like a decade of

8
00:00:21,520 --> 00:00:23,677
abuse of small companies, then they're
going to give them the money to grow.

9
00:00:23,760 --> 00:00:26,957
>> Ronald Reagan had a line about this,
which is if it moves, tax it. If it

10
00:00:27,040 --> 00:00:30,557
keeps moving, regulate it.
If it stops moving, subsidize it.

11
00:00:30,640 --> 00:00:36,040
>> Yeah. >> Uh the Europeans are
definitely at the subsidize it stage.

12
00:00:37,120 --> 00:00:39,037
>> David, welcome to the Asz podcast.

13
00:00:39,120 --> 00:00:40,957
Thanks for joining.
>> Yeah, good to be here.

14
00:00:41,040 --> 00:00:43,197
>> So, David, you're the AI and cryptozar.

15
00:00:43,280 --> 00:00:45,917
Why don't you first talk about why
it makes sense to have those as a

16
00:00:46,000 --> 00:00:48,797
portfolio? What do they have to do with
each other? And then I'll have you lay

17
00:00:48,880 --> 00:00:51,677
out what's the what's the what's the Trump
plan on those those two categories

18
00:00:51,760 --> 00:00:55,277
and how we're doing. >> Well,
there are two technologies that I

19
00:00:55,360 --> 00:00:58,317
guess are relatively new. And so there's

20
00:00:58,400 --> 00:01:03,517
a lot of fear of them and uh I think
people don't actually know that much

21
00:01:03,600 --> 00:01:07,597
about them. They don't really know what
to make of them. I think that from a

22
00:01:07,680 --> 00:01:10,317
policy standpoint and we can talk about
the similarities and differences. The

23
00:01:10,400 --> 00:01:11,917
the approaches are a little different.

24
00:01:12,000 --> 00:01:15,677
Um I think with crypto, the main thing

25
00:01:15,760 --> 00:01:18,877
that's needed is regulatory certainty.

26
00:01:18,960 --> 00:01:22,877
You know, all the entrepreneurs I've
talked to over the years, they they all

27
00:01:22,960 --> 00:01:25,837
say the same thing, which is just tell
us what the rules are. were happy to

28
00:01:25,920 --> 00:01:29,277
comply, but people, you know, Washington
won't tell us what they are. And in

29
00:01:29,360 --> 00:01:31,677
fact, during the Biden years, you had um

30
00:01:31,760 --> 00:01:35,757
an SEC chairman who uh took an approach,

31
00:01:35,840 --> 00:01:40,397
which I guess has been called um
regulation through uh enforcement, which

32
00:01:40,480 --> 00:01:41,917
basically means you just get prosecuted.

33
00:01:42,000 --> 00:01:43,037
They don't tell you what the rules are.

34
00:01:43,120 --> 00:01:45,197
You just basically get indicted and then

35
00:01:45,280 --> 00:01:47,357
everyone else is supposed to divine what

36
00:01:47,440 --> 00:01:50,637
the rules are um as you get prosecuted

37
00:01:50,720 --> 00:01:55,517
and fined and imprisoned. So, that was
the approach for several years. and um

38
00:01:55,600 --> 00:01:59,437
you know and and and there and as
a result of that basically the whole

39
00:01:59,520 --> 00:02:01,597
crypto industry was in the process of

40
00:02:01,680 --> 00:02:04,957
moving offshore and uh and America I

41
00:02:05,040 --> 00:02:07,037
think was being deprived of this

42
00:02:07,120 --> 00:02:09,757
industry of the future. And so President

43
00:02:09,840 --> 00:02:13,517
Trump during his campaign and last year

44
00:02:13,600 --> 00:02:17,917
he gave a now famous speech in Nashville
in which he declared that he would make

45
00:02:18,000 --> 00:02:21,357
the United States the crypto capital
of the planet and that he would fire

46
00:02:21,440 --> 00:02:24,637
Gendler. That was like the big applause
line. I think he He's talked about I

47
00:02:24,720 --> 00:02:27,917
applauded [laughter] >> he's
talked about how surprised he was

48
00:02:28,000 --> 00:02:32,797
at you know what a big uh you know
ovation he got at that. So he said it

49
00:02:32,880 --> 00:02:35,917
again and you know the crowd erupted
again but in any event he promised

50
00:02:36,000 --> 00:02:41,277
basically to provide this clarity so
that um you know the the industry would

51
00:02:41,360 --> 00:02:45,917
um understand what the rules are be able
to comply in turn that gives that should

52
00:02:46,000 --> 00:02:48,077
provide uh greater protection for

53
00:02:48,160 --> 00:02:51,037
consumers and um businesses everyone

54
00:02:51,120 --> 00:02:55,437
who's part of the ecosystem. uh and it
makes America more competitive. So I

55
00:02:55,520 --> 00:02:57,277
think that's sort of the the the the

56
00:02:57,360 --> 00:03:01,357
mandate on on crypto is uh it in a way

57
00:03:01,440 --> 00:03:04,157
it's pro- regulation. It's basically we

58
00:03:04,240 --> 00:03:09,517
we want to put in place regulations
in a in a way AI is kind of the opposite

59
00:03:09,600 --> 00:03:13,437
where I think the B administration was
too heavy-handed. They were starting to

60
00:03:13,520 --> 00:03:17,517
really regulate this area without even
understanding what it was. they weren't

61
00:03:17,600 --> 00:03:21,677
they had no one had really taken
the time to understand uh how AI was even

62
00:03:21,760 --> 00:03:24,397
being used what the real u dangers were

63
00:03:24,480 --> 00:03:26,877
there was this intense fear-mongering

64
00:03:26,960 --> 00:03:29,437
and as a result of that you know the

65
00:03:29,520 --> 00:03:33,917
approach of the b administration
was uh they were in the process of of

66
00:03:34,000 --> 00:03:37,677
implementing very heavy-handed
regulations on both the software and

67
00:03:37,760 --> 00:03:40,717
hardware side and we can drill into that

68
00:03:40,800 --> 00:03:45,037
I think that with the Trump
administration um the approach has been

69
00:03:45,120 --> 00:03:49,437
that we want the United States
to win the AI race. Um, it's a global

70
00:03:49,520 --> 00:03:53,917
competition. Um, sometimes we mention
the fact that China is our probably our

71
00:03:54,000 --> 00:03:57,517
main competitor in this area. They're
the only other country that has the

72
00:03:57,600 --> 00:04:00,717
technological capability, the talent,

73
00:04:00,800 --> 00:04:04,157
the knowhow, the um expertise to to beat

74
00:04:04,240 --> 00:04:08,717
us in this area. And we want to make
sure the United States wins. Um, and of

75
00:04:08,800 --> 00:04:12,237
course in the US it's not really
the government that's responsible for

76
00:04:12,320 --> 00:04:15,837
innovation. It's the private sector.
So that means that our companies have to

77
00:04:15,920 --> 00:04:19,757
win and if you're imposing all sorts
of crazy um burdensome regulation on them

78
00:04:19,840 --> 00:04:21,597
then that's going to hurt not not help.

79
00:04:21,680 --> 00:04:24,717
So um the president gave a I think very

80
00:04:24,800 --> 00:04:29,357
important AI policy speech a couple
months ago on July 23rd where he

81
00:04:29,440 --> 00:04:33,917
declared in no uncertain terms that we
had to win the AI race and he laid out

82
00:04:34,000 --> 00:04:36,157
several pillars on on how we do that. it

83
00:04:36,240 --> 00:04:40,957
was pro- innovation, pro infrastructure,
which also means uh pro energy and pro

84
00:04:41,040 --> 00:04:44,477
export. And we can drill into all those
things if you want, but that was sort of

85
00:04:44,560 --> 00:04:47,117
the the high line. And so I think that,

86
00:04:47,200 --> 00:04:52,317
you know, again with with AI, um
the idea is kind of like how do we unleash

87
00:04:52,400 --> 00:04:55,437
innovation? And I think with crypto,
it's been more about how do we create

88
00:04:55,520 --> 00:05:00,877
regulatory certainty. Um but you know,
in terms of my role, like what you know,

89
00:05:00,960 --> 00:05:04,077
why why am I doing both? I mean, I think
the common denominator is just again,

90
00:05:04,160 --> 00:05:09,357
these are new technologies. They um,
you know, they're both um, you know,

91
00:05:09,440 --> 00:05:11,837
obviously come from from the tech
industry, which has a very different

92
00:05:11,920 --> 00:05:16,957
culture than Washington does. And I kind
of see it as my role to help be a bridge

93
00:05:17,040 --> 00:05:19,437
between what's happening in Silicon
Valley and what's happening in

94
00:05:19,520 --> 00:05:23,277
Washington and helping Washington

95
00:05:23,360 --> 00:05:25,837
understand not just the policy that's

96
00:05:25,920 --> 00:05:30,317
needed or or the innovation that's
happening but also kind of culturally,

97
00:05:30,400 --> 00:05:32,237
you know, what what makes the tech

98
00:05:32,320 --> 00:05:33,837
industry different and special and and

99
00:05:33,920 --> 00:05:37,437
how that needs to be protected um from

100
00:05:37,520 --> 00:05:41,517
uh government doing something uh
excessively heavy-handed. So, um, so

101
00:05:41,600 --> 00:05:44,637
David, you know, we're going to talk
a lot about AI today, but just on crypto,

102
00:05:44,720 --> 00:05:47,597
I've I've had this interesting
experience this year kind of after the,

103
00:05:47,680 --> 00:05:50,557
you know, election, you know, kind
of people adjusted to the change, uh, of

104
00:05:50,640 --> 00:05:53,837
government. And I've had this discussion
with, you know, a number of people who,

105
00:05:53,920 --> 00:05:57,677
let's say, in in politics who were
previously anti-crypto, who have been,

106
00:05:57,760 --> 00:06:00,637
you know, trying to figure out how to kind
of get to a more sensible position.

107
00:06:00,720 --> 00:06:03,437
Um, and then also actually people in the
in the financial services industry who

108
00:06:03,520 --> 00:06:06,397
kind of followed it from a distance,
um, and maybe, you know, participated in

109
00:06:06,480 --> 00:06:08,957
the, you know, the various debunking
things without really understanding what

110
00:06:09,040 --> 00:06:12,717
was happening. Uh but the the common
denominator has been they're like you

111
00:06:12,800 --> 00:06:17,437
know Mark I didn't really understand um
how bad it was. Um, I basically thought

112
00:06:17,520 --> 00:06:21,757
you guys in tech were basically
just whining a lot. Um, and you know,

113
00:06:21,840 --> 00:06:25,437
pleading as a special interest, um,
and kind of doing the normal thing. And I

114
00:06:25,520 --> 00:06:28,317
figured, you know, the the, you know,
the horror stories were kind of made up,

115
00:06:28,400 --> 00:06:31,437
you know, of people getting prosecuted
and, you know, entrepreneurs getting

116
00:06:31,520 --> 00:06:34,717
their houses raided by the FBI and,
you know, like the whole paniply of things

117
00:06:34,800 --> 00:06:38,797
that, you know, that happened and like
I didn't really I I now in retrospect now

118
00:06:38,880 --> 00:06:41,517
that I now that I go back and look, I'm
like, "Oh my god, this was actually much

119
00:06:41,600 --> 00:06:45,677
worse than I thought." Um, what do you
do you do you have that experience? And

120
00:06:45,760 --> 00:06:48,637
as as as you're in there and kind
of as you now have like a complete view of

121
00:06:48,720 --> 00:06:51,517
everything that happens, like do do you
think people understand actually how bad

122
00:06:51,600 --> 00:06:54,317
it was? >> I mean,
I think it's a a great point. I

123
00:06:54,400 --> 00:06:58,077
mean, I didn't really know either.
You kind of heard um generally I mean, we

124
00:06:58,160 --> 00:07:01,757
knew that there was debanking going
on and and by the way, it wasn't just

125
00:07:01,840 --> 00:07:03,997
crypto companies that were being debanked,
but their founders were being

126
00:07:04,080 --> 00:07:08,317
debanked personally. So, if you were
the founder of a crypto company, you

127
00:07:08,400 --> 00:07:11,277
couldn't open a bank account. I mean,
that's a huge problem. It's like how do

128
00:07:11,360 --> 00:07:16,157
you how do you transact? How do you make
payments? um how do you pay people? I

129
00:07:16,240 --> 00:07:19,197
mean it basically deprivives you of a
livelihood. It's a very extreme form of

130
00:07:19,280 --> 00:07:22,157
censorship. So that was definitely
happening. And then of course you have

131
00:07:22,240 --> 00:07:25,357
all the the prosecutions uh that the SEC

132
00:07:25,440 --> 00:07:27,277
was behind. So yeah, it was it was

133
00:07:27,360 --> 00:07:30,077
really bad. I remember um back in I

134
00:07:30,160 --> 00:07:32,477
think it was in March uh we had a crypto

135
00:07:32,560 --> 00:07:35,117
summit at the White House and um one of

136
00:07:35,200 --> 00:07:39,837
the attendees said that you know a year
ago uh I would have thought it was more

137
00:07:39,920 --> 00:07:44,317
likely that I'd be in jail than that I'd
be at the White House. Um and so it was

138
00:07:44,400 --> 00:07:47,757
you know it was a really big milestone
for for the industry. They had never

139
00:07:47,840 --> 00:07:52,797
ever received any kind of recognition
like that. The idea that this was even a

140
00:07:52,880 --> 00:07:54,637
industry that you would do an event at

141
00:07:54,720 --> 00:07:57,837
the White House. I mean um at a minimum

142
00:07:57,920 --> 00:08:00,237
I think crypto is seen as very day

143
00:08:00,320 --> 00:08:03,677
class. Um but uh but in any event yeah

144
00:08:03,760 --> 00:08:05,277
no it's been a it's been a huge shift. I

145
00:08:05,360 --> 00:08:07,437
mean, we basically have have um have

146
00:08:07,520 --> 00:08:12,877
stopped that. And um and you know,
it was very unfair because again, these

147
00:08:12,960 --> 00:08:16,717
founders wanted to comply with the rules,
but they weren't told what they

148
00:08:16,800 --> 00:08:21,117
were. Um and that was all part of a
deliberate strategy, I think, to drive

149
00:08:21,200 --> 00:08:22,957
crypto offshore.

150
00:08:23,040 --> 00:08:26,717
>> Yeah. One of the things that um is very

151
00:08:26,800 --> 00:08:30,317
different between crypto and AI
that we've noticed is that on the crypto

152
00:08:30,400 --> 00:08:32,637
front, everybody just wanted rules. um

153
00:08:32,720 --> 00:08:35,837
and the industry was relatively unified.

154
00:08:35,920 --> 00:08:39,197
Whereas in AI, we've seen very like

155
00:08:39,280 --> 00:08:44,077
interesting kind of calls coming
from inside the house with uh certain

156
00:08:44,160 --> 00:08:47,757
companies really going for regulatory
capture. You know, people who have early

157
00:08:47,840 --> 00:08:51,997
leads saying let's cut off all new
companies from developing AI and so

158
00:08:52,080 --> 00:08:54,317
forth. Um what do you make of that and

159
00:08:54,400 --> 00:08:56,477
and where do you think that's going?

160
00:08:56,560 --> 00:08:59,037
>> I think it's I think it's a very big

161
00:08:59,120 --> 00:09:03,517
problem. Um I actually recently
criticized uh one of our AI model

162
00:09:03,600 --> 00:09:06,557
companies for engaging
in regulatory capture strategy.

163
00:09:06,640 --> 00:09:09,757
>> Yes. >> Um very fair
criticism by [laughter] the way.

164
00:09:09,840 --> 00:09:12,797
>> It is very fair
and actually um you know

165
00:09:12,880 --> 00:09:16,797
of course they denied it and then
uh should I tell the story? I mean

166
00:09:16,880 --> 00:09:19,117
>> yeah sure rarely do you get vindicated

167
00:09:19,200 --> 00:09:24,477
on on X so so thoroughly and completely
as I did on this because after this

168
00:09:24,560 --> 00:09:28,477
company it was basically anthropic after
they denied it. There was um I mean what

169
00:09:28,560 --> 00:09:31,597
basically happened is that Jack Clark
who's a co-founder and head of policy

170
00:09:31,680 --> 00:09:32,717
philanthropic

171
00:09:32,800 --> 00:09:35,837
>> gave a speech at a
a conference where he

172
00:09:35,920 --> 00:09:40,317
said that uh he compared fear of AI to a

173
00:09:40,400 --> 00:09:45,037
child um you know seeing monsters
in the dark or or or thinking there were

174
00:09:45,120 --> 00:09:48,237
monsters in the dark and but then you
turn the lights on and the monsters are

175
00:09:48,320 --> 00:09:51,357
there. I thought that was such a um

176
00:09:51,440 --> 00:09:56,317
ridiculous analogy. I mean it's
basically purile. I mean, it's it's so

177
00:09:56,400 --> 00:09:59,117
childish just to be almost
self-indicting because you're basically

178
00:09:59,200 --> 00:10:02,157
admitting the fear is made up, not real.

179
00:10:02,240 --> 00:10:03,837
>> Uh, in any event,
so I said, "Well, this

180
00:10:03,920 --> 00:10:06,797
is like fear-mongering and part of their
regulatory capture strategy." And of

181
00:10:06,880 --> 00:10:11,277
course, they denied it. But then um
a lawyer who was in the crowd at his

182
00:10:11,360 --> 00:10:14,797
speech said, "Well, yeah, but Jack's not
telling you what he said during the Q&A,

183
00:10:14,880 --> 00:10:19,757
which he basically admitted that
everything that Anthropic was doing was

184
00:10:19,840 --> 00:10:24,077
uh you know, with with like things
like SB53, which is supposedly just

185
00:10:24,160 --> 00:10:27,677
implementing transparency." He said,
"No, that's all that he admitted that

186
00:10:27,760 --> 00:10:29,917
was just a stepping
stone to their real goal,

187
00:10:30,000 --> 00:10:32,637
>> which was to get an approval, a system

188
00:10:32,720 --> 00:10:34,957
of pre-approvals in Washington before

189
00:10:35,040 --> 00:10:37,837
you can release new models." and he

190
00:10:37,920 --> 00:10:42,077
admitted as part of the Q&A that making
people very afraid was part of their

191
00:10:42,160 --> 00:10:44,237
strategy. So again, just as much as a

192
00:10:44,320 --> 00:10:47,037
smoking gun as you could ever get um in

193
00:10:47,120 --> 00:10:49,837
a in a spaton X. But the reason why I

194
00:10:49,920 --> 00:10:53,037
think that that approach is so damaging

195
00:10:53,120 --> 00:10:56,317
is that the thing that's uh really made

196
00:10:56,400 --> 00:11:00,317
I think Silicon Valley special over
the past several decades is permissionless

197
00:11:00,400 --> 00:11:03,837
innovation. Right?
It's that two guys in a garage

198
00:11:03,920 --> 00:11:07,437
>> can just pursue their idea. you know,
maybe they raise some capital from

199
00:11:07,520 --> 00:11:11,837
angels or VC firms. B basically people
who are willing to lose all of their

200
00:11:11,920 --> 00:11:15,677
money and these are people who,
you know, are young founders. They could

201
00:11:15,760 --> 00:11:18,637
also be, you know, it could be the the
the dropout, the future dropout in a

202
00:11:18,720 --> 00:11:23,037
dorm room and they're able just
to pursue their idea. And um and the only

203
00:11:23,120 --> 00:11:26,877
reason that I think has happened
in Silicon Valley whereas you look at in

204
00:11:26,960 --> 00:11:29,757
you know industries like um I don't know

205
00:11:29,840 --> 00:11:34,397
like pharma or um you know healthcare
or defense or banking or these highly

206
00:11:34,480 --> 00:11:37,837
regulated industries where you just
don't see a lot of startups is because

207
00:11:37,920 --> 00:11:40,397
they're all heavily regulated which
means you have to go to Washington to

208
00:11:40,480 --> 00:11:44,717
get permission to do things. And the
thing I've seen in Washington is just

209
00:11:44,800 --> 00:11:47,037
that, you know, the approvals get set up

210
00:11:47,120 --> 00:11:50,317
for reasons, but those reasons very

211
00:11:50,400 --> 00:11:54,557
quickly stop mattering. And it just
matters like how good your government

212
00:11:54,640 --> 00:11:57,997
affairs team is at navigating through
the bureaucracy and figuring out how to

213
00:11:58,080 --> 00:12:01,757
get those approvals. And it's not
something that your typical startup

214
00:12:01,840 --> 00:12:05,197
founders are going to be good at. Um,
it's something that big companies get

215
00:12:05,280 --> 00:12:07,517
good at because they've got
the resources and that's exactly what

216
00:12:07,600 --> 00:12:09,997
regulatory capture means. So the the

217
00:12:10,080 --> 00:12:12,397
whole basis of Silicon Valley success,

218
00:12:12,480 --> 00:12:15,757
the reason why it's really the the crown

219
00:12:15,840 --> 00:12:20,237
jewel of the American economy and the
envy of the rest of the world. We see

220
00:12:20,320 --> 00:12:23,597
all these attempts by all these other
countries to create their own Silicon

221
00:12:23,680 --> 00:12:27,277
Valley. The the reason that's the case
is because of permissionless innovation.

222
00:12:27,360 --> 00:12:29,997
And what is being contemplated and

223
00:12:30,080 --> 00:12:34,397
discussed and implemented with respect
to AI is an approval system for both

224
00:12:34,480 --> 00:12:38,477
software and hardware. Uh and this is
not theoretical. This has already been

225
00:12:38,560 --> 00:12:42,717
happening on the the hardware side.
One of the last things that the Biden

226
00:12:42,800 --> 00:12:47,357
administration did the last week
the Biden administration was um impose the

227
00:12:47,440 --> 00:12:49,437
so-called Biden diffusion rule which

228
00:12:49,520 --> 00:12:53,917
requires that every sale of a GPU on

229
00:12:54,000 --> 00:12:56,877
Earth be licensed by the government

230
00:12:56,960 --> 00:12:59,197
which is to say preapproved um unless it

231
00:12:59,280 --> 00:13:01,837
fits into some category of exception but

232
00:13:01,920 --> 00:13:04,237
basically the overall idea is that you

233
00:13:04,320 --> 00:13:06,877
know uh compute is now going to be um a

234
00:13:06,960 --> 00:13:09,197
licensed and and pre-approved um

235
00:13:09,280 --> 00:13:12,237
category. uh we rescended that. Um and

236
00:13:12,320 --> 00:13:14,077
then on the the software side like I

237
00:13:14,160 --> 00:13:16,717
said I mean the goal very clearly is to

238
00:13:16,800 --> 00:13:19,597
start with these reporting requirements

239
00:13:19,680 --> 00:13:22,317
to the government to the states and then

240
00:13:22,400 --> 00:13:26,797
where that ramps up to is you have
to go to Washington to get permission uh

241
00:13:26,880 --> 00:13:29,117
before you release a new model and um

242
00:13:29,200 --> 00:13:30,957
you know it's this would drastically

243
00:13:31,040 --> 00:13:33,997
slow down innovation and make America

244
00:13:34,080 --> 00:13:38,557
less competitive. I mean, you know,
these approvals can take months, they

245
00:13:38,640 --> 00:13:42,477
can take years. Uh, when models are when
when a new chip is released every year

246
00:13:42,560 --> 00:13:46,237
and we have licenses that have been
sitting in the hopper for 2 years, I

247
00:13:46,320 --> 00:13:50,397
mean, the requests are obsolete by the
time they finally get approved. And um,

248
00:13:50,480 --> 00:13:54,397
that would be even more true with models
where, you know, the the cycle time is,

249
00:13:54,480 --> 00:13:57,917
you know, like three three or four
months for a new model. I mean, you

250
00:13:58,000 --> 00:14:01,757
know, and and and what exactly is a
bureaucracy in Washington going to know

251
00:14:01,840 --> 00:14:05,357
about uh this technology that, you know,
that they're that they're going to be in

252
00:14:05,440 --> 00:14:09,037
a good position to to approve it.
Any event, but this is what is being

253
00:14:09,120 --> 00:14:12,477
contemplated right now and I think
it would be a disaster for for Silicon

254
00:14:12,560 --> 00:14:14,877
Valley, but also and for innovation, but

255
00:14:14,960 --> 00:14:16,797
therefore for American competitiveness

256
00:14:16,880 --> 00:14:20,717
and I think we will lose the AI race uh

257
00:14:20,800 --> 00:14:24,637
to countries like China if you know if
this is if this is the set of rules that

258
00:14:24,720 --> 00:14:27,277
we have. >> Yeah.
One of the really diabolical

259
00:14:27,360 --> 00:14:29,757
things about their argument is if they

260
00:14:29,840 --> 00:14:34,717
really believed there was a monster then
why are they buying GPUs like at a rate

261
00:14:34,800 --> 00:14:37,997
faster than anybody? And then the other
thing that we know from being in the

262
00:14:38,080 --> 00:14:41,117
industry is their reputation is the

263
00:14:41,200 --> 00:14:45,037
literally the worst security practices
in the entire industry with respect to

264
00:14:45,120 --> 00:14:48,397
their own code. So if you were building
this monster, the last thing you'd want

265
00:14:48,480 --> 00:14:52,637
to do is like leave a bunch of holes
around for people to hack it. Um so they

266
00:14:52,720 --> 00:14:53,917
don't believe anything they're saying.

267
00:14:54,000 --> 00:14:56,637
It's like completely made up to try and

268
00:14:56,720 --> 00:14:58,957
maintain their lead by the same

269
00:14:59,040 --> 00:15:01,597
psychotic. I think there is um I think

270
00:15:01,680 --> 00:15:05,757
it's a heady drug to basically say that

271
00:15:05,840 --> 00:15:08,077
um that you know we're we're creating

272
00:15:08,160 --> 00:15:12,797
this you know new super intelligence
that is going to could destroy humanity

273
00:15:12,880 --> 00:15:16,877
but we're the only ones who are virt
virtuous enough >> to ensure [laughter]

274
00:15:16,960 --> 00:15:20,477
that this is done correctly
right and I think that you know

275
00:15:20,560 --> 00:15:24,237
>> it's a good recruiting tool
yeah join the virtuous team

276
00:15:24,320 --> 00:15:27,117
>> yes um I think that's right >> the

277
00:15:27,200 --> 00:15:31,757
>> but yeah but I think that is definitely
um you So I I I think of of all the

278
00:15:31,840 --> 00:15:35,277
companies that particular one
has been the most aggressive

279
00:15:35,360 --> 00:15:36,877
uh in terms of the regulatory capture

280
00:15:36,960 --> 00:15:39,517
and pushing these um for the these

281
00:15:39,600 --> 00:15:44,397
regulations. And just I mean let's just
bring it up a level just doesn't have to

282
00:15:44,480 --> 00:15:46,477
be about them. Um there's now something

283
00:15:46,560 --> 00:15:48,797
like 1,200 bills going through state uh

284
00:15:48,880 --> 00:15:52,797
legislators right now to regulate AI. uh

285
00:15:52,880 --> 00:15:56,157
25% of them are in um the the top four

286
00:15:56,240 --> 00:15:58,397
blue states which are California, New

287
00:15:58,480 --> 00:16:03,677
York, Colorado, and Illinois. Uh over
a hundred measures have already passed. Um

288
00:16:03,760 --> 00:16:07,917
I think three of them just got signed
in the last month in California alone.

289
00:16:08,000 --> 00:16:12,077
>> I'll tell you just let me tell you what
Colorado is actually Colorado, Illinois,

290
00:16:12,160 --> 00:16:16,717
and California have all done some
version of a thing called algorithmic

291
00:16:16,800 --> 00:16:20,957
discrimination.
um which I think is it's really

292
00:16:21,040 --> 00:16:22,957
troubling where where it's headed. Um

293
00:16:23,040 --> 00:16:25,277
what what this concept means is that if

294
00:16:25,360 --> 00:16:29,277
the model produces an output

295
00:16:29,360 --> 00:16:31,277
that has a disperate impact on a

296
00:16:31,360 --> 00:16:33,757
protected group then that is algorithmic

297
00:16:33,840 --> 00:16:36,317
discrimination and the list of protected

298
00:16:36,400 --> 00:16:39,117
groups is um is very long. It's more

299
00:16:39,200 --> 00:16:41,437
than just the the the usual ones. So,

300
00:16:41,520 --> 00:16:43,517
for example, in Colorado, they've

301
00:16:43,600 --> 00:16:47,117
defined um people who um may not have

302
00:16:47,200 --> 00:16:52,157
English language proficiency as a
protected group. So, I guess if the

303
00:16:52,240 --> 00:16:55,997
model says something bad about, you know,
illegal aliens and that would be,

304
00:16:56,080 --> 00:17:00,957
you know, that would basically violate
the law. I don't know exactly how model

305
00:17:01,040 --> 00:17:05,677
companies are even supposed to comply
with this rule. I mean, presumably,

306
00:17:05,760 --> 00:17:08,557
presumably discrimination is already
illegal. So if you're a business and you

307
00:17:08,640 --> 00:17:11,357
violate the civil rights laws and you
engage in discrimination, you're already

308
00:17:11,440 --> 00:17:12,957
liable for that. You know, there's no

309
00:17:13,040 --> 00:17:15,917
reason you know if if you happen to um

310
00:17:16,000 --> 00:17:20,637
you know make that mistake and you use
any kind of tool in the process of doing

311
00:17:20,720 --> 00:17:23,837
it, we don't really need to go after the
tool developer because we can already go

312
00:17:23,920 --> 00:17:26,637
after the business that's made that

313
00:17:26,720 --> 00:17:29,037
decision. But the whole purpose of of

314
00:17:29,120 --> 00:17:31,997
these laws is to get at the tool.

315
00:17:32,080 --> 00:17:34,877
They're making not just the the business

316
00:17:34,960 --> 00:17:39,757
that is using AI liable, they're making
the tool developer liable. And I don't

317
00:17:39,840 --> 00:17:42,877
even know how the tool developer
is supposed to anticipate this because how

318
00:17:42,960 --> 00:17:44,797
do you know all the ways that your tool

319
00:17:44,880 --> 00:17:47,277
is going to be used? How do you know

320
00:17:47,360 --> 00:17:49,997
that um that this output, you know,

321
00:17:50,080 --> 00:17:51,997
especially if the output is 100% true

322
00:17:52,080 --> 00:17:54,317
and accurate and the model is doing its

323
00:17:54,400 --> 00:17:57,277
job? um you know then how are you

324
00:17:57,360 --> 00:18:01,597
supposed to know that that was that
output was used as part of a decision

325
00:18:01,680 --> 00:18:05,357
that had a disperate impact?
Nevertheless, you're liable. And the

326
00:18:05,440 --> 00:18:08,957
only way that I can see for model
developers to even attempt to comply

327
00:18:09,040 --> 00:18:12,157
with this is to build a DEI layer into

328
00:18:12,240 --> 00:18:17,357
their models that tries to anticipate
could this answer have a disparit

329
00:18:17,440 --> 00:18:18,877
impact? And if it does, we either can't

330
00:18:18,960 --> 00:18:20,957
give you the answer. We have to sanitize

331
00:18:21,040 --> 00:18:24,957
or distort the answer. Uh, and you know,

332
00:18:25,040 --> 00:18:28,637
you just take this to its logical
conclusion and we're back to, you know,

333
00:18:28,720 --> 00:18:33,197
woke AI, which by the way was a major
objective of the Biden administration.

334
00:18:33,280 --> 00:18:37,357
Um, that that Biden executive order
on AI that we rescended as part of the

335
00:18:37,440 --> 00:18:40,877
Trump administration had something like
20 pages of DEI language in it. They

336
00:18:40,960 --> 00:18:42,877
were very much trying to promote

337
00:18:42,960 --> 00:18:46,557
>> uh DEI values,
they called it, in models

338
00:18:46,640 --> 00:18:49,597
>> and and we saw what the result of that
was. you know, uh, it was, you know, we

339
00:18:49,680 --> 00:18:53,197
saw the whole black George Washington
thing where history was being rewritten

340
00:18:53,280 --> 00:18:55,837
in real time because somebody built, you

341
00:18:55,920 --> 00:18:58,717
know, a DEI layer into the model. Um,

342
00:18:58,800 --> 00:19:03,517
and um, and I, you know, and I I
almost feel like the term woke AI is

343
00:19:03,600 --> 00:19:06,637
insufficient >> to explain
what's going on because it

344
00:19:06,720 --> 00:19:09,757
somehow trivializes it. I mean, what
we're really talking about is Orwellian

345
00:19:09,840 --> 00:19:12,317
AI. You know, we're talking about AI

346
00:19:12,400 --> 00:19:16,797
that um that lies to you, that um that

347
00:19:16,880 --> 00:19:19,917
distorts an an answer that rewrites

348
00:19:20,000 --> 00:19:23,997
history in real time to serve a current
political agenda of the people who are

349
00:19:24,080 --> 00:19:26,557
in power. I mean, it's it's um it's very

350
00:19:26,640 --> 00:19:29,357
Orwellian and um and we were definitely

351
00:19:29,440 --> 00:19:31,837
on that path before President Trump's

352
00:19:31,920 --> 00:19:33,917
election. Uh it was part of the Biden

353
00:19:34,000 --> 00:19:39,037
EO. We saw it happen, you know, in in
the release of that first Gemini model.

354
00:19:39,120 --> 00:19:40,877
uh that was not an accident that you

355
00:19:40,960 --> 00:19:43,197
know that th those distorted outputs

356
00:19:43,280 --> 00:19:47,117
came from somewhere. Um so it was you

357
00:19:47,200 --> 00:19:50,797
know to just to me this is the the

358
00:19:50,880 --> 00:19:54,637
biggest risk of AI actually is it's not

359
00:19:54,720 --> 00:19:58,317
it was not described by James Cameron it
was described by George Orwell you know

360
00:19:58,400 --> 00:20:00,797
it's in my view it's not the Terminator

361
00:20:00,880 --> 00:20:04,717
it's 1984 that um you know that as AI

362
00:20:04,800 --> 00:20:07,117
eats the internet and becomes the main

363
00:20:07,200 --> 00:20:09,597
way that we interact and get our

364
00:20:09,680 --> 00:20:13,357
information online that it'll be used by

365
00:20:13,440 --> 00:20:16,717
the people in power to um to control the

366
00:20:16,800 --> 00:20:18,717
information we receive. Um that it'll

367
00:20:18,800 --> 00:20:21,357
contain anid ideological bias that

368
00:20:21,440 --> 00:20:26,317
essentially it'll censor us. All
that trust and safety apparatus that was

369
00:20:26,400 --> 00:20:29,597
created for social media will be ported
over to this new world of AI. Mark, I

370
00:20:29,680 --> 00:20:32,157
know that you've spoken about this quite
a bit. I think you're absolutely right

371
00:20:32,240 --> 00:20:36,317
about that. And then on top of that,
you've got the surveillance um issues

372
00:20:36,400 --> 00:20:39,917
where, you know, AI is going to know
everything about you. Uh it's going to

373
00:20:40,000 --> 00:20:42,237
be your kind of personal assistant. And

374
00:20:42,320 --> 00:20:45,677
so it's kind of the perfect tool for the

375
00:20:45,760 --> 00:20:48,557
government to um monitor and control

376
00:20:48,640 --> 00:20:53,277
you. And to me that is by far
the biggest risk of of AI. And that's the

377
00:20:53,360 --> 00:20:57,117
thing we should be working towards
preventing. And the problem is a lot of

378
00:20:57,200 --> 00:20:59,917
these regulations that are being whipped

379
00:21:00,000 --> 00:21:02,637
up by these um fear-mongering

380
00:21:02,720 --> 00:21:04,477
techniques, they're actually empowering

381
00:21:04,560 --> 00:21:07,677
the government to uh to engage in this

382
00:21:07,760 --> 00:21:10,397
type of control that I think we should

383
00:21:10,480 --> 00:21:14,957
all be very afraid of. Actually, >>
Sam Alman earlier this week said that in

384
00:21:15,040 --> 00:21:19,917
2028 or by 2028, he expects to have
automated researchers. I'm I'm I'm

385
00:21:20,000 --> 00:21:24,237
curious just for your sort of state
of of AI sort of model development or or

386
00:21:24,320 --> 00:21:28,317
just progress in general and what do you
think are the implications? Some people

387
00:21:28,400 --> 00:21:32,717
have been sort of, you know, saying that,
you know, AGI is two years away,

388
00:21:32,800 --> 00:21:36,957
sort of the AI 2027 papers, the old
Ashen Brener's situational awareness

389
00:21:37,040 --> 00:21:40,397
papers. I'm curious kind of what's your
reading of the of the state of play in

390
00:21:40,480 --> 00:21:43,677
terms of a AI development and what
are the implications from that?

391
00:21:43,760 --> 00:21:47,917
>> So, so my sense is that um people are
in Silicon Valley are kind of pulling back

392
00:21:48,000 --> 00:21:49,757
from the let's call it imminent AGI

393
00:21:49,840 --> 00:21:52,877
narrative. Um I saw Andre Carpathy gave

394
00:21:52,960 --> 00:21:58,237
an interview where all of a sudden he's
he's um rewritten this and he says AGI

395
00:21:58,320 --> 00:22:02,317
is at least a decade away. He's
basically saying that um that you know

396
00:22:02,400 --> 00:22:05,677
reinforcement learning has its limits.
I mean it's it's very useful. It's the

397
00:22:05,760 --> 00:22:09,117
main paradigm right now that they're
making a lot of progress with. But he

398
00:22:09,200 --> 00:22:12,397
says that actually the way that
humans learn is not really through

399
00:22:12,480 --> 00:22:15,837
reinforcement. We do something a little
different which I think is a good thing

400
00:22:15,920 --> 00:22:19,837
because it means that human and AI will
be synergistic, right? Right. I mean the

401
00:22:19,920 --> 00:22:23,757
the AI's um understanding if it's based
on RL will be a little different than

402
00:22:23,840 --> 00:22:27,277
the way that we intuit it [clears
throat] and and reason. Uh but

403
00:22:27,360 --> 00:22:31,117
in any event, I I sense more
of a pullback from this imminent AGI

404
00:22:31,200 --> 00:22:35,117
narrative. Um you know, the idea that
AGI is two years away. Of course, it's

405
00:22:35,200 --> 00:22:38,957
like kind of unclear what people mean
by AGI, but it's kind of, you know, it was

406
00:22:39,040 --> 00:22:43,037
used in this like scary way that um
it's kind of the super intelligence that

407
00:22:43,120 --> 00:22:45,757
would grow beyond our control. Um I feel

408
00:22:45,840 --> 00:22:50,237
like people are pulling back from that
and understanding that yes we're still

409
00:22:50,320 --> 00:22:52,477
making a lot of progress and the

410
00:22:52,560 --> 00:22:55,997
progress is amazing but at the same time

411
00:22:56,080 --> 00:22:59,677
you know what we mean
by intelligence is multifaceted

412
00:22:59,760 --> 00:23:02,877
and it's not like you know there's
progress being made along some

413
00:23:02,960 --> 00:23:04,957
dimensions but it's not along every

414
00:23:05,040 --> 00:23:06,477
dimension.

415
00:23:06,560 --> 00:23:10,477
Um, and so therefore I think again I

416
00:23:10,560 --> 00:23:13,517
would just I mean I've described
actually the situation we're in right

417
00:23:13,600 --> 00:23:17,757
now as a little bit of a Goldilock
scenario where um you know the extremes

418
00:23:17,840 --> 00:23:23,197
would be you know you kind of have the
the scary Terminator situation imminent

419
00:23:23,280 --> 00:23:26,557
super intelligence that'll grow beyond
our control on the other the other

420
00:23:26,640 --> 00:23:29,677
narrative you hear in the press a lot
is that we're in a big bubble. So in other

421
00:23:29,760 --> 00:23:33,437
words the whole thing is fake >>
and the media is basically pushing both

422
00:23:33,520 --> 00:23:37,197
narratives at the same time. >> Yeah.
Um, but any event, I think that

423
00:23:37,280 --> 00:23:41,197
the truth is is more in the middle.
It's kind of a Goldilock scenario where um

424
00:23:41,280 --> 00:23:43,677
we're seeing a lot of innovation. I

425
00:23:43,760 --> 00:23:47,357
think the um the progress is impressive.

426
00:23:47,440 --> 00:23:49,517
Uh it's I think we're going to see big

427
00:23:49,600 --> 00:23:54,557
productivity gains in the economy from
this. But uh but I like the observations

428
00:23:54,640 --> 00:23:57,517
that Bology made recently where he said
there's a couple of things that really

429
00:23:57,600 --> 00:24:00,237
struck me. One was um AI is

430
00:24:00,320 --> 00:24:03,117
polytheistic, not monotheistic. Meaning

431
00:24:03,200 --> 00:24:05,757
what we're seeing is many instead of

432
00:24:05,840 --> 00:24:08,477
just one all- knowing all powerful god.

433
00:24:08,560 --> 00:24:10,717
What we're seeing is a bunch of smaller

434
00:24:10,800 --> 00:24:14,077
deities more specialized models

435
00:24:14,160 --> 00:24:17,197
um you know uh you know it's it's not

436
00:24:17,280 --> 00:24:21,357
that sort of um we're not on that
kind of recursive self-improvement

437
00:24:21,440 --> 00:24:25,197
uh track just yet but you know we're
seeing many different kinds of models

438
00:24:25,280 --> 00:24:27,517
make progress in different areas. Um and

439
00:24:27,600 --> 00:24:32,637
then the other the other um one was just
his observation that AI was um was

440
00:24:32,720 --> 00:24:36,077
middle to middle whereas humans are end
to end and therefore the relationship is

441
00:24:36,160 --> 00:24:40,477
is pretty synergistic and um I think
that's right. I mean I think I think all

442
00:24:40,560 --> 00:24:44,397
th those observations are like resonate
with me in terms of where we're at right

443
00:24:44,480 --> 00:24:47,677
now. >> Yeah. And that that that's very
consistent with what we're seeing as

444
00:24:47,760 --> 00:24:50,877
well where um you know ideas that we

445
00:24:50,960 --> 00:24:53,917
thought would for sure get subsumed by

446
00:24:54,000 --> 00:24:57,997
the big models are becoming amazingly

447
00:24:58,080 --> 00:25:00,477
differentiated businesses. Um just

448
00:25:00,560 --> 00:25:04,477
because the fat tail of the universe is

449
00:25:04,560 --> 00:25:08,157
very fat and you need really kind of

450
00:25:08,240 --> 00:25:12,477
specific understanding of certain
scenarios to build an effective model

451
00:25:12,560 --> 00:25:16,477
and that that's just how it's going. you
know, no model has just like figured out

452
00:25:16,560 --> 00:25:19,917
how to do everything. >> Yeah.
I mean, and the models work best

453
00:25:20,000 --> 00:25:23,117
when they have context, you know, and um

454
00:25:23,200 --> 00:25:26,957
the more I mean, we've all seen this,
the more general your prompt, the less

455
00:25:27,040 --> 00:25:29,197
likely it is that you're going to be

456
00:25:29,280 --> 00:25:31,997
able to um you know, get get a great

457
00:25:32,080 --> 00:25:35,117
response. And um I don't know, if you

458
00:25:35,200 --> 00:25:38,477
tell the AI uh you know, uh something

459
00:25:38,560 --> 00:25:44,237
very general like um what business can
I create to make a billion dollars? it

460
00:25:44,320 --> 00:25:47,517
it's not going to give you something
actionable. You know, you have to get

461
00:25:47,600 --> 00:25:51,517
very specific about what you're trying
to do and it has to have access to

462
00:25:51,600 --> 00:25:53,037
relevant data and then it can give you

463
00:25:53,120 --> 00:25:56,717
some specific answers to to um to a

464
00:25:56,800 --> 00:25:58,877
prompt. Um and I think this is, you

465
00:25:58,960 --> 00:26:03,997
know, partly Bali's point, which is,
you know, the the the AI does not come up

466
00:26:04,080 --> 00:26:06,717
with its own objective. You know,
it needs to be prompted. It needs to be

467
00:26:06,800 --> 00:26:08,637
told what to do. We've seen no evidence

468
00:26:08,720 --> 00:26:11,117
that that's at this stage that that's

469
00:26:11,200 --> 00:26:13,917
changing. we're still at step zero in

470
00:26:14,000 --> 00:26:17,677
terms of AI kind of, you know, somehow

471
00:26:17,760 --> 00:26:22,477
coming up with its own objective. And um
uh and as a result of that, you know,

472
00:26:22,560 --> 00:26:25,357
the model has to be prompted and then it
gives you an output and that output has

473
00:26:25,440 --> 00:26:29,197
to be validated. You have to somehow
make sure it's correct because models

474
00:26:29,280 --> 00:26:33,997
can still be wrong. Um and more likely
you have to iterate a few times because

475
00:26:34,080 --> 00:26:37,597
it doesn't give you exactly what you want.
So now you kind of reprompt and

476
00:26:37,680 --> 00:26:38,797
we've all had this experience, right?

477
00:26:38,880 --> 00:26:41,517
you this is why like the chat the chat

478
00:26:41,600 --> 00:26:46,237
um interface is so is so necessary is
because it takes you a few times to kind

479
00:26:46,320 --> 00:26:50,797
of iterate to get to the output that
actually has value for you. Um again you

480
00:26:50,880 --> 00:26:55,277
know humans are antenn and the the um
the AI is middle to middle. I just don't

481
00:26:55,360 --> 00:26:58,957
you know we haven't seen any evidence
that that fundamental dynamic is is

482
00:26:59,040 --> 00:27:03,117
changing. Um I mean I think you know
we're at the I mean I'd love to hear

483
00:27:03,200 --> 00:27:05,437
what you guys think about this. this.
I mean, we're obviously at the outset of

484
00:27:05,520 --> 00:27:10,477
of agents and um you know, and agents
you can give an objective to and then'll

485
00:27:10,560 --> 00:27:12,077
be able to take tasks on your behalf.

486
00:27:12,160 --> 00:27:15,757
But I suspect that the agents will work
better as well when they have a much

487
00:27:15,840 --> 00:27:19,997
more narrow context. They're much less
likely to go off the rails, start going

488
00:27:20,080 --> 00:27:22,477
in weird directions. If you give it like

489
00:27:22,560 --> 00:27:25,357
a very broad, you know, a very broad

490
00:27:25,440 --> 00:27:29,757
task, it's just not likely to completely
figure it out before it needs human

491
00:27:29,840 --> 00:27:33,357
intervention. But if you give it
something very narrow to do, then it's

492
00:27:33,440 --> 00:27:36,637
much more likely to be successful. Um,

493
00:27:36,720 --> 00:27:37,997
so, you know, I would just guess like,

494
00:27:38,080 --> 00:27:42,557
okay, just you tell the AI, um, you

495
00:27:42,640 --> 00:27:46,717
know, sell my product. You know, it's
it's very unlikely that it's just going

496
00:27:46,800 --> 00:27:48,557
to figure out like what that means and

497
00:27:48,640 --> 00:27:51,197
how to do that. But if you're a sales

498
00:27:51,280 --> 00:27:56,717
rep and you're using the AI to help you,
there's probably very specific tasks

499
00:27:56,800 --> 00:28:00,717
that you can tell it to do and it would
be much more successful doing that. So,

500
00:28:00,800 --> 00:28:03,757
I I just tend to think I mean this also
kind of speaks to the whole job loss

501
00:28:03,840 --> 00:28:08,397
narrative. I just think that this is
going to be a very synergistic tool for

502
00:28:08,480 --> 00:28:10,317
a long time. I don't think it's going to

503
00:28:10,400 --> 00:28:13,757
wipe out um human human jobs. I don't

504
00:28:13,840 --> 00:28:16,957
think the need for human cognition is

505
00:28:17,040 --> 00:28:21,357
going away. Um it's something that, you
know, we'll we'll all use to kind of get

506
00:28:21,440 --> 00:28:25,837
this big productivity boost at least for
the foreseeable future. I mean, I don't

507
00:28:25,920 --> 00:28:28,237
know. I don't know if anyone can any
of us can predict what's going to happen

508
00:28:28,320 --> 00:28:31,917
beyond five or 10 years, but I mean
that's just what I'm seeing right now. I

509
00:28:32,000 --> 00:28:35,517
don't know. I'm curious what what
are you guys seeing this front?

510
00:28:35,600 --> 00:28:39,357
>> Yeah, generally consistent with that
things are improving. So like on agents

511
00:28:39,440 --> 00:28:43,277
um the the early agents, the longer
the running task, the more they would go

512
00:28:43,360 --> 00:28:45,517
like completely bananas and off off the

513
00:28:45,600 --> 00:28:48,557
rails. People are working on that. I do

514
00:28:48,640 --> 00:28:51,437
think like everything's working better

515
00:28:51,520 --> 00:28:54,477
in a context. I at least from what we've

516
00:28:54,560 --> 00:28:57,517
seen that will continue and even you

517
00:28:57,600 --> 00:28:59,757
know to your point on like super smart

518
00:28:59,840 --> 00:29:03,757
models there's like a dozen video models

519
00:29:03,840 --> 00:29:07,437
out there and there's not one that's the

520
00:29:07,520 --> 00:29:11,197
best at everything or even close
to the best at everything. There's like

521
00:29:11,280 --> 00:29:13,917
literally a dozen that are all the best

522
00:29:14,000 --> 00:29:17,917
at one thing. Um, and

523
00:29:18,000 --> 00:29:21,917
which is a little surprising at least
to me because you would think, you know,

524
00:29:22,000 --> 00:29:24,157
just the sheer size of the data would be

525
00:29:24,240 --> 00:29:28,557
an advantage, but even that um hasn't

526
00:29:28,640 --> 00:29:31,997
quite proven out. Uh, you know, it it is

527
00:29:32,080 --> 00:29:33,597
like depending on what you want. Do you

528
00:29:33,680 --> 00:29:36,637
want a meme? Do you want a movie? Do you

529
00:29:36,720 --> 00:29:40,397
want, you know, an ad? Like it's all

530
00:29:40,480 --> 00:29:45,197
very very different. And I think this
gets to your main point, which is, and

531
00:29:45,280 --> 00:29:47,677
Mark Zuckerberg said something
that I really liked. He's like, look,

532
00:29:47,760 --> 00:29:50,797
intelligence is not life. Um, and these

533
00:29:50,880 --> 00:29:53,197
things that we associate

534
00:29:53,280 --> 00:29:56,957
with life, um, like we have an

535
00:29:57,040 --> 00:29:59,837
objective. Um, we have free will, we're

536
00:29:59,920 --> 00:30:02,877
sentient. uh th those just aren't part

537
00:30:02,960 --> 00:30:06,877
of a mathematical model that is you know

538
00:30:06,960 --> 00:30:10,557
searching through a distribution and
figuring out an answer or even you know

539
00:30:10,640 --> 00:30:13,277
a model that you know through a

540
00:30:13,360 --> 00:30:15,117
reinforcement learning technique can

541
00:30:15,200 --> 00:30:18,797
kind of improve its logic. So it's just

542
00:30:18,880 --> 00:30:22,157
like the the the comparison to humans I

543
00:30:22,240 --> 00:30:25,917
think is it just falls short in a lot of
ways is what we're seeing you know we're

544
00:30:26,000 --> 00:30:28,717
just different >> you know
then the models are very good

545
00:30:28,800 --> 00:30:32,557
at things they're better than
humans at things already many things

546
00:30:32,640 --> 00:30:35,277
>> the other [clears throat] thing I'd
bring up related to this which is sort

547
00:30:35,360 --> 00:30:39,437
of this which I think is a little
orthogonal but also quite related which

548
00:30:39,520 --> 00:30:43,597
is sort of is is the future of the world
going to be one or a small number of

549
00:30:43,680 --> 00:30:46,797
companies or for that matter governments
or super AIs that kind of own and

550
00:30:46,880 --> 00:30:50,317
control anything. Um, and sort of all
the value rolls up into a handful of of

551
00:30:50,400 --> 00:30:53,037
entities and and you know, and there
you get into this there's like the hyper

552
00:30:53,120 --> 00:30:55,437
capitalist version of it where a few
companies make all the money or there's

553
00:30:55,520 --> 00:30:58,397
like the hyper communist version of it
where you have total state control or

554
00:30:58,480 --> 00:31:01,197
whatever. >> Um, you know,
or is this a technology

555
00:31:01,280 --> 00:31:05,517
that's going to diffuse out um and be
like in everybody's hands um and be a

556
00:31:05,600 --> 00:31:08,477
tool of empowerment and creativity
um and individual effort, you know,

557
00:31:08,560 --> 00:31:12,637
expressiveness and um and um and as a
tool for basically everybody to use. I

558
00:31:12,720 --> 00:31:15,117
think one of the really striking things
about this period of time and you being

559
00:31:15,200 --> 00:31:19,197
in this role is that this this is the
period in which the scenario number two

560
00:31:19,280 --> 00:31:23,037
is very I think very clearly playing
out um which is that this is the time in

561
00:31:23,120 --> 00:31:25,597
which AI is actually hyperdemocratizing.

562
00:31:25,680 --> 00:31:28,957
I think actually AI is actually
hyperdemocratized. It has spread to more

563
00:31:29,040 --> 00:31:32,317
individuals um uh both in the country
and around the world uh in the shortest

564
00:31:32,400 --> 00:31:35,037
period of time of any new technology I
think in history. you know, we're we're

565
00:31:35,120 --> 00:31:38,237
it's something like 600 million,
you know, users today on on rapidly on the

566
00:31:38,320 --> 00:31:41,437
way to a billion, rapidly on the way
to to 5 billion, you know, kind of across

567
00:31:41,520 --> 00:31:44,717
all the all the consumer products. And
then and then the best AIs in the world

568
00:31:44,800 --> 00:31:46,397
are are in the consumer products, right?

569
00:31:46,480 --> 00:31:50,077
Um and so if you use the if you use,
you know, current day Chad GPT or Grock or

570
00:31:50,160 --> 00:31:53,837
any of these things like, you know,
I I I can't spend more money, you know, uh

571
00:31:53,920 --> 00:31:55,677
and uh and and get access to better AI.

572
00:31:55,760 --> 00:31:57,037
It's it's it's in the consumer products.

573
00:31:57,120 --> 00:32:00,877
And so just in practice, what you have
like playing out in real time is this

574
00:32:00,960 --> 00:32:04,157
technology is going to be in everybody's
hands. Um, and everybody is going to be

575
00:32:04,240 --> 00:32:06,637
able to use it to optim, you know,
optimize the things that they do. Have

576
00:32:06,720 --> 00:32:10,157
it be a thought partner. Have it be, you
know, somebody, you know, a a a um, you

577
00:32:10,240 --> 00:32:13,197
know, an assistant for for for building
companies, you know, starting companies

578
00:32:13,280 --> 00:32:17,117
or creating art or, you know, doing all
the things that people want to do. Um,

579
00:32:17,200 --> 00:32:20,237
you know, my wife was just using it
this morning to design a new uh an

580
00:32:20,320 --> 00:32:23,327
entrepreneurship curriculum for our
10-year-old, right? Um, you know,

581
00:32:23,410 --> 00:32:25,437
[laughter] right? Like,
you know, like literally

582
00:32:25,520 --> 00:32:28,717
it's like, oh, wow, that's like a really
great idea. Um, and it took her a couple

583
00:32:28,800 --> 00:32:31,677
hours and she has like a full curriculum
for him to be able to start his first

584
00:32:31,760 --> 00:32:33,997
video game company and here's all the
different skills that he needs to learn

585
00:32:34,080 --> 00:32:37,277
and here's all the resources and and like
that's just a level of capability.

586
00:32:37,360 --> 00:32:40,797
I mean to to to you know to have
done that without without these modern

587
00:32:40,880 --> 00:32:43,997
consumer AI tools you know you'd have
to go like hire a you know special ed

588
00:32:44,080 --> 00:32:47,277
education specialist or something you
know which is basically impossible um to

589
00:32:47,360 --> 00:32:49,757
do that kind of thing. So and and you
know everybody has these stories now in

590
00:32:49,840 --> 00:32:53,757
their lives among people they know. So
we're we have the I I think we have like

591
00:32:53,840 --> 00:32:56,477
a lot of proof that the the track that
this on is that this is going to be in

592
00:32:56,560 --> 00:32:59,117
everybody's hands and in fact that is
going to be a really good thing and and

593
00:32:59,200 --> 00:33:01,997
I think and I think David I think you
guys are really playing a key role in

594
00:33:02,080 --> 00:33:03,757
making making it happen.

595
00:33:03,840 --> 00:33:06,557
>> I think it's so important that this um

596
00:33:06,640 --> 00:33:10,957
technology remain decentralized because
the kind of the Orwellian concern is

597
00:33:11,040 --> 00:33:15,197
kind of the ultimate centralization >>
and fortunately so far what we're seeing

598
00:33:15,280 --> 00:33:17,277
in the market is that it's hyper

599
00:33:17,360 --> 00:33:19,437
competitive. There's five major model

600
00:33:19,520 --> 00:33:22,557
companies all making huge investments.

601
00:33:22,640 --> 00:33:27,117
Um, and the uh the the benchmark the the

602
00:33:27,200 --> 00:33:31,757
the model performance the evaluations
are relatively clustered and there's a

603
00:33:31,840 --> 00:33:35,837
lot of leaprogging going on. So,
you know, Grock releases new model, it

604
00:33:35,920 --> 00:33:39,917
leaprogs chat GBT and but then CGT
releases something new, they they leap

605
00:33:40,000 --> 00:33:42,637
frog. So they're all like very
competitive and close to each other. And

606
00:33:42,720 --> 00:33:45,197
I think that's uh that's a good thing.

607
00:33:45,280 --> 00:33:47,117
And it's the opposite of what was

608
00:33:47,200 --> 00:33:49,437
predicted, you know, through this like

609
00:33:49,520 --> 00:33:52,317
imminent AGI story where the the the

610
00:33:52,400 --> 00:33:54,637
sort of the the storytelling there was

611
00:33:54,720 --> 00:33:58,877
that one model would get a lead and then

612
00:33:58,960 --> 00:34:02,637
it would direct its own intelligence
to making itself better and then so

613
00:34:02,720 --> 00:34:05,277
therefore its lead would get bigger
and bigger and you kind of get this

614
00:34:05,360 --> 00:34:09,277
recursive self-improvement and pretty
soon you're off to the the singularity.

615
00:34:09,360 --> 00:34:13,037
And um and and we haven't really seen
that, you know, we haven't seen one

616
00:34:13,120 --> 00:34:18,237
model completely pull away in terms
of capabilities. Um and I think that's a a

617
00:34:18,320 --> 00:34:21,117
good thing. thing. And so Eric to your
point about this narrative about the

618
00:34:21,200 --> 00:34:23,837
virtual AI researcher that that was one

619
00:34:23,920 --> 00:34:26,797
variant of this sort of um imminent AGI

620
00:34:26,880 --> 00:34:30,557
narrative is that the steps would be um

621
00:34:30,640 --> 00:34:32,877
you know models get smarter then models

622
00:34:32,960 --> 00:34:35,757
create virtual AI researcher and then

623
00:34:35,840 --> 00:34:40,157
you get a million virtual AI researchers
and then you know it's singularity and I

624
00:34:40,240 --> 00:34:42,557
think just the slight of hand in that is

625
00:34:42,640 --> 00:34:45,597
um what is a virtual AI researcher right

626
00:34:45,680 --> 00:34:48,717
it's like a very easy thing to say, but

627
00:34:48,800 --> 00:34:50,957
like what does that really mean? And you

628
00:34:51,040 --> 00:34:54,317
know, Toby's point about um you know, AI

629
00:34:54,400 --> 00:34:58,237
is still middle to middle. It doesn't
it's not end to end. So if an AI

630
00:34:58,320 --> 00:35:02,237
researcher is endtoend, there's like
things that has to you things the person

631
00:35:02,320 --> 00:35:04,637
has to figure out. They've got to set
their own objective. They've got to be

632
00:35:04,720 --> 00:35:08,317
able to pivot in ways that AI can't. Um

633
00:35:08,400 --> 00:35:12,237
you know, like is it really is that
really feasible to create a virtual AI

634
00:35:12,320 --> 00:35:16,717
researcher? I think there's like parts
of the job that you know AI could get

635
00:35:16,800 --> 00:35:18,397
really good at or even better than

636
00:35:18,480 --> 00:35:21,517
humans, but probably that tool has to be

637
00:35:21,600 --> 00:35:25,117
used by a human AI researcher. So I

638
00:35:25,200 --> 00:35:27,837
guess the the the the the argument I

639
00:35:27,920 --> 00:35:31,277
suspect could be um sort of teological

640
00:35:31,360 --> 00:35:34,157
in the sense that you might need AGI to

641
00:35:34,240 --> 00:35:38,397
create a virtual AI researcher
as opposed to the other way around. And if

642
00:35:38,480 --> 00:35:41,597
that's the case, you don't just,
you know, you're not going to get like

643
00:35:41,680 --> 00:35:45,997
singularity. So, I'm a little bit
skeptical of that claim. Um, you know,

644
00:35:46,080 --> 00:35:48,317
we'll see. S what Sam say he could do it

645
00:35:48,400 --> 00:35:50,957
in 2028. I mean, I guess we'll see. And

646
00:35:51,040 --> 00:35:55,357
>> I [laughter] I I think all those
claims tend to be like recruiting ideas as

647
00:35:55,440 --> 00:35:57,277
opposed to actual predictions.

648
00:35:57,360 --> 00:36:00,637
>> He's he's not the first to to mention
that idea. Other other model companies

649
00:36:00,720 --> 00:36:05,037
have been promoting it, but um yeah,
Leopold mentioned that too. Um you know,

650
00:36:05,120 --> 00:36:06,957
we'll we'll we'll see. But I suspect

651
00:36:07,040 --> 00:36:09,357
that that's what's wrong with that

652
00:36:09,440 --> 00:36:13,437
argument is that virtual AI researcher

653
00:36:13,520 --> 00:36:16,717
um requires AGI and so the idea that

654
00:36:16,800 --> 00:36:21,677
you're going to get AGI through
a virtual AI researcher is um is backwards

655
00:36:21,760 --> 00:36:23,597
but we'll see you know we'll see.

656
00:36:23,680 --> 00:36:25,677
>> David, you've also you and the
administration I think have also been

657
00:36:25,760 --> 00:36:28,797
very supportive of open source AI which
I think also dovtales into this in terms

658
00:36:28,880 --> 00:36:30,717
of like the market being very competitive.

659
00:36:30,800 --> 00:36:32,637
>> Yes. >> Do you want
to spend a moment on what

660
00:36:32,720 --> 00:36:35,677
you guys have been able to do
on that and how you think about it?

661
00:36:35,760 --> 00:36:39,757
Yeah, I mean so so just to the you open
source is very important because I just

662
00:36:39,840 --> 00:36:44,557
think it's synonymous with um you know
freedom I mean software freedom. You

663
00:36:44,640 --> 00:36:48,557
basically could run your own models
on your own hardware and retain control

664
00:36:48,640 --> 00:36:52,477
over your own information.
And by the way this is what enterprises

665
00:36:52,560 --> 00:36:55,037
typically do all the time is you know um

666
00:36:55,120 --> 00:37:00,077
about half the global data center
market is you know is on prem meaning

667
00:37:00,160 --> 00:37:02,797
enterprises and governments create their
own data centers. they don't go to the

668
00:37:02,880 --> 00:37:06,317
big clouds and by the way I've got
nothing against the hyperscalers but

669
00:37:06,400 --> 00:37:09,677
just you know people like to run their

670
00:37:09,760 --> 00:37:14,477
own um you know own data centers and you
know maintain control over their own

671
00:37:14,560 --> 00:37:18,237
data and that kind of thing and I think
that will be true for consumers to some

672
00:37:18,320 --> 00:37:22,237
degree as well um so I do think it's
an important area that we should you know

673
00:37:22,320 --> 00:37:24,717
want to encourage and promote the irony

674
00:37:24,800 --> 00:37:28,637
right now in the market is that um is

675
00:37:28,720 --> 00:37:31,677
that uh the best open source models are

676
00:37:31,760 --> 00:37:36,957
Chinese. Uh, and it's it's sort of a
quirk, right? It's it's the opposite of

677
00:37:37,040 --> 00:37:40,557
what you'd expect. You'd expect like
the American system would promote open and

678
00:37:40,640 --> 00:37:41,997
somehow the Chinese system would promote

679
00:37:42,080 --> 00:37:45,757
closed. That has kind of um, you know,

680
00:37:45,840 --> 00:37:47,917
ended up being a little backwards. Um, I

681
00:37:48,000 --> 00:37:49,917
think there's like good reasons for it.

682
00:37:50,000 --> 00:37:53,277
Uh, it could just be it could just be

683
00:37:53,360 --> 00:37:57,677
kind of a historical accident. the fact
that the Deep Seek founder was like very

684
00:37:57,760 --> 00:38:01,197
committed to open source and kind
of just kind of that like got things

685
00:38:01,280 --> 00:38:05,197
started that way or it could be part
of a deliberate strategy. Uh if you're

686
00:38:05,280 --> 00:38:08,237
China and you're trying to catch up,
open source is a really good way to do

687
00:38:08,320 --> 00:38:12,077
that because you get all the non-aligned
developers to want to help your project

688
00:38:12,160 --> 00:38:15,837
which they can't do if you know with a
closed project. So it's a great strategy

689
00:38:15,920 --> 00:38:18,957
for catching up. And then also if you

690
00:38:19,040 --> 00:38:21,197
think that your business model you know

691
00:38:21,280 --> 00:38:23,437
as a company or as a country is let's

692
00:38:23,520 --> 00:38:26,877
say scale manufacturing of hardware then

693
00:38:26,960 --> 00:38:31,517
you'd want the the software part
to be free or cheap because it's your

694
00:38:31,600 --> 00:38:35,997
compliment right so you try
to commoditize your compliment and I don't

695
00:38:36,080 --> 00:38:39,357
know whether it's by accident or part
of design that that seems to be what the

696
00:38:39,440 --> 00:38:41,517
Chinese strategy has been I think that

697
00:38:41,600 --> 00:38:46,477
the the right answer for the US in this
is to you know to encourage encourage

698
00:38:46,560 --> 00:38:50,717
our own open source. I mean, we I think
it'd be a great thing if we saw more

699
00:38:50,800 --> 00:38:52,877
open source initiatives get get going. I

700
00:38:52,960 --> 00:38:55,117
guess there's one promising one called

701
00:38:55,200 --> 00:38:58,957
um Reflection, which um was founded by

702
00:38:59,040 --> 00:39:02,477
former um you know, engineers from uh

703
00:39:02,560 --> 00:39:06,797
from Google Deep Mind. Uh so I hope we
see more open source innovation in in

704
00:39:06,880 --> 00:39:10,957
the west. But um but look, I think it's
it's very important. is critical and

705
00:39:11,040 --> 00:39:14,797
like I said I in my view it's synonymous
with with freedom and it's definitely

706
00:39:14,880 --> 00:39:19,037
not something we want to suppress.
Now just back to the the closed ecosystem

707
00:39:19,120 --> 00:39:23,037
for a second. It's true we have five
major competitors there and they're all

708
00:39:23,120 --> 00:39:28,077
spending a lot of money. I do worry
a little bit that at some point some point

709
00:39:28,160 --> 00:39:32,317
in time that the model consol not the
model the um the market consolidates and

710
00:39:32,400 --> 00:39:36,157
we end up with you know like a monopoly
or duopoly or something like that as

711
00:39:36,240 --> 00:39:38,237
we've seen in other technology markets.

712
00:39:38,320 --> 00:39:41,037
We saw this with search and you know and

713
00:39:41,120 --> 00:39:45,997
so on down the line. And I just think
that it would be good if this market

714
00:39:46,080 --> 00:39:47,837
stayed more competitive than just one or

715
00:39:47,920 --> 00:39:52,637
two winners. Um and um I I don't I don't

716
00:39:52,720 --> 00:39:55,517
really know what to do about that. I'm
just making that observation. And I I do

717
00:39:55,600 --> 00:39:57,597
think that having open source as an

718
00:39:57,680 --> 00:40:00,557
option always then ensures that even if

719
00:40:00,640 --> 00:40:05,837
the market does consolidate that you
do have an alternative and it's an

720
00:40:05,920 --> 00:40:07,197
alternative that's more fully within

721
00:40:07,280 --> 00:40:09,517
your control as opposed to a large

722
00:40:09,600 --> 00:40:13,277
corporation or um you know or or the

723
00:40:13,360 --> 00:40:16,477
deep state you know working with that

724
00:40:16,560 --> 00:40:20,317
corporation as we saw in the Twitter
files that you know the deep state was

725
00:40:20,400 --> 00:40:22,957
working with all these um you know

726
00:40:23,040 --> 00:40:28,237
social media companies um you know
in in implementing much more widespread

727
00:40:28,320 --> 00:40:32,237
censorship than I think we
any of us thought possible. So

728
00:40:32,320 --> 00:40:35,197
uh we've seen evidence in the past
and again in the social networking space

729
00:40:35,280 --> 00:40:37,597
about um about how the government could

730
00:40:37,680 --> 00:40:40,477
get involved in nefarious ways and it

731
00:40:40,560 --> 00:40:42,957
would be good to have alternatives so

732
00:40:43,040 --> 00:40:46,877
that so you know to to prevent that or

733
00:40:46,960 --> 00:40:50,877
to to make it less likely that
that scenario comes about with AI.

734
00:40:50,960 --> 00:40:54,077
>> Yeah. Well, as you know, we we
and others uh um are very aggressively

735
00:40:54,160 --> 00:40:57,917
investing um in not new model companies
of many kinds, including new foundation

736
00:40:58,000 --> 00:41:01,037
model companies and then also, you know,
as you're probably there are a whole

737
00:41:01,120 --> 00:41:04,797
bunch of new open source efforts um you
know, that are not yet public that, you

738
00:41:04,880 --> 00:41:08,157
know, hopefully will bear fruit over the
next couple years. So, I think at least

739
00:41:08,240 --> 00:41:11,437
at least in the in the medium term,
I think we're looking at an explosion of

740
00:41:11,520 --> 00:41:13,757
model development as opposed
to consolidation and then, you know, we'll

741
00:41:13,840 --> 00:41:15,117
we'll see what happens from there.

742
00:41:15,200 --> 00:41:18,397
>> Yeah, that's really good to hear. I
mean I think you know if we assess kind of

743
00:41:18,480 --> 00:41:21,277
the the state of the AI race um visav v

744
00:41:21,360 --> 00:41:26,397
China uh this is the only area where we
appear to be behind is in this area of

745
00:41:26,480 --> 00:41:29,517
open source models. Um yeah >> I
I think you know if you don't care

746
00:41:29,600 --> 00:41:32,077
whether it's open or closed
I think we have the lead.

747
00:41:32,160 --> 00:41:36,637
>> Uh I think our top models uh model
companies are ahead of the top Chinese

748
00:41:36,720 --> 00:41:40,717
companies although they're quite good
but uh but just this narrow area of open

749
00:41:40,800 --> 00:41:43,837
source seems to be where they have
an advantage. So, it's great to hear that

750
00:41:43,920 --> 00:41:49,037
you guys are seeing, you know, a lot
more um efforts um coming to market.

751
00:41:49,120 --> 00:41:50,717
>> Yeah. Yeah, there's more coming. Yep.

752
00:41:50,800 --> 00:41:52,797
Good. >> Yeah. Yeah.
Definitely more coming.

753
00:41:52,880 --> 00:41:57,197
>> The um Peter Teal quipped, you know,
many years ago that he thought, you

754
00:41:57,280 --> 00:42:01,357
know, crypto would be libertarian
or decentralizing and that uh AI would be

755
00:42:01,440 --> 00:42:04,317
communist or or centralizing.
And I think one thing we've perhaps we've

756
00:42:04,400 --> 00:42:07,677
learned that uh technology isn't
deterministic and that there are a set

757
00:42:07,760 --> 00:42:11,357
of choices that determine whether these
technologies are decentralizing or or

758
00:42:11,440 --> 00:42:15,517
centralizing. And um maybe we could
segue use that as a segue to go deeper

759
00:42:15,600 --> 00:42:19,277
into the the state of the race with China.
Uh maybe David you could lay out

760
00:42:19,360 --> 00:42:23,037
the sort of the what's most important
to to get right. You've already indicated

761
00:42:23,120 --> 00:42:27,197
you know open source is one example. Um
you know you you alluded earlier to sort

762
00:42:27,280 --> 00:42:31,517
of our strategy as relates to chips.
Um you know some people say that yes it's

763
00:42:31,600 --> 00:42:35,037
it's a good idea to do what we're doing
because it'll you know limit domestic

764
00:42:35,120 --> 00:42:38,957
semiconductor production. Other people
say, "Oh, well, you know, some of these

765
00:42:39,040 --> 00:42:42,557
companies say chips are their biggest,
you know, limiting factors and so are we

766
00:42:42,640 --> 00:42:46,397
enabling them in some way." Why don't
you talk about the sort of state of play

767
00:42:46,480 --> 00:42:47,917
and then our strategy?

768
00:42:48,000 --> 00:42:50,557
>> Yeah. So, you know, when we talk about

769
00:42:50,640 --> 00:42:55,117
winning the AI race, um, sometimes
we say we're in a race against China,

770
00:42:55,200 --> 00:42:57,597
sometimes we just leave it a little bit
more vague because I I don't think we

771
00:42:57,680 --> 00:43:02,477
should become overly obsessed with our uh,
competitors or adversaries. I think

772
00:43:02,560 --> 00:43:06,157
whether we win or not will mostly have
to do with the decisions we make about

773
00:43:06,240 --> 00:43:08,877
our own technology ecosystem, not about,

774
00:43:08,960 --> 00:43:12,077
you know, what we do v visa v them. And

775
00:43:12,160 --> 00:43:15,277
so the president in in his July 23rd

776
00:43:15,360 --> 00:43:17,277
speech on AI policy, I think mentioned a

777
00:43:17,360 --> 00:43:20,077
few of the key pillars of of, you know,

778
00:43:20,160 --> 00:43:24,397
of of how we win this this AI race.
Um, and by the way, I'm not saying it ever

779
00:43:24,480 --> 00:43:26,317
ends. This it might be an infinite game,

780
00:43:26,400 --> 00:43:28,477
but um but we want to be in the lead at

781
00:43:28,560 --> 00:43:31,757
least. And um and I I I I do think that

782
00:43:31,840 --> 00:43:36,077
um there could be a period of time where
like you know take the internet where I

783
00:43:36,160 --> 00:43:39,917
mean the internet's still going on. It's
happened but we understand that kind of

784
00:43:40,000 --> 00:43:43,757
who the winners are is kind of baked now.
Um so there could be a period of

785
00:43:43,840 --> 00:43:47,197
time in which you know it's kind of baked
who the the winners in AI are. But

786
00:43:47,280 --> 00:43:51,757
um in any event um you know in terms
of how we win this race um you know I

787
00:43:51,840 --> 00:43:52,797
mentioned a few of the key pillars.

788
00:43:52,880 --> 00:43:56,317
Number one is innovation. You know it's
very important to support the private

789
00:43:56,400 --> 00:43:59,117
sector because they're the ones who
do the innovation. We're not going to

790
00:43:59,200 --> 00:44:03,277
regulate our way to beating our our
adversary. We just have to out innovate

791
00:44:03,360 --> 00:44:06,077
them. Um I mentioned I I think right now

792
00:44:06,160 --> 00:44:09,597
the biggest obstacle is the the the um

793
00:44:09,680 --> 00:44:12,157
frenzy of overregulation happening at

794
00:44:12,240 --> 00:44:17,197
the states. I desperately think we need
um a federal a single federal standard.

795
00:44:17,280 --> 00:44:20,157
Uh a patchwork of 50 different
regulatory regimes is going to be

796
00:44:20,240 --> 00:44:24,237
incredibly burdensome to comply with.
I think even the people who support a lot

797
00:44:24,320 --> 00:44:27,197
of this regulation are now acknowledging
that we're going to need a federal

798
00:44:27,280 --> 00:44:32,077
standard. Uh the problem is that when
they talk about it, what they really

799
00:44:32,160 --> 00:44:35,757
want is to federalize the most
ownorous version of all the state laws

800
00:44:35,840 --> 00:44:39,437
>> and that can't be allowed either.
So, you know, there's going to be a battle

801
00:44:39,520 --> 00:44:42,317
to come. I think as the states um become

802
00:44:42,400 --> 00:44:47,197
more and more unwieldy, you know, as it
becomes more of a trap for startups that

803
00:44:47,280 --> 00:44:50,477
they now have to report into 50
different states at 50 different times

804
00:44:50,560 --> 00:44:53,517
to 50 different agencies about 50
different things, people are going to

805
00:44:53,600 --> 00:44:56,717
realize this is crazy and they're going
to try to federalize it. And then the

806
00:44:56,800 --> 00:45:00,237
question, I think, is whether we get
preemption heavy or preeemption light.

807
00:45:00,320 --> 00:45:04,637
Um you know, do we do we get a I think
everyone's gonna ultimately be in favor

808
00:45:04,720 --> 00:45:08,637
of a of a single federal standard.
Um because I think one of America's

809
00:45:08,720 --> 00:45:12,957
greatest advantages is that we have
a large national market, right? Not 50

810
00:45:13,040 --> 00:45:16,877
separate state markets like kind of,
you know, Europe before the EU wasn't

811
00:45:16,960 --> 00:45:19,917
competitive at all in the internet
because it's 30 different regulatory

812
00:45:20,000 --> 00:45:23,597
regimes. And so, you know, if you're
a European startup and even if you won

813
00:45:23,680 --> 00:45:28,077
your country, it didn't get you very far
because you still had to like, you know,

814
00:45:28,160 --> 00:45:31,117
figure out how to compete in 30 other
countries before you could even win

815
00:45:31,200 --> 00:45:34,477
Europe. And then meanwhile your American
competitors won the entire American

816
00:45:34,560 --> 00:45:38,637
market and is ready to scale up globally.
So the fact that we have a

817
00:45:38,720 --> 00:45:42,077
single national market is just
fundamental to our competitiveness and

818
00:45:42,160 --> 00:45:43,837
it's why you know winners in America

819
00:45:43,920 --> 00:45:45,757
then go on to kind of win the whole

820
00:45:45,840 --> 00:45:50,877
world. So we have to preserve that and I
think we will eventually get some some

821
00:45:50,960 --> 00:45:53,997
federal preeemption. I think
the question will just again be whether we

822
00:45:54,080 --> 00:45:57,197
preempt heavy or preempt light. Um,

823
00:45:57,280 --> 00:46:01,037
second big area is infrastructure,
you know, and energy. You know, we want to

824
00:46:01,120 --> 00:46:06,237
help this this um amazing infrastructure
boom that's happening. And the biggest,

825
00:46:06,320 --> 00:46:10,317
I think, limiting factor there is just
going to be around energy. Uh, I think

826
00:46:10,400 --> 00:46:12,797
President Trump's been incredibly
far-sighted in this. I mean, he was

827
00:46:12,880 --> 00:46:16,957
talking about drill baby drill many
years ago. Um, he understood that energy

828
00:46:17,040 --> 00:46:20,637
is the basis for everything. It's
definitely the basis for this AI boom.

829
00:46:20,720 --> 00:46:24,637
And we want to basically get all
of these unnecessary regulations, the

830
00:46:24,720 --> 00:46:26,637
permitting restrictions, a lot of the

831
00:46:26,720 --> 00:46:30,397
nimism out of the way so that um AI

832
00:46:30,480 --> 00:46:33,917
companies can build data centers and get
power for them. And we can talk about

833
00:46:34,000 --> 00:46:35,197
that more if you want, but I think that

834
00:46:35,280 --> 00:46:39,357
that's a second really huge um uh part

835
00:46:39,440 --> 00:46:42,797
of the of what it's going to take to win
the AI race. And then the third area is

836
00:46:42,880 --> 00:46:45,437
is around exports. And maybe this has

837
00:46:45,520 --> 00:46:48,717
been the most controversial one. Um, and

838
00:46:48,800 --> 00:46:52,477
it really speaks to the cultural
divide between Silicon Valley and and

839
00:46:52,560 --> 00:46:55,997
Washington. So, I think
all of us in Silicon Valley

840
00:46:56,080 --> 00:46:59,917
understand that the way that you win
a technology race is by building the

841
00:47:00,000 --> 00:47:04,477
biggest ecosystem, right? You get
the most developers building on your your

842
00:47:04,560 --> 00:47:08,237
platform, you get the most apps in your
app store, everyone just uses you. I

843
00:47:08,320 --> 00:47:12,477
mean, you know, those are the companies
that typically win are the ones that get

844
00:47:12,560 --> 00:47:16,477
all the users, all the developers, and so
on. And so we in Silicon Valley have

845
00:47:16,560 --> 00:47:19,917
a partnership mentality. You know,
we want to just publish the APIs and get

846
00:47:20,000 --> 00:47:23,437
everyone using them. Uh Washington has
a different mentality, right? It's much

847
00:47:23,520 --> 00:47:27,677
more of a command and control. We want
you to get approved. Um you know, we

848
00:47:27,760 --> 00:47:29,117
kind of want to hoard this technology.

849
00:47:29,200 --> 00:47:32,157
Only America should have it. Um and and

850
00:47:32,240 --> 00:47:36,397
this this was really fundamental,
I think, to the Biden diffusion rule where

851
00:47:36,480 --> 00:47:40,077
the the point of that rule is to stop
diffusion, right? Diffusion is a is a is

852
00:47:40,160 --> 00:47:44,797
a bad word. Um but in Silicon Valley we
understand that diffusion is how you win.

853
00:47:44,880 --> 00:47:47,357
>> Um [snorts] I know I
mean I don't think we ever

854
00:47:47,440 --> 00:47:51,277
called it diffusion before. That was a new
word for me. We just called it usage.

855
00:47:51,360 --> 00:47:53,117
>> Yeah. >> Uh but we
understand that like getting

856
00:47:53,200 --> 00:47:55,517
the most users is how you win. Um so

857
00:47:55,600 --> 00:47:57,437
there's like a fundamental culture clash

858
00:47:57,520 --> 00:48:00,557
going on right now. And um you know the

859
00:48:00,640 --> 00:48:04,957
way I kind of parse it is that what we
decide to sell to China is always going

860
00:48:05,040 --> 00:48:08,637
to be comp uh complicated because you
know they're our competitor and our

861
00:48:08,720 --> 00:48:13,117
adversary and there's the whole
potential dual use and so the the the

862
00:48:13,200 --> 00:48:17,117
question of what you sell to China is is
nuanced. Um but what we sell to the rest

863
00:48:17,200 --> 00:48:20,237
of the world that should be an easy
question which is we should want to do

864
00:48:20,320 --> 00:48:23,517
business with the rest of the world.
We should want to have the largest

865
00:48:23,600 --> 00:48:28,477
ecosystem possible. uh and every country
we exclude from our technology alliance

866
00:48:28,560 --> 00:48:30,957
were basically driving into the arms
of China and it makes their ecosystem

867
00:48:31,040 --> 00:48:33,837
bigger and what we saw under the the the

868
00:48:33,920 --> 00:48:36,957
the Biden years is that they were

869
00:48:37,040 --> 00:48:39,917
constantly pushing other countries uh

870
00:48:40,000 --> 00:48:42,237
into the arms of China starting with the

871
00:48:42,320 --> 00:48:45,517
Gulf states in um October of 2023.

872
00:48:45,600 --> 00:48:48,717
Basically the the Gulf States and I'm
talking about countries like Saudi

873
00:48:48,800 --> 00:48:53,837
Arabia, UAE, long-standing US allies,
they weren't allowed to buy chips from

874
00:48:53,920 --> 00:48:57,437
the US. In other words, they weren't
allowed to set up data centers and

875
00:48:57,520 --> 00:49:00,557
participate in in AI. And um you know,

876
00:49:00,640 --> 00:49:04,637
here we are telling all these countries
that, you know, AI is fundamental to the

877
00:49:04,720 --> 00:49:08,157
future. It's going to be the basis of
the economy, and yet we're excluding you

878
00:49:08,240 --> 00:49:11,277
from participating in the
American tech stack. Well,

879
00:49:11,360 --> 00:49:14,477
>> you know, it's obvious what they're
going to do. You know, the only play

880
00:49:14,560 --> 00:49:17,917
we're giving them is to go to China.
And so, you know, all of these rules

881
00:49:18,000 --> 00:49:19,757
basically just create pent-up demand for

882
00:49:19,840 --> 00:49:22,957
for um for Chinese chips and models. and

883
00:49:23,040 --> 00:49:25,517
it creates a Huawei belt and road. And

884
00:49:25,600 --> 00:49:28,237
we are hearing that um that Huawei is

885
00:49:28,320 --> 00:49:32,477
starting to uh proliferate or diffuse um

886
00:49:32,560 --> 00:49:36,797
in the Middle East and in Southeast Asia.
And I just think it's a really

887
00:49:36,880 --> 00:49:39,277
counterproductive strategy.
We're completely shooting ourselves in the

888
00:49:39,360 --> 00:49:43,277
foot. And like the the greatest irony is

889
00:49:43,360 --> 00:49:46,557
that the people who've been pushing this
strategy of driving all these countries

890
00:49:46,640 --> 00:49:48,877
into China's arms have called themselves

891
00:49:48,960 --> 00:49:53,837
China hawks. you know, as if what
they're doing is uh is hurting China.

892
00:49:53,920 --> 00:49:57,357
No, it's like it's it's helping China.
I mean, it's basically just handing them

893
00:49:57,440 --> 00:50:00,397
markets. And our products are better. Uh

894
00:50:00,480 --> 00:50:04,797
but if you don't give these countries
a choice to buy the American tech stack,

895
00:50:04,880 --> 00:50:06,797
obviously they're going to go with the

896
00:50:06,880 --> 00:50:09,437
Chinese tech stack. And um you know,

897
00:50:09,520 --> 00:50:13,517
China China's out there promoting um you

898
00:50:13,600 --> 00:50:18,317
know, uh Deepseek models and Huawei
chips and they're not like ringing their

899
00:50:18,400 --> 00:50:20,157
hands about

900
00:50:20,240 --> 00:50:23,277
you know whether um you know exporting

901
00:50:23,360 --> 00:50:27,277
chips for a data center in UAE is going
to like create the Terminator and you

902
00:50:27,360 --> 00:50:31,357
know all these like ridiculous
narratives that we've invented to you

903
00:50:31,440 --> 00:50:36,637
know reasons we've invented not to sell
American technology to our friends. So,

904
00:50:36,720 --> 00:50:40,077
um, you know, that that has ended up
being, I think, surprisingly maybe the

905
00:50:40,160 --> 00:50:45,197
most controversial part of what we've
advocated for. Um, but, uh, but there

906
00:50:45,280 --> 00:50:48,157
you have it. So, in any event, I'll
stop there. Those are kind of some of the

907
00:50:48,240 --> 00:50:51,197
major pillars of what
we've been advocating.

908
00:50:51,280 --> 00:50:54,477
>> Should we go deeper on sort of the
infrastructure energy point in terms of

909
00:50:54,560 --> 00:50:57,517
what it's really going to take to to
to get enough capacity or what's what's

910
00:50:57,600 --> 00:51:01,597
most important in that second bullet you
were talking about? Yeah. I mean, well,

911
00:51:01,680 --> 00:51:04,717
so I mean, there there are definitely
people who are much more knowledgeable

912
00:51:04,800 --> 00:51:07,997
about energy than I am and are experts
in the space. But here's what I've been

913
00:51:08,080 --> 00:51:11,197
able to kind of divine is um so first of

914
00:51:11,280 --> 00:51:15,437
all, the administration, President Trump
has signed multiple executive orders to

915
00:51:15,520 --> 00:51:17,197
uh to allow for nuclear to make

916
00:51:17,280 --> 00:51:19,277
permitting easier. We've even freed up

917
00:51:19,360 --> 00:51:22,957
federal land uh for for data centers to

918
00:51:23,040 --> 00:51:26,477
hopefully trying to help get around some
of these state and local restrictions.

919
00:51:26,560 --> 00:51:29,517
And obviously the president has made it

920
00:51:29,600 --> 00:51:32,957
a lot easier to um to stand up new um uh

921
00:51:33,040 --> 00:51:35,437
energy projects, uh power generation,

922
00:51:35,520 --> 00:51:38,157
all that kind of stuff. Um I still think

923
00:51:38,240 --> 00:51:41,277
though that we have um a growing nimi

924
00:51:41,360 --> 00:51:46,077
problem at the state and local level
uh in in the US that that is becoming a

925
00:51:46,160 --> 00:51:51,197
little bit worrisome.
uh and um and uh and if if we don't

926
00:51:51,280 --> 00:51:55,277
figure out a way to to address it, then
it could really slow down uh the build

927
00:51:55,360 --> 00:51:57,837
out of this infrastructure. Um in terms

928
00:51:57,920 --> 00:52:02,797
of of power, so my understanding is that
nuclear is going to take 5 or 10 years.

929
00:52:02,880 --> 00:52:05,597
It's just it's just not something that
we're going to be able to do in the next

930
00:52:05,680 --> 00:52:09,357
two or three years. So in the short term,
it really means that like gas is

931
00:52:09,440 --> 00:52:13,997
the way that these data centers are
going to get um powered. And the the

932
00:52:14,080 --> 00:52:17,437
issue with gas is the shortage there is
not I mean America has plenty of natural

933
00:52:17,520 --> 00:52:20,477
gas and um there's uh it exists in

934
00:52:20,560 --> 00:52:24,877
plenty of enough red states where you
could just build out data centers like

935
00:52:24,960 --> 00:52:28,717
close to the source which be smart.
Uh the issue is there's like a shortage of

936
00:52:28,800 --> 00:52:32,797
these um gas turbines. You know there's
only like two or three companies that

937
00:52:32,880 --> 00:52:35,437
make these things and like there's like

938
00:52:35,520 --> 00:52:40,237
a backlog of two or three years.
So I think that's probably the the the

939
00:52:40,320 --> 00:52:43,997
immediate problem there
that needs to get solved. Um

940
00:52:44,080 --> 00:52:47,437
uh however I do think that in the next
2 or 3 years we could get a lot more out

941
00:52:47,520 --> 00:52:50,317
of the grid. So, I've had um you know,

942
00:52:50,400 --> 00:52:53,837
energy executives tell me that uh that

943
00:52:53,920 --> 00:52:57,997
if we could just shed um 40 hours a year

944
00:52:58,080 --> 00:53:02,477
of peak load from the grid to like
backup generators, to diesel, things

945
00:53:02,560 --> 00:53:04,397
like that, you could free up an

946
00:53:04,480 --> 00:53:07,357
additional 80 gawatts of power, which is

947
00:53:07,440 --> 00:53:12,157
a lot because I guess
the way that it works is um

948
00:53:12,240 --> 00:53:14,157
is, you know, the the grid is only used

949
00:53:14,240 --> 00:53:19,277
about 50% uh 50% of the capacity is used
throughout the here uh because they have

950
00:53:19,360 --> 00:53:23,437
to build enough capacity for the the peak
the peak days like the hottest day

951
00:53:23,520 --> 00:53:24,717
in summer, the coldest day in winter.

952
00:53:24,800 --> 00:53:29,997
Those are your peak days. Uh and they
don't want to um commit to a bunch of

953
00:53:30,080 --> 00:53:32,957
the capacity being used and then you
know you find out that you have a really

954
00:53:33,040 --> 00:53:34,877
cold day in winter and people can't get

955
00:53:34,960 --> 00:53:40,237
enough heat for their their homes.
And so um so they can't like overcommit to

956
00:53:40,320 --> 00:53:44,077
you know to say contracts or data
centers, things like that. Um, but if

957
00:53:44,160 --> 00:53:46,957
you could again just like um if you

958
00:53:47,040 --> 00:53:51,837
could shed that like 40 hours
a year of peak load to backup

959
00:53:51,920 --> 00:53:55,757
uh then you would be able to free up 80
80 gigawatts which is a lot. Um and that

960
00:53:55,840 --> 00:53:59,277
would definitely get us through the next
you know two or three years until the

961
00:53:59,360 --> 00:54:02,157
gas turbine bottleneck's been alleviated.
Um and then eventually you

962
00:54:02,240 --> 00:54:05,837
get to to nuclear. Uh so that would be

963
00:54:05,920 --> 00:54:09,277
very good. I think um the the issue
there is just there's a whole bunch of

964
00:54:09,360 --> 00:54:11,997
insane regulations preventing

965
00:54:12,080 --> 00:54:14,077
uh you know load shedding. So like for

966
00:54:14,160 --> 00:54:19,117
example, you can't use diesel.
Um and uh this you know uh Chris Wright

967
00:54:19,200 --> 00:54:21,997
who's the secretary of energy is is very
good on all this stuff and I think he's

968
00:54:22,080 --> 00:54:26,157
working on unraveling all of this so
we could actually uh actually do this.

969
00:54:26,240 --> 00:54:29,437
>> It's funny David as you talk about
this stuff. It's I I can't help but it's a

970
00:54:29,520 --> 00:54:31,677
little bit like it's a little bit like
the principle is just do the opposite of

971
00:54:31,760 --> 00:54:36,157
the EU. [laughter] Yeah. >> Yeah.

972
00:54:36,240 --> 00:54:39,517
>> Like basically I think everything
we've talked about so far is basically the

973
00:54:39,600 --> 00:54:42,477
opposite of the European approach.

974
00:54:42,560 --> 00:54:46,157
>> Yeah. Well, I mean the Europeans I mean
they have a really different mindset for

975
00:54:46,240 --> 00:54:47,597
all this stuff. when when they talk

976
00:54:47,680 --> 00:54:51,277
about AI leadership, um what they mean

977
00:54:51,360 --> 00:54:55,837
is that they're taking the lead
in defining the regulations and they're

978
00:54:55,920 --> 00:54:58,877
all, you know, it's like that's what
they're proud of is that like they think

979
00:54:58,960 --> 00:55:02,157
that's what their comparative advantage
is is that, you know, they get together

980
00:55:02,240 --> 00:55:04,957
in Brussels and figure out
what all the rules should be

981
00:55:05,040 --> 00:55:08,077
>> and that's what
they call leadership. Um

982
00:55:08,160 --> 00:55:11,597
>> the the EU the EU just announced a uh
I shouldn't beat on them too much, but

983
00:55:11,680 --> 00:55:14,797
they just announced a big new growth fund,
a big new public private sector uh

984
00:55:14,880 --> 00:55:18,957
tech growth fund to grow uh EU companies
to scale. And I I just I would literally

985
00:55:19,040 --> 00:55:21,597
I was just like, well, they it's
actually quite it's almost like a game

986
00:55:21,680 --> 00:55:25,837
show or something. They do everything
they can to strangle them in their crib

987
00:55:25,920 --> 00:55:27,517
>> and then if they
if they make it if they

988
00:55:27,600 --> 00:55:30,797
make it through like a decade of abuse
of small companies, then they're going

989
00:55:30,880 --> 00:55:34,477
to give the money to grow.
>> Um [laughter] >> Well,

990
00:55:34,560 --> 00:55:38,157
well, it's what Yeah. Ronald Reagan had
a line about this which is if it moves

991
00:55:38,240 --> 00:55:40,477
tax it. If it keeps moving regulate it.

992
00:55:40,560 --> 00:55:42,557
If it stops moving subsidize it.

993
00:55:42,640 --> 00:55:44,317
>> Yeah. >> Uh the Europeans
are definitely at the

994
00:55:44,400 --> 00:55:47,997
subsidiz it stage. [laughter] >> Yeah.
And I Yeah. I shouldn't be on them

995
00:55:48,080 --> 00:55:51,517
too much but I just like I've always
been proud to be an American but

996
00:55:51,600 --> 00:55:54,717
particularly now u because like we we
just the fact that we've we've we've

997
00:55:54,800 --> 00:55:57,517
been we we are we it really [snorts]
feels like we're reentering on core

998
00:55:57,600 --> 00:55:59,837
American values in a lot of the things
that we're talking about which is just

999
00:55:59,920 --> 00:56:03,837
really great. >> Yeah.
I mean again it's you know our our

1000
00:56:03,920 --> 00:56:07,997
view is that the first of all we have
to win the I race. We want America to lead

1001
00:56:08,080 --> 00:56:11,837
in this critical area. It's like
fundamental for our economy and our

1002
00:56:11,920 --> 00:56:14,157
national security. How do you do that?

1003
00:56:14,240 --> 00:56:16,557
Well, our companies have to be
successful because they're the ones who

1004
00:56:16,640 --> 00:56:19,917
do the innovation. Again, you're not
going to regulate your way to winning

1005
00:56:20,000 --> 00:56:23,277
the AI race. I'm not saying we don't
need any regulations, but the point is

1006
00:56:23,360 --> 00:56:26,077
just that's not how that's not what's
going to determine whether we're the

1007
00:56:26,160 --> 00:56:30,877
winners or not. David, you recently
tweeted that climate dumerism perhaps is

1008
00:56:30,960 --> 00:56:34,637
giving way to AI dumerism based on,
you know, Bill Bill Gates recent comments.

1009
00:56:34,720 --> 00:56:38,157
Um, what do you mean by this? Do do you
mean it's going to be a major flank, you

1010
00:56:38,240 --> 00:56:41,197
know, of of of the, you know, US left or

1011
00:56:41,280 --> 00:56:42,957
what do what do you mean by this

1012
00:56:43,040 --> 00:56:46,077
comment? Well, I I think the left needs

1013
00:56:46,160 --> 00:56:48,877
a central organizing catastrophe to

1014
00:56:48,960 --> 00:56:53,677
justify their uh takeover of the economy

1015
00:56:53,760 --> 00:56:57,277
and to to regulate everything
and especially to control the information

1016
00:56:57,360 --> 00:57:01,917
space. Uh and I think that you're seeing
that kind of the allure of the whole

1017
00:57:02,000 --> 00:57:06,317
climate change doomer narrative has kind
of faded. Maybe it's the fact that they

1018
00:57:06,400 --> 00:57:08,877
predicted 10 years ago that the whole
world would be underwater in 10 years

1019
00:57:08,960 --> 00:57:12,717
and that hasn't happened. So it's like
at a certain point you get discredited

1020
00:57:12,800 --> 00:57:16,397
by your own catastrophic predictions.
I suspect that's where we'll be with AI

1021
00:57:16,480 --> 00:57:19,837
dumerism in a few years. But in the
meantime, it's a really good narrative

1022
00:57:19,920 --> 00:57:23,117
to kind of take the place of the climate
dumerism. There's actually a lot of

1023
00:57:23,200 --> 00:57:25,757
similarities, I would say. Um, you know,

1024
00:57:25,840 --> 00:57:27,757
you've kind of got um there's a lot of

1025
00:57:27,840 --> 00:57:31,597
kind of um pre-existing Hollywood um

1026
00:57:31,680 --> 00:57:35,277
storytelling and pop culture that supports
this idea. You know, you've got

1027
00:57:35,360 --> 00:57:37,757
the Terminator movies and the Matrix and

1028
00:57:37,840 --> 00:57:40,797
all this kind of stuff. Uh so people

1029
00:57:40,880 --> 00:57:42,637
have been taught to be afraid of this.

1030
00:57:42,720 --> 00:57:46,317
Um and then you know you you um there's

1031
00:57:46,400 --> 00:57:51,197
enough kind of pseudocience behind it,
you know, kind of like um you know with

1032
00:57:51,280 --> 00:57:55,917
uh like you've got all these contrived
studies that like the one where they

1033
00:57:56,000 --> 00:58:00,157
claimed that the AI researcher got
blackmailed by his own AI model or

1034
00:58:00,240 --> 00:58:02,477
whatever. Look, it's very easy to steer

1035
00:58:02,560 --> 00:58:07,597
the model towards um the the answer
that you want and uh and a lot of these

1036
00:58:07,680 --> 00:58:11,517
studies have been very contrived,
but there's this patina of pseudocience to

1037
00:58:11,600 --> 00:58:15,357
it. It's certainly technical enough
that the average person doesn't feel

1038
00:58:15,440 --> 00:58:19,037
comfortable saying that this doesn't
make any sense. I mean, it's more like

1039
00:58:19,120 --> 00:58:21,117
you're not an expert. What do you know?

1040
00:58:21,200 --> 00:58:25,597
Uh and even Republican politicians,
I think, um are kind of falling for this.

1041
00:58:25,680 --> 00:58:29,997
So yeah, I mean it's a rarely desirable
nar and of course you know as AI touches

1042
00:58:30,080 --> 00:58:33,437
more and more things, more and more
parts of the economy, every business is

1043
00:58:33,520 --> 00:58:37,277
going to use it to some degree. If you
can regulate AI, then that kind of gives

1044
00:58:37,360 --> 00:58:42,317
you a lot of control over lots of other
things. And like I mentioned, AI is kind

1045
00:58:42,400 --> 00:58:44,397
of eating the internet.
It's like the main way that you're getting

1046
00:58:44,480 --> 00:58:49,517
information. So again, if you can kind
of get your hooks into what the AI is

1047
00:58:49,600 --> 00:58:54,397
showing people, now you can control what
they see and hear and think. uh which

1048
00:58:54,480 --> 00:58:57,277
dovetales with the whole with the left
censorship, you know, agenda, which

1049
00:58:57,360 --> 00:59:02,077
they've never given up on. Uh dovetales
with their agenda to brainwash kids, uh

1050
00:59:02,160 --> 00:59:05,997
which we, you know, which is kind of
the whole woke thing. So, I mean, this is

1051
00:59:06,080 --> 00:59:09,597
this is going to be very desirable
for the left. And and this is why I mean,

1052
00:59:09,680 --> 00:59:11,037
look, they're they're already doing

1053
00:59:11,120 --> 00:59:12,637
this. This is not like some prediction

1054
00:59:12,720 --> 00:59:16,717
on my part. Basically, after um scam,

1055
00:59:16,800 --> 00:59:20,557
bankr run fraud uh did what he did with
FTX and got sent to jail. Well, he was

1056
00:59:20,640 --> 00:59:25,677
like a big effective altruist and he had
made um pandemics like their big cause.

1057
00:59:25,760 --> 00:59:29,517
They needed a new cause and they got
behind this idea of of X risk which

1058
00:59:29,600 --> 00:59:34,797
existential risk. The idea being if
there's like a 1% chance of um AI ending

1059
00:59:34,880 --> 00:59:38,317
the world then we should drop everything
and just focus on that because you do

1060
00:59:38,400 --> 00:59:41,677
the expected value calculation and so
if it ends humanity then that's the only

1061
00:59:41,760 --> 00:59:45,997
thing you should focus on even if it's
you know very small percentage chance.

1062
00:59:46,080 --> 00:59:50,477
But they really like reorganized behind
this with all and you know they've got

1063
00:59:50,560 --> 00:59:52,717
quite a few advocates and actually it's

1064
00:59:52,800 --> 00:59:54,477
it's an amazing story about how much

1065
00:59:54,560 --> 00:59:58,237
influence they were able to achieve um

1066
00:59:58,320 --> 01:00:02,157
largely behind the scenes or in the
shadows during the Biden years. They

1067
01:00:02,240 --> 01:00:06,317
basically convinced all of the major
Biden staffers of this um view of this

1068
01:00:06,400 --> 01:00:09,997
like imminent super intelligence
is coming. We should be really afraid of

1069
01:00:10,080 --> 01:00:13,757
it. We need to consolidate control
over it. there should only be, you know,

1070
01:00:13,840 --> 01:00:17,277
ideally two or three companies that have
it. We don't want anyone in the rest of

1071
01:00:17,360 --> 01:00:19,997
the world to get it. And then, you know,
what they said is, you know, once we

1072
01:00:20,080 --> 01:00:23,837
make sure that there's only two or three
American companies, we'll u we'll solve

1073
01:00:23,920 --> 01:00:26,717
the coordination problems. That's what
they consider to be the, you know, the

1074
01:00:26,800 --> 01:00:30,237
free market. Uh we'll solve those
coordination problems for buying those

1075
01:00:30,320 --> 01:00:32,637
um companies and we'll be able to

1076
01:00:32,720 --> 01:00:37,437
control this whole thing and prevent
the the genie from escaping the bottle. I

1077
01:00:37,520 --> 01:00:41,437
think it was like this totally paranoid
version of what would happen. And it's

1078
01:00:41,520 --> 01:00:42,797
already being it's already in the

1079
01:00:42,880 --> 01:00:46,717
process of being refuted. But this this

1080
01:00:46,800 --> 01:00:51,277
vision is fundamentally what animated
the Biden executive order on AI. It's

1081
01:00:51,360 --> 01:00:54,717
what animated the Biden diffusion rule
and Mark. I mean, you've talked about

1082
01:00:54,800 --> 01:00:58,157
how you were in a meeting with Biden
folks and they were going to basically

1083
01:00:58,240 --> 01:01:01,037
ban open source and >> they
were going to they're basically

1084
01:01:01,120 --> 01:01:04,557
going to anoint two or three
winners and that was it. Um

1085
01:01:04,640 --> 01:01:07,837
>> yeah, they told us that they told
us that explicitly. Um and uh yeah, and

1086
01:01:07,920 --> 01:01:09,597
they told us exactly what you just said.

1087
01:01:09,680 --> 01:01:11,997
They told us they're going to ban open
source. And when we challenged them on

1088
01:01:12,080 --> 01:01:14,877
the ability to ban open source because
it's, you know, we're talking about, you

1089
01:01:14,960 --> 01:01:18,557
know, math, you know, like mathematical
algorithms that are taught in textbooks

1090
01:01:18,640 --> 01:01:20,317
and YouTube videos and at universities.

1091
01:01:20,400 --> 01:01:23,597
Um, you know, they they they said we'll
during the Cold War, we banned entire

1092
01:01:23,680 --> 01:01:26,877
areas of physics um and put them off
limits and we'll do the same thing for

1093
01:01:26,960 --> 01:01:31,277
math if we have to. >> Um, yeah.
And that's the um Yeah, that's

1094
01:01:31,360 --> 01:01:35,197
the uh Yeah, that was the >>
And you'll be happy to know that the guy

1095
01:01:35,280 --> 01:01:40,477
who actually said that is now
an anthropic employee. It's not.

1096
01:01:40,560 --> 01:01:44,237
>> No, that's exactly right. All all those
and and I mean literally the minute the

1097
01:01:44,320 --> 01:01:47,677
Biden administration was over, all
the top Biden AI employees went to go work

1098
01:01:47,760 --> 01:01:50,957
at Enthropic, which tells you
who they were working with during

1099
01:01:51,040 --> 01:01:53,197
>> the Biden years. Um, >> yeah.

1100
01:01:53,280 --> 01:01:55,597
>> Yeah. >> But no, but I
mean this was this is very

1101
01:01:55,680 --> 01:01:59,757
much the narrative. you sort of had this
imminent uh super intelligence and then

1102
01:01:59,840 --> 01:02:01,357
you know the one of the refrains you

1103
01:02:01,440 --> 01:02:04,557
heard um was that uh AI is like nuclear

1104
01:02:04,640 --> 01:02:07,517
weapons and uh and GPUs are like uranium

1105
01:02:07,600 --> 01:02:12,477
or plutonium or something >> and um
and so and and therefore we need

1106
01:02:12,560 --> 01:02:15,837
like an you know the way the proper
way to to regulate this is with like an

1107
01:02:15,920 --> 01:02:18,317
international atomic energy commission

1108
01:02:18,400 --> 01:02:23,357
>> and so you know again everything would
be sort of centralized and um controlled

1109
01:02:23,440 --> 01:02:24,877
and they would anoint two or three

1110
01:02:24,960 --> 01:02:27,597
winners and you Now this I think this

1111
01:02:27,680 --> 01:02:32,557
narrative really started to fall
apart >> uh with the launch of DeepSeek

1112
01:02:32,640 --> 01:02:35,277
>> which really happened in the first
I don't know couple weeks of the Trump

1113
01:02:35,360 --> 01:02:37,437
administration because you know if you

1114
01:02:37,520 --> 01:02:42,717
asked any of these people what they
thought of um of China during this time

1115
01:02:42,800 --> 01:02:46,077
when they were pushing all these
regulations they basically you know and

1116
01:02:46,160 --> 01:02:49,437
specifically well wait if if we shoot
ourselves in the foot by overregulating

1117
01:02:49,520 --> 01:02:53,277
AI you know won't China just win the AI
race if you were to ask them that what

1118
01:02:53,360 --> 01:02:57,197
they what they would have said they did
say is that China is so far behind us it

1119
01:02:57,280 --> 01:03:00,637
doesn't matter. And furthermore, uh,
and this was said completely without

1120
01:03:00,720 --> 01:03:04,957
evidence that if we basically slow
down to impose, you know, all these

1121
01:03:05,040 --> 01:03:08,317
supposedly healthy regulations, well,
China will just copy us and do the same

1122
01:03:08,400 --> 01:03:13,117
thing. I think it was an absurdly naive
view. I think that if we shoot ourselves

1123
01:03:13,200 --> 01:03:15,917
in the foot, China will just be like,
"Thank you very much. We'll just

1124
01:03:16,000 --> 01:03:19,917
>> take leadership in this technology.
Uh, why wouldn't we?" Um, but this is what

1125
01:03:20,000 --> 01:03:22,397
they said. And um and you know when when

1126
01:03:22,480 --> 01:03:25,197
when the Biden executive order on AI was

1127
01:03:25,280 --> 01:03:28,317
um crafted, there was no discussion

1128
01:03:28,400 --> 01:03:32,797
whatsoever of the China compete,
you know, it was just it was just again

1129
01:03:32,880 --> 01:03:37,277
assumed that we were so far ahead that
um that we could basically do anything

1130
01:03:37,360 --> 01:03:40,557
to our companies and it would just be
it wouldn't it wouldn't really affect our

1131
01:03:40,640 --> 01:03:44,637
our um competitiveness.
And I think I think that narrative

1132
01:03:44,720 --> 01:03:48,877
really started to fall apart with um
DeepSeek at the model level. uh back in

1133
01:03:48,960 --> 01:03:53,277
April uh Huawei launched a technology
called cloud matrix in which they

1134
01:03:53,360 --> 01:03:56,957
compensated for the fact that their
chips individually are not as good as

1135
01:03:57,040 --> 01:04:00,237
Nvidia's chips by networking more of

1136
01:04:00,320 --> 01:04:03,597
them together. They took 384 of them you

1137
01:04:03,680 --> 01:04:05,757
use their prowess in networking to

1138
01:04:05,840 --> 01:04:10,637
create you know this rack system cloud
matrix and it was demonstrated to show

1139
01:04:10,720 --> 01:04:14,157
that you know yes Nvidia chips are
better they're much more power efficient

1140
01:04:14,240 --> 01:04:17,117
but at the rack level at the system

1141
01:04:17,200 --> 01:04:19,677
level you know um Huawei could get the

1142
01:04:19,760 --> 01:04:24,477
job done with these you know asend chips
and cloud matrix and so again I think

1143
01:04:24,560 --> 01:04:27,357
that showed that you know we're not
the only game in town on chips which means

1144
01:04:27,440 --> 01:04:31,197
that if we don't sell our chips to you
know our our friends and allies in the

1145
01:04:31,280 --> 01:04:36,157
Middle East and other places,
then Huawei certainly will. Uh so I think

1146
01:04:36,240 --> 01:04:39,197
it's just been kind of one revelation
after another in which we've learned

1147
01:04:39,280 --> 01:04:42,557
that um that a lot of their
preconceptions and belief were wrong.

1148
01:04:42,640 --> 01:04:46,157
And we've talked about the fact that um
that the that the markets ended up being

1149
01:04:46,240 --> 01:04:49,677
much more decentralized than they ever
could have predicted. And and I'd also

1150
01:04:49,760 --> 01:04:52,637
say one other thing which is they they

1151
01:04:52,720 --> 01:04:57,437
um they also believe that there'd be
you know um imminent catastrophes that

1152
01:04:57,520 --> 01:05:00,077
haven't so this is kind of like
the equivalent to the global warming thing

1153
01:05:00,160 --> 01:05:02,157
where we're all supposed
to be underwater by now.

1154
01:05:02,240 --> 01:05:05,437
>> They [snorts] they were saying that
models trained on I think I don't know

1155
01:05:05,520 --> 01:05:09,117
like 10 to 25 >> flops
or whatever were like way too

1156
01:05:09,200 --> 01:05:14,237
risky. Well, every single model now at
at the frontier is trained on that level

1157
01:05:14,320 --> 01:05:17,597
of compute. And so they would have
banned us from even being at the place

1158
01:05:17,680 --> 01:05:19,757
we're at today if we had listened to

1159
01:05:19,840 --> 01:05:24,797
these people back in 2023. So just
a couple years ago. So that's like really

1160
01:05:24,880 --> 01:05:27,757
important to keep in mind that their
predictions of imminent catastrophe have

1161
01:05:27,840 --> 01:05:32,877
already been refuted. And so things
are moving in a direction that um that I

1162
01:05:32,960 --> 01:05:35,917
think are very different than you know
what they thought in you know let's call

1163
01:05:36,000 --> 01:05:38,637
it the first year after
the launch of chat GPT.

1164
01:05:38,720 --> 01:05:42,157
>> Right. So David just to uh come back
real quick while we still have you on on

1165
01:05:42,240 --> 01:05:46,317
crypto. Um so the administration um
and I think the country had a significant

1166
01:05:46,400 --> 01:05:47,997
victory uh earlier this year with the

1167
01:05:48,080 --> 01:05:50,477
president signing the u stablecoin bill

1168
01:05:50,560 --> 01:05:55,437
into law uh which was the the genius act.
Um and I I think that I'll just

1169
01:05:55,520 --> 01:05:58,797
tell you what we see is like
the positive consequences of that law have

1170
01:05:58,880 --> 01:06:03,037
been even bigger than we thought. Um
and I would say that's both for the stable

1171
01:06:03,120 --> 01:06:06,637
coin industry and you now see actually
a lot of a lot of financial institutions

1172
01:06:06,720 --> 01:06:10,317
of all kinds embracing stable coins um
in a way that that they weren't before.

1173
01:06:10,400 --> 01:06:12,957
um and you know sort of the the the
phenomenon spreading in America you know

1174
01:06:13,040 --> 01:06:15,997
by the way doing you know being being
in the lead and doing very well there. Um

1175
01:06:16,080 --> 01:06:19,357
but but but just also more broadly just
as a signal to the crypto industry that

1176
01:06:19,440 --> 01:06:22,717
like this you know this really is a you
know this really is a new day. Um and

1177
01:06:22,800 --> 01:06:25,197
there really there really are going
to be regulatory frameworks that that make

1178
01:06:25,280 --> 01:06:28,077
these things possible and and you know
that that are responsible but also make

1179
01:06:28,160 --> 01:06:31,677
this industry really possible to flourish.
um and in the US um as you

1180
01:06:31,760 --> 01:06:34,317
know there's a second piece
of legislation um you know being

1181
01:06:34,400 --> 01:06:37,837
constructed right now which is the market
structure bill uh called the the

1182
01:06:37,920 --> 01:06:41,597
clarity act um which is sort of the
phase two of the legislative agenda um

1183
01:06:41,680 --> 01:06:44,157
and I wondered maybe if you could just
tell us a little bit about your view of

1184
01:06:44,240 --> 01:06:46,957
of the importance of that of that bill
and and then you know kind of how do you

1185
01:06:47,040 --> 01:06:51,917
think that process is going >> I
think it's extremely important um so

1186
01:06:52,000 --> 01:06:55,357
as you mentioned we we passed the the
Genius Act few few months ago but that

1187
01:06:55,440 --> 01:06:57,437
was just for stable coins stable coins

1188
01:06:57,520 --> 01:06:59,677
are about 6% of the total market cap in

1189
01:06:59,760 --> 01:07:02,717
terms of tokens. So 94% are all the

1190
01:07:02,800 --> 01:07:07,037
other types of tokens and the Clarity
Act would apply to all of that and

1191
01:07:07,120 --> 01:07:09,117
provide the regulatory framework for all

1192
01:07:09,200 --> 01:07:10,957
those other crypto projects and

1193
01:07:11,040 --> 01:07:14,957
companies. Um, you know, if we could be

1194
01:07:15,040 --> 01:07:19,517
sure that, you know, currently we have
a great SEC chairman, Paul Atkins, and if

1195
01:07:19,600 --> 01:07:23,037
we could be sure that Paul Atkins and
a person like Paul Atkins was always at

1196
01:07:23,120 --> 01:07:26,477
the SEC forever, then we wouldn't
necessarily need legislation because

1197
01:07:26,560 --> 01:07:29,517
they're already in the process
of implementing like much better rules and

1198
01:07:29,600 --> 01:07:32,877
providing regulatory clarity. But the
truth is that we don't know for sure.

1199
01:07:32,960 --> 01:07:37,117
And if you're a founder who's trying
to make a decision now about where you're

1200
01:07:37,200 --> 01:07:40,157
going to build your company, you want
to have certainty for 10 years out, 20

1201
01:07:40,240 --> 01:07:44,637
years out. uh you know we want
to encourage long-term projects and so

1202
01:07:44,720 --> 01:07:48,237
again I think it's very
important to pro to canonize

1203
01:07:48,320 --> 01:07:52,077
uh the the the rules that that's first
to provide the clarity and then to make

1204
01:07:52,160 --> 01:07:56,797
sure there's enough stability around
them and sort of canonize those rules in

1205
01:07:56,880 --> 01:08:00,157
legislation. That's the only way that you
provide that long-term stability. I

1206
01:08:00,240 --> 01:08:03,117
think that we will get the Clarity Act
done. Like you mentioned it passed the

1207
01:08:03,200 --> 01:08:07,597
House with about 300 votes of about
78 Democrats. So it was substantially

1208
01:08:07,680 --> 01:08:11,437
bipartisan. I think it will ultimat
we're it's now going through the Senate.

1209
01:08:11,520 --> 01:08:14,557
Um I think it will ultimately get done.

1210
01:08:14,640 --> 01:08:17,757
Uh we're negotiating with a dozen or so

1211
01:08:17,840 --> 01:08:19,597
Democrat. We have to get to 60 votes. So

1212
01:08:19,680 --> 01:08:24,877
that's the the hard part is under
the filibuster. We got to get to 60. Uh so

1213
01:08:24,960 --> 01:08:28,077
but we're negotiating with about a dozen
Democrats. And I I do think that we will

1214
01:08:28,160 --> 01:08:30,717
ultimately get to to that number. We by

1215
01:08:30,800 --> 01:08:33,677
the way we ended up having 68 uh votes

1216
01:08:33,760 --> 01:08:38,557
in the Senate for genius uh including
18 Democrats. So, I do think that even if

1217
01:08:38,640 --> 01:08:41,757
we just get, you know,
twothirds of the number of

1218
01:08:41,840 --> 01:08:44,957
Democrats that we got for Genius, then,
you know, we'll we'll be we'll be fine

1219
01:08:45,040 --> 01:08:49,757
on on clarity. But I, you know,
it this will provide the the regulatory

1220
01:08:49,840 --> 01:08:53,197
framework again for all the other tokens
besides stable coins. And um I think

1221
01:08:53,280 --> 01:08:58,237
it's just a critical piece of of
legislation. And uh yeah, this this

1222
01:08:58,320 --> 01:09:02,797
would ultimately I think kind
of complete the crypto agenda where we've

1223
01:09:02,880 --> 01:09:07,117
kind of you know moving from you know
Biden's war on crypto to Trump's crypto

1224
01:09:07,200 --> 01:09:11,277
capital of the planet. Um and then you
know I think the industry will have the

1225
01:09:11,360 --> 01:09:14,717
stability it needs and can
just focus on innovating and

1226
01:09:14,800 --> 01:09:17,437
>> there'll be you know rule updates
and things like that but you know we'll

1227
01:09:17,520 --> 01:09:20,877
fundamentally have the foundation
for the industry in place.

1228
01:09:20,960 --> 01:09:26,077
>> On Genius Act um you know President
Trump really made that bill possible. I

1229
01:09:26,160 --> 01:09:29,837
mean, first of all, it was his election
that completely shifted the conversation

1230
01:09:29,920 --> 01:09:34,637
>> on crypto. We would still be, you know,
if a different result had been reached,

1231
01:09:34,720 --> 01:09:38,397
we would have again like figure out
at the SEC. The founders would still be

1232
01:09:38,480 --> 01:09:41,037
getting prosecuted. We wouldn't know
what the rules are. Elizabeth Warren

1233
01:09:41,120 --> 01:09:43,837
would be calling the shots. So,
President Trump's election made

1234
01:09:43,920 --> 01:09:48,797
everything possible. And it's his
commitment to the industry and he and

1235
01:09:48,880 --> 01:09:52,237
his commitment to keeping his um
promises during the election that's made

1236
01:09:52,320 --> 01:09:56,317
all of this possible. But also I mean he
got directly involved in making sure the

1237
01:09:56,400 --> 01:09:59,517
Genius Act passed. It was the
legislation was declared dead many

1238
01:09:59,600 --> 01:10:02,637
times. Uh I saw it with my own eyes that

1239
01:10:02,720 --> 01:10:04,797
you know he was able to persuade um Rick

1240
01:10:04,880 --> 01:10:09,517
Calstrren votes and uh twist arms cajul

1241
01:10:09,600 --> 01:10:12,877
uh and um charm and um you know he

1242
01:10:12,960 --> 01:10:17,437
ultimately got it done and um and I
think that clarity will be a similar

1243
01:10:17,520 --> 01:10:22,557
result. people are always prematurely
declaring these things to be dead or

1244
01:10:22,640 --> 01:10:25,917
whatever. Um there are a lot of twists
and turns in the legislative process.

1245
01:10:26,000 --> 01:10:28,557
It's definitely true that you don't want
to see the sausage getting made, but um

1246
01:10:28,640 --> 01:10:31,037
but anyway, I think we're
on a good track right now.

1247
01:10:31,120 --> 01:10:35,677
>> Good. Fantastic. >> Great.
>> Pete Buddhajes went on allin recently

1248
01:10:35,760 --> 01:10:40,077
and you guys talked about the the left's
identity crisis and you he's hoping for

1249
01:10:40,160 --> 01:10:43,917
more of a moderate, you know, center
center left um to to to emerge. At the

1250
01:10:44,000 --> 01:10:48,077
same time, we see Mdani in New York.
I'm I'm curious what you you think of what

1251
01:10:48,160 --> 01:10:50,637
are you seeing in terms of what
is the future in terms of for the for the

1252
01:10:50,720 --> 01:10:53,997
Democratic party in terms of it is there
a more moderate presence or is it kind

1253
01:10:54,080 --> 01:10:57,597
of this Mani style pop
you know woke populism?

1254
01:10:57,680 --> 01:10:59,597
>> I mean it certainly seems to me that

1255
01:10:59,680 --> 01:11:04,877
Mani and and um I don't know like the woke
socialism seems to be the future of

1256
01:11:04,960 --> 01:11:08,317
the party. I mean that's where all the
energy is and their base. I mean I don't

1257
01:11:08,400 --> 01:11:12,717
want that to be the case. I'd rather
have a rational Democrat party, but um

1258
01:11:12,800 --> 01:11:14,797
but that seems to be where their base

1259
01:11:14,880 --> 01:11:17,677
is, where the energy is. And um you

1260
01:11:17,760 --> 01:11:20,477
don't really hear um Democrats um within

1261
01:11:20,560 --> 01:11:24,717
the party trying to self police
and distance themselves from that. Uh you

1262
01:11:24,800 --> 01:11:28,477
saw I mean all the major figures in the
Democrat party have endorsed Ma Donnie.

1263
01:11:28,560 --> 01:11:29,997
So yeah, I mean that's where that party

1264
01:11:30,080 --> 01:11:34,877
seems to be headed. Um I I I think that

1265
01:11:34,960 --> 01:11:37,677
um partly it's where their base is at. I

1266
01:11:37,760 --> 01:11:41,117
think partly it it might be um a a

1267
01:11:41,200 --> 01:11:44,557
misread of of uh it sort of it could be

1268
01:11:44,640 --> 01:11:46,797
kind of like a partial reaction to to

1269
01:11:46,880 --> 01:11:51,597
Trump where they feel like um you
know establishment politics has kind of

1270
01:11:51,680 --> 01:11:55,037
failed and so they need a populism
of the left to compete with a populism to

1271
01:11:55,120 --> 01:11:57,517
the right. And so I think that's maybe

1272
01:11:57,600 --> 01:12:02,157
part of the calculation for why they're
going in this direction. But I don't you

1273
01:12:02,240 --> 01:12:04,797
know fundamentally I don't think it works.
I don't think socialism works. I

1274
01:12:04,880 --> 01:12:07,037
don't think the um you know, defund the

1275
01:12:07,120 --> 01:12:08,717
police, empty all the jails policies

1276
01:12:08,800 --> 01:12:12,157
work. Um so, you know, I think we're

1277
01:12:12,240 --> 01:12:15,917
about to get another um you know, uh

1278
01:12:16,000 --> 01:12:18,717
case uh a teaching moment in New York.

1279
01:12:18,800 --> 01:12:19,997
Unfortunately, it's not going to be good

1280
01:12:20,080 --> 01:12:23,197
for the city. Uh but, uh you know, we

1281
01:12:23,280 --> 01:12:27,437
we've seen this movie before, but uh
but yeah, that's that's where I mean, it

1282
01:12:27,520 --> 01:12:30,397
does appear that's where the Democrat
party is. Um I don't completely get it.

1283
01:12:30,480 --> 01:12:32,157
I mean, other people have made this

1284
01:12:32,240 --> 01:12:33,677
observation, but they do seem to be on

1285
01:12:33,760 --> 01:12:39,597
the 20% side of every 8020 issue. Um,

1286
01:12:39,680 --> 01:12:43,837
uh, you know, opening the border, um,

1287
01:12:43,920 --> 01:12:45,997
you know, on the, um, soft on crime

1288
01:12:46,080 --> 01:12:48,157
stuff, you know, releasing all the the

1289
01:12:48,240 --> 01:12:51,997
repeat offenders and, um,

1290
01:12:52,080 --> 01:12:56,957
and just sort of this, um, you know,
anti- capitalist approach um, you know,

1291
01:12:57,040 --> 01:13:00,397
which I think will be disastrous for the
economy. I mean this but this is this is

1292
01:13:00,480 --> 01:13:01,517
kind of where the party's at right now.

1293
01:13:01,600 --> 01:13:06,797
It is it's a little scary because it
does mean that if we lose elections in

1294
01:13:06,880 --> 01:13:10,637
places where we do lose elections it's
like you know there you could end up

1295
01:13:10,720 --> 01:13:14,317
with something really horrible. Not just
like you know we're not just playing in

1296
01:13:14,400 --> 01:13:18,317
the 40 yard lines anymore in American
politics. Uh and that that is a little

1297
01:13:18,400 --> 01:13:22,557
bit scary. >> Yeah. >> And I
I do think that you know if it

1298
01:13:22,640 --> 01:13:25,837
weren't for Donald Trump I think in a
way we we might already be there. Um,

1299
01:13:25,920 --> 01:13:27,357
you know, I think, you know, but we have

1300
01:13:27,440 --> 01:13:33,117
to make sure that this um this um the

1301
01:13:33,200 --> 01:13:36,717
Trump revolution continues.

1302
01:13:36,800 --> 01:13:41,117
La lastly, um we were just talking about
New York. Um recently in an episode all

1303
01:13:41,200 --> 01:13:45,357
in in San Francisco, you you you know
endorsed bringing the the National

1304
01:13:45,440 --> 01:13:48,397
Guard. You know, Ben off had his comments.
He he sort of, you know, went

1305
01:13:48,480 --> 01:13:51,997
back and forth with those comments. I'm
curious if speaking of teaching moments,

1306
01:13:52,080 --> 01:13:56,637
I'm curious if you see San Francisco
as savable in in some sense and and what

1307
01:13:56,720 --> 01:13:59,757
what needs to be true
to to get there if so.

1308
01:13:59,840 --> 01:14:03,597
>> Well, Daniel Luri is the best mayor
we've had in decades. So, I think he's

1309
01:14:03,680 --> 01:14:08,717
doing a very good job um within the
constraints that San Francisco presents.

1310
01:14:08,800 --> 01:14:12,717
Uh so, the the the mayor job,
unfortunately, we have a weak mayor in

1311
01:14:12,800 --> 01:14:15,837
San Francisco. I I don't mean him,
I just mean like the way it's all set up.

1312
01:14:15,920 --> 01:14:18,397
the board of supervisors has, you know,

1313
01:14:18,480 --> 01:14:21,517
a ton of power and um over time they've

1314
01:14:21,600 --> 01:14:26,557
been able to kind of transfer power from
the mayor to to themselves. Um and then

1315
01:14:26,640 --> 01:14:29,917
of course you got all these leftwing
judges. I mean, it's just amazing to me

1316
01:14:30,000 --> 01:14:34,717
that um there's a case right now,
this is a case that galvanized me several

1317
01:14:34,800 --> 01:14:36,797
years ago, the case of Troy Mallister

1318
01:14:36,880 --> 01:14:40,397
who was a repeat offender who uh killed

1319
01:14:40,480 --> 01:14:42,317
pe two people on New Year's Eve. I think

1320
01:14:42,400 --> 01:14:45,597
it was like uh 2020.

1321
01:14:45,680 --> 01:14:48,717
Uh and um he was arrested four times,

1322
01:14:48,800 --> 01:14:51,997
you know, uh in the year before that. Uh

1323
01:14:52,080 --> 01:14:55,997
he he he ended up killing these two
people and he had a very very long

1324
01:14:56,080 --> 01:15:01,437
criminal history. He had committed
armed robbery before, stolen many cars.

1325
01:15:01,520 --> 01:15:04,157
Uh and he should have been in jail.
He should not have been released. But he

1326
01:15:04,240 --> 01:15:07,917
was basically released thanks to the
zero bail policies of Chase Bodin, who

1327
01:15:08,000 --> 01:15:11,677
was then the district attorney who we
got recalled. There was a huge outcry. I

1328
01:15:11,760 --> 01:15:15,997
mean, even in San Francisco for there
to be a recall of a polit I mean, you got

1329
01:15:16,080 --> 01:15:18,877
to be like seriously leftwing to um recall

1330
01:15:18,960 --> 01:15:20,637
>> to basically alienate San Francisco and

1331
01:15:20,720 --> 01:15:23,757
Chase Bin managed to be so far out there

1332
01:15:23,840 --> 01:15:26,397
that he alienated even San Francisco.

1333
01:15:26,480 --> 01:15:29,197
>> And yet, I don't
know why Tory Mallister

1334
01:15:29,280 --> 01:15:33,677
isn't sentenced already and in jail
for 20 years plus, but his case is still

1335
01:15:33,760 --> 01:15:37,757
pending through the courts, never ending.
And there's a leftwing judge

1336
01:15:37,840 --> 01:15:41,037
who's considering just giving him
diversion. basically means you just get

1337
01:15:41,120 --> 01:15:45,997
released maybe with an ankle bracelet
or something. Um, that's insane. So, I

1338
01:15:46,080 --> 01:15:48,397
mean, that's what we're dealing with
in San Francisco. I mean, like crazy

1339
01:15:48,480 --> 01:15:49,677
leftwing judges who want to release all

1340
01:15:49,760 --> 01:15:53,917
the criminals. Uh, and um, you know, and

1341
01:15:54,000 --> 01:15:57,197
so I just wonder like is Daniel
up against too many constraints and

1342
01:15:57,280 --> 01:16:00,797
therefore I know he doesn't want the
president to send in the National Guard,

1343
01:16:00,880 --> 01:16:05,277
but maybe ultimately it would be helpful.
Um but in any event, I think

1344
01:16:05,360 --> 01:16:06,957
the president has has agreed to kind of

1345
01:16:07,040 --> 01:16:11,037
hold off on that um out of you,

1346
01:16:11,120 --> 01:16:13,997
you know, Daniel had a good conversation
with the president and asked him to hold

1347
01:16:14,080 --> 01:16:16,637
back and um and you know, the president

1348
01:16:16,720 --> 01:16:19,197
uh agreed and is giving him time to

1349
01:16:19,280 --> 01:16:22,397
implement his solutions. And look, if if

1350
01:16:22,480 --> 01:16:27,597
um if Daniel and his team can keep
making progress and fix the problems

1351
01:16:27,680 --> 01:16:30,877
without the National Guard having
to come in, then so much the better. Um,

1352
01:16:30,960 --> 01:16:33,917
we'll just see if and and I know he
wants to and like I said, he's the best

1353
01:16:34,000 --> 01:16:37,677
mayor we've had in decades. It's just
a question of whether he'll be too

1354
01:16:37,760 --> 01:16:41,517
constrained by the other
powers that be in the city.

1355
01:16:41,600 --> 01:16:43,197
>> David, thank you so much
for coming on the podcast.

1356
01:16:43,280 --> 01:16:46,157
>> Yeah. >> Fantastic. Thank you, David.

1357
01:16:46,240 --> 01:16:47,917
>> Yeah. And thank you for the work. It's

1358
01:16:48,000 --> 01:16:51,277
we we as much as anybody appreciate the

1359
01:16:51,360 --> 01:16:55,357
work that you've done to uh fix
the things in the past and and put us on a

1360
01:16:55,440 --> 01:16:57,197
great road to the future. Yep.

1361
01:16:57,280 --> 01:17:00,797
>> Well, thanks and I appreciate what you
guys have done as well. So, thank you for

1362
01:17:00,880 --> 01:17:03,917
for your support, everything you're doing.
So, >> yeah, >> I appreciate it.

1363
01:17:04,000 --> 01:17:07,680
>> Definitely. [music]

1364
01:17:16,800 --> 01:17:18,820
[music]

1365
01:17:22,040 --> 01:17:24,060
[music]