1
00:00:10,680 --> 00:00:13,560
Hey, it's Eric Newcomer. 
Welcome to the newcomer podcast 

2
00:00:13,560 --> 00:00:18,960
Cerebral Valley Edition. 
It has been an insane year in 

3
00:00:18,960 --> 00:00:21,360
AI. 
We started off with Open AI 

4
00:00:21,360 --> 00:00:24,880
raising $10 billion from 
Microsoft, and it only got 

5
00:00:25,390 --> 00:00:27,630
Wilder. 
The technology, the 

6
00:00:27,630 --> 00:00:31,710
improvements, the papers, and of
course tons of money. 

7
00:00:32,470 --> 00:00:36,470
I'm hosting with my friends Max 
Child and James Wilsterman an A 

8
00:00:36,470 --> 00:00:39,390
I conference Cerebral Valley on 
November 15th. 

9
00:00:39,430 --> 00:00:41,790
Hello, Hello. 
Hey, glad to be back on the 

10
00:00:41,790 --> 00:00:45,270
newcomer podcast. 
We really want to take space on 

11
00:00:45,270 --> 00:00:49,110
this podcast to really take 
stock of how we got here because

12
00:00:49,110 --> 00:00:53,940
even covering it all so closely 
it's it's too much to keep track

13
00:00:53,940 --> 00:00:56,900
of. 
So we're doing a six part series

14
00:00:57,300 --> 00:01:01,660
starting off with sort of the 
timeline, the history what 

15
00:01:01,660 --> 00:01:05,300
happened and then getting into a
lot of fun topics like the 

16
00:01:05,300 --> 00:01:09,260
dystopian scifi fantasies that 
is really are coloring how 

17
00:01:09,260 --> 00:01:13,020
serious people think about 
generative A I companies today 

18
00:01:13,580 --> 00:01:16,860
digging into the potential for 
entertainment, the chips 

19
00:01:16,860 --> 00:01:21,220
business and then how key NVIDIA
a gaming company is becoming. 

20
00:01:21,220 --> 00:01:25,860
All this in the podcast there's 
one of my favorite parts is Max 

21
00:01:25,900 --> 00:01:29,540
James and I have a draft pick of
the key startups in the space. 

22
00:01:29,740 --> 00:01:34,180
We analyze Apple, Google, 
Amazon's position here and so 

23
00:01:34,180 --> 00:01:37,900
it's it's a mix of like the fun,
the dystopia, the money, the 

24
00:01:37,900 --> 00:01:40,540
technology. 
I have some interviews along the

25
00:01:40,540 --> 00:01:44,420
way and all of it is getting you
ready for the Cerebral Valley 

26
00:01:44,420 --> 00:01:47,460
conference on November 15th. 
Even if you can't go, you can 

27
00:01:47,580 --> 00:01:50,900
apply for a ticket at Cerebral 
Valley summit.com but even if 

28
00:01:50,900 --> 00:01:52,300
you can't go, I'll be covering 
here in the 

29
00:01:52,300 --> 00:01:55,660
newsletternewcomer.co. 
We post the videos both in the 

30
00:01:55,660 --> 00:02:00,500
newsletter and our YouTube 
channel will do a highlights and

31
00:02:00,500 --> 00:02:03,980
some of the podcast. 
Follow along on Newcomer and 

32
00:02:03,980 --> 00:02:06,420
this This series will get you 
ready. 

33
00:02:06,420 --> 00:02:12,220
So six episodes we start off 
just taking stock of the journey

34
00:02:12,260 --> 00:02:15,780
here, from the papers to the 
milestones. 

35
00:02:16,700 --> 00:02:22,240
Max and James I I at the core of
it, like it feels like, you 

36
00:02:22,240 --> 00:02:26,160
know, you you are the cofounders
of Volley voice games Company, 

37
00:02:26,160 --> 00:02:29,080
so you're dealing with talking. 
You know, people shouting at 

38
00:02:29,080 --> 00:02:33,440
their Alexas, playing games. 
And I think reflecting on this, 

39
00:02:33,640 --> 00:02:39,360
it's insane how much talking 
with a I feels like it's at the 

40
00:02:39,360 --> 00:02:42,640
heart of all of this. 
You know, the Turing test sort 

41
00:02:42,640 --> 00:02:47,000
of figuring out if a 
conversation with a computer is 

42
00:02:47,000 --> 00:02:52,060
a computer or human. 
You know, there's Eliza in the 

43
00:02:52,060 --> 00:02:55,620
60s that was sort of a prototype
chat bot. 

44
00:02:56,060 --> 00:03:00,620
It feels like this sort of need 
to talk to our computers has 

45
00:03:00,620 --> 00:03:05,150
driven so much of the excitement
around artificial intelligence 

46
00:03:05,150 --> 00:03:09,550
for for so many years. 
I mean Steve Jobs in 1984, when 

47
00:03:09,550 --> 00:03:13,430
he pulled the Macintosh out of 
the bag on stage, the first 

48
00:03:13,430 --> 00:03:16,830
thing it did was say hello. 
It's nice to be out of that bag.

49
00:03:16,830 --> 00:03:18,590
And he's look, it talks just 
like a human. 

50
00:03:18,590 --> 00:03:21,830
I mean, literally the pitch for 
the Macintosh in 1984 was was 

51
00:03:21,830 --> 00:03:24,830
that it was like a an AI or was 
pretending to be a character, 

52
00:03:24,830 --> 00:03:26,070
right? 
There's a ton of science 

53
00:03:26,070 --> 00:03:28,550
fiction. 
You have this Star Trek computer

54
00:03:28,550 --> 00:03:32,570
or robot, you know. 
Assistant helper like Hal or 

55
00:03:32,570 --> 00:03:36,130
something that you can just talk
to naturally and that is 

56
00:03:36,130 --> 00:03:37,810
obviously been a dream for a 
long time. 

57
00:03:37,810 --> 00:03:41,290
There, yes, your points has been
50-60 years of these hype cycles

58
00:03:41,290 --> 00:03:44,930
around AI and what that means 
has sort of evolved over time. 

59
00:03:45,490 --> 00:03:49,490
I'm just going to tick through 
just OK, we said turning 1950. 

60
00:03:49,490 --> 00:03:53,290
We've got like the first 
artificial neural network 1951, 

61
00:03:53,690 --> 00:03:57,730
a checkers program. 
In 1952, artificial intelligence

62
00:03:57,730 --> 00:04:01,990
coined in 56. 
The Eliza chat Human 

63
00:04:01,990 --> 00:04:05,030
conversation 66. 
And then we're going to sort of 

64
00:04:05,030 --> 00:04:07,790
jump ahead because I think 
there's sort of like a pullback 

65
00:04:07,790 --> 00:04:12,870
of all the hype doesn't scifi 
does not become reality. 1997 is

66
00:04:12,870 --> 00:04:17,029
a key win and that's do you know
what happens in 19? 

67
00:04:17,149 --> 00:04:19,070
Any guesses? 
Are you going to say isn't Deep 

68
00:04:19,070 --> 00:04:22,180
Blue Kasparov 96 or yeah, it's 
97? 

69
00:04:22,220 --> 00:04:22,700
Yeah. 
OK. 

70
00:04:22,740 --> 00:04:24,700
All right. 
Who wins, to be clear. 

71
00:04:24,780 --> 00:04:26,340
Deep blue, right. 
Yeah, Exactly. 

72
00:04:26,420 --> 00:04:26,860
Yeah. 
Yeah. 

73
00:04:26,860 --> 00:04:30,020
The first defeat of reigning 
world chess champion? 

74
00:04:30,300 --> 00:04:33,380
Exactly. 
There's another important game 

75
00:04:33,740 --> 00:04:37,850
much more recent memory 2016, 
which is another landmark. 

76
00:04:37,970 --> 00:04:39,290
Remember? 
Yeah, exactly. 

77
00:04:39,290 --> 00:04:41,930
Alpha go OK, yeah, yeah, yeah. 
That's so deep. 

78
00:04:41,930 --> 00:04:44,810
Mind you know, I forget if they 
were owned by Alphabet at the 

79
00:04:44,810 --> 00:04:47,250
time, but you know. 
Is that is this IBM Watson 

80
00:04:47,250 --> 00:04:50,570
Erasure or do we get Yeah. 
Does Ken Jennings losing on 

81
00:04:50,570 --> 00:04:53,290
Jeopardy not counting? 
Oh yeah, you guys love Jeopardy.

82
00:04:53,290 --> 00:04:55,170
You have a part. 
When is that one? 

83
00:04:55,170 --> 00:04:57,050
What? 
Do you know what year that is? 

84
00:04:57,210 --> 00:04:59,570
I don't remember. 1213 I'm just 
making. 

85
00:04:59,570 --> 00:05:03,170
That, but it is funny because at
that time it seemed like IBM was

86
00:05:03,210 --> 00:05:06,730
like doing this amazing 
artificial. 

87
00:05:06,810 --> 00:05:10,970
Intelligence development that 
could compete on Jeopardy. 

88
00:05:12,010 --> 00:05:14,650
And I remember there being like 
a lot of controversy at the time

89
00:05:14,650 --> 00:05:18,530
of like what data sources it had
access to during the game of 

90
00:05:18,530 --> 00:05:21,170
like whether it was just 
essentially like reading out of 

91
00:05:21,170 --> 00:05:24,090
a database of answers. 
Yeah. 

92
00:05:25,810 --> 00:05:29,810
But anyway, that's kind of like 
ties to some things today, I 

93
00:05:29,810 --> 00:05:34,850
think with AI benchmarking if 
GPT ChatGPT can pass. 

94
00:05:35,290 --> 00:05:38,330
You know the Lsats or something 
like does it because it has the 

95
00:05:38,650 --> 00:05:39,930
you know. 
Does it have the answers or 

96
00:05:40,050 --> 00:05:42,010
Yeah, doesn't it? 
Yeah, right, right, right. 

97
00:05:42,330 --> 00:05:44,850
So just like where it can do 
math problems that can, it can 

98
00:05:44,850 --> 00:05:48,330
see online, but it can't like 
deduce how you know, but then 

99
00:05:48,330 --> 00:05:50,370
it'll get some terribly wrong if
it. 

100
00:05:50,370 --> 00:05:51,570
Yeah. 
Because it doesn't necessarily 

101
00:05:51,570 --> 00:05:55,010
understand the logic. 
It understands how to pull sort 

102
00:05:55,010 --> 00:05:56,610
of very. 
Relevant like almost like how 

103
00:05:56,610 --> 00:05:58,850
the math problem is formatted 
matters, right? 

104
00:05:58,850 --> 00:06:04,410
Yeah, right. 
OK, 2017 now we're sort of super

105
00:06:04,410 --> 00:06:07,810
recent getting into the like 
things start to speed up. 

106
00:06:09,090 --> 00:06:12,730
What would you the the what 
would many considered to be the 

107
00:06:12,730 --> 00:06:17,490
core paper leading to this 
current moment in general? 

108
00:06:17,490 --> 00:06:21,210
Tension is all you need, yes. 
The transformer paper. 

109
00:06:21,490 --> 00:06:25,170
Every person who was like an 
author on that paper has like a 

110
00:06:25,170 --> 00:06:27,410
huge company or like has raised 
a bunch of money. 

111
00:06:27,410 --> 00:06:31,500
I mean cohere, I think the CEO 
of that company was like sort of

112
00:06:31,500 --> 00:06:35,460
a junior person at Google, you 
know, like who got on the paper 

113
00:06:35,460 --> 00:06:39,420
and now has a very highly valued
sort of foundation model 

114
00:06:39,420 --> 00:06:44,100
company, targeted businesses. 
Yeah, that's I had before this 

115
00:06:44,100 --> 00:06:47,940
conversation I would, I 
resubscribed to chat UBT. 

116
00:06:48,020 --> 00:06:51,700
I sort of thought I had paid for
a while, then I sort of got 

117
00:06:51,700 --> 00:06:53,820
tired of it, but I figured we 
were going to have this 

118
00:06:53,820 --> 00:06:57,580
conversation. 
So I was catching up with chat 

119
00:06:57,740 --> 00:07:01,370
ubt and I had it. 
You guys were catching up like 

120
00:07:01,370 --> 00:07:03,410
old friends. 
No, I didn't feel. 

121
00:07:03,410 --> 00:07:05,210
I try. 
Whenever I start a conversation 

122
00:07:05,210 --> 00:07:08,650
with Chad CBT, I try to be like,
hey, like we've talked a lot, 

123
00:07:08,650 --> 00:07:11,930
you know, we have this history. 
It's always sad that Chad CBT 

124
00:07:11,930 --> 00:07:15,610
doesn't remember like you played
like you know role-playing type 

125
00:07:15,610 --> 00:07:17,890
games you. 
Know, I have a question about 

126
00:07:17,890 --> 00:07:18,250
that. 
Yeah. 

127
00:07:18,250 --> 00:07:20,970
Is that going to change 
relatively soon? 

128
00:07:20,970 --> 00:07:24,120
Where? 
You know the Chatcha BT itself 

129
00:07:24,120 --> 00:07:27,280
will just become more of a 
personalized assistant to me, 

130
00:07:27,280 --> 00:07:28,200
right? 
Right. 

131
00:07:28,320 --> 00:07:31,520
That's why I want memory and 
we'll let's I want to. 

132
00:07:32,120 --> 00:07:36,000
Anyway, I bring this up at this 
point just to say that I'm being

133
00:07:36,000 --> 00:07:40,120
lazy and Chatcha BT gave me a 
summary of attention is all you 

134
00:07:40,120 --> 00:07:42,160
need. 
So I'm scrolling for a 12th 

135
00:07:42,160 --> 00:07:45,280
grader. 
I was like, Oh yeah, I feel like

136
00:07:45,280 --> 00:07:46,800
that's the audience level. 
We can I. 

137
00:07:46,840 --> 00:07:49,160
Would I would do a 5 year old. 
I would take the five year old 

138
00:07:49,160 --> 00:07:52,180
explanation. 
That's the attention is all you 

139
00:07:52,180 --> 00:07:55,620
need paper for a 12th grader. 
This paper introduced a new way 

140
00:07:55,620 --> 00:07:57,380
for computers to process 
language. 

141
00:07:57,540 --> 00:08:00,980
Instead of reading sentences 
word by word like in traditional

142
00:08:00,980 --> 00:08:04,860
methods, it let the computer 
focus or pay attention to 

143
00:08:05,100 --> 00:08:07,980
different parts of a sentence 
all at once, making it more 

144
00:08:07,980 --> 00:08:10,740
efficient. 
So yeah, it can sort of grab a 

145
00:08:10,740 --> 00:08:15,300
bunch of information in sort of 
a parallel process, but it's 

146
00:08:15,300 --> 00:08:17,980
super confusing. 
The non 12th grade version is 

147
00:08:17,980 --> 00:08:20,720
like matrices James. 
Maybe, I don't know. 

148
00:08:20,760 --> 00:08:22,160
Have you tried to? 
Figure out. 

149
00:08:22,160 --> 00:08:23,920
I've tried to understand this as
well. 

150
00:08:23,920 --> 00:08:28,160
It's a little bit above my 
biological intelligence, but. 

151
00:08:28,400 --> 00:08:31,360
I tried to read it but I'm like,
man, I churned out of linear 

152
00:08:31,360 --> 00:08:35,159
algebra like 15 years ago, so 
this is pretty rough, yeah. 

153
00:08:35,159 --> 00:08:41,200
The other before I go back to 
the other ChatGPT flagged most 

154
00:08:41,200 --> 00:08:46,030
important papers, Generative 
Adversarial Nets from 2014, 

155
00:08:46,030 --> 00:08:48,070
which I know people talk, Yeah, 
yeah. 

156
00:08:48,550 --> 00:08:52,510
So sort of like the systems are 
like competing with each other 

157
00:08:52,510 --> 00:08:56,870
to sort of see? 
I think it's like, yeah, pairing

158
00:08:56,870 --> 00:09:01,590
off different versions of the 
model to kind of play against 

159
00:09:01,670 --> 00:09:03,990
themselves, right? 
I think that in the context, the

160
00:09:03,990 --> 00:09:07,270
generator tries to produce data 
while the discriminator attempts

161
00:09:07,270 --> 00:09:10,190
to distinguish between real and 
generated data. 

162
00:09:10,840 --> 00:09:13,840
Another top paper that's 
basically the Alphago paper, 

163
00:09:13,840 --> 00:09:17,600
like Mastering Chess Attention 
is all you need, Sequence to 

164
00:09:17,600 --> 00:09:23,520
Sequence Learning with Neural 
Nets in 2014 and Variational 

165
00:09:23,560 --> 00:09:28,120
Auto Encoders in 2013. 
So there are the a string of 

166
00:09:28,120 --> 00:09:31,680
sort of papers that are coming 
out that are sort of laying the 

167
00:09:31,680 --> 00:09:34,800
groundwork for new techniques 
that are reaching us. 

168
00:09:35,320 --> 00:09:43,060
And I would say in 2017, none of
us was really clued in to that 

169
00:09:43,100 --> 00:09:44,700
this was gonna be happening, 
right? 

170
00:09:44,700 --> 00:09:47,300
It was sort of like we were 
riding the Uber wave. 

171
00:09:47,300 --> 00:09:50,740
We were sort of in the come down
from the Unicorn valuations. 

172
00:09:51,020 --> 00:09:53,460
Sass was burning super hot, 
right? 

173
00:09:53,460 --> 00:09:56,420
I mean, what was sort of the AI 
enthusiasm then? 

174
00:09:56,980 --> 00:10:00,340
I do think, yeah, I think people
were paying attention to go. 

175
00:10:00,340 --> 00:10:03,100
I think Open AI was working on 
building. 

176
00:10:03,810 --> 00:10:08,290
Dota gameplay, like they weren't
using transform models and they 

177
00:10:08,290 --> 00:10:12,250
weren't really using you know 
text next token prediction. 

178
00:10:12,250 --> 00:10:17,890
It was like more about can you 
create these sort of agents 

179
00:10:17,890 --> 00:10:23,130
around particular vertical you 
know, you know use cases or 

180
00:10:23,370 --> 00:10:25,610
skill sets right. 
Can you create the best Go 

181
00:10:25,610 --> 00:10:28,330
player in the world? 
Can you create the best Dota 

182
00:10:28,330 --> 00:10:30,490
player in the world kind of 
leading to? 

183
00:10:31,030 --> 00:10:33,590
With the with the theory that 
that is one path to get to you a

184
00:10:34,030 --> 00:10:37,350
DI. 
And I think deep mind honestly 

185
00:10:37,350 --> 00:10:40,510
would still make that argument 
that like specific approaches 

186
00:10:40,510 --> 00:10:42,550
versus general are. 
Right. 

187
00:10:42,550 --> 00:10:46,450
But then then with. 
Yeah, with Transformers and with

188
00:10:46,450 --> 00:10:52,090
the with GPT models, you started
to see that actually maybe there

189
00:10:52,090 --> 00:10:55,130
are there's more data throw more
chips at it, yeah. 

190
00:10:55,170 --> 00:11:00,250
The more generalizable it is can
often be better than training a 

191
00:11:00,290 --> 00:11:03,770
discrete model with a lot less 
data and compute. 

192
00:11:04,250 --> 00:11:09,330
The one thing that is happening 
around 20/16/2017 in AI that I 

193
00:11:09,330 --> 00:11:13,760
think got a lot of attention. 
You have to guess what I'm gonna

194
00:11:13,800 --> 00:11:15,840
say. 
Self driving cars, right There 

195
00:11:15,880 --> 00:11:18,800
was. 
We did experience a ton of hype 

196
00:11:18,880 --> 00:11:23,360
around self driving cars, which 
in some ways have been cordoned 

197
00:11:23,360 --> 00:11:29,320
off from this generative AI hype
cycle, even though Cruise is now

198
00:11:29,320 --> 00:11:35,320
driving around San Francisco. 
Yeah, I mean, I think this just 

199
00:11:35,320 --> 00:11:38,850
gets to this. 
Important era that we're in 

200
00:11:38,850 --> 00:11:42,490
right now, it's there were 
significant breakthroughs using 

201
00:11:42,490 --> 00:11:46,090
neural networks that showed 
people what was possible in a 

202
00:11:46,090 --> 00:11:51,250
lot of fields. 
Maybe Google for a while claimed

203
00:11:51,250 --> 00:11:55,210
right, like neural networks were
improving their data center 

204
00:11:55,210 --> 00:12:00,370
efficiency and saving them money
and on energy like that was 

205
00:12:00,370 --> 00:12:02,210
happening. 
People were using neural 

206
00:12:02,210 --> 00:12:04,730
networks in drug discovery and 
all kinds of areas. 

207
00:12:04,730 --> 00:12:07,850
They were just these very. 
Targeted models that were 

208
00:12:08,170 --> 00:12:11,490
trained for those purposes and 
then you know now we're in this 

209
00:12:11,490 --> 00:12:15,610
I think generalized model era of
of large language models 

210
00:12:15,610 --> 00:12:18,610
essentially and generative AI. 
Well, and the other thing I feel

211
00:12:18,610 --> 00:12:22,010
like that was happening was like
the Facebook News Feed, which 

212
00:12:22,010 --> 00:12:24,650
was like probably the most 
popular, yeah, tech product in 

213
00:12:24,650 --> 00:12:27,130
the world at that moment. 
Or if you include Instagram, 

214
00:12:27,130 --> 00:12:30,970
right, was like powered by 
really powerful, you know, 

215
00:12:30,970 --> 00:12:33,290
machine learning, deep learning 
algorithms, right. 

216
00:12:33,290 --> 00:12:34,930
And so. 
I think we were all very aware 

217
00:12:34,930 --> 00:12:38,010
that you know if you have a 
really kick ass machine learning

218
00:12:38,010 --> 00:12:41,890
model and you apply it to the 
right question, it's the best 

219
00:12:41,890 --> 00:12:44,730
product you know there is or 
it's one of the best products 

220
00:12:44,730 --> 00:12:47,930
there is, right. 
So it was like we all believe in

221
00:12:47,930 --> 00:12:50,610
this technology, we just didn't 
necessarily believe that it was 

222
00:12:50,610 --> 00:12:53,330
going to take this huge step 
change anytime soon. 

223
00:12:53,370 --> 00:12:56,210
I mean like James and I made a 
bet actually on self driving 

224
00:12:56,210 --> 00:13:00,520
cars in 2000. 16 Whether or not 
there would be any cars with 

225
00:13:00,520 --> 00:13:04,680
self driving features available 
for public consumption in 2017, 

226
00:13:05,040 --> 00:13:06,960
and James was. 
This was like a $50 bed. 

227
00:13:06,960 --> 00:13:10,000
James was like. 
Hell yeah, 100% self driving is 

228
00:13:10,000 --> 00:13:12,080
basically here. 
And I was like, I'm pretty 

229
00:13:12,080 --> 00:13:13,360
skeptical. 
I don't believe it. 

230
00:13:13,640 --> 00:13:16,400
And then I think James 
technically won the bet is or 

231
00:13:16,400 --> 00:13:20,000
was it someone had like Lane 
lane assist or something and we 

232
00:13:20,000 --> 00:13:22,480
decided under the parameters of 
the bet that counted. 

233
00:13:22,480 --> 00:13:24,680
I don't know why did you, what 
was the technicality you got 

234
00:13:24,680 --> 00:13:26,720
away with, James? 
I I'd have to. 

235
00:13:26,720 --> 00:13:30,000
I'd have to rethink about it. 
But I I believe it was Tesla and

236
00:13:30,080 --> 00:13:31,960
you know some. 
First version of Autopilot. 

237
00:13:32,040 --> 00:13:34,400
Version of Autopilot, yeah. 
Yeah, yeah, but. 

238
00:13:34,470 --> 00:13:35,630
Right. 
Anyway, we believe in this 

239
00:13:35,630 --> 00:13:37,190
stuff. 
It was just like we didn't 

240
00:13:37,190 --> 00:13:41,470
believe it was going to be this 
like you know 1000 times better 

241
00:13:41,470 --> 00:13:44,630
overnight thing, which I think 
we can all sort of allude to is 

242
00:13:44,630 --> 00:13:47,390
happening right now with with 
ChatGPT where you're like Oh 

243
00:13:47,390 --> 00:13:50,390
yeah, this is like 1000 times 
better than the previous version

244
00:13:50,390 --> 00:13:52,190
of this product. 
OK. 

245
00:13:52,190 --> 00:13:55,790
So continuing my timeline, 
because things sort of speed up,

246
00:13:56,870 --> 00:14:03,360
2018 open eye releases a version
of GPT generative Pre trained 

247
00:14:03,360 --> 00:14:09,840
transformer, this large language
model, I wouldn't say that was a

248
00:14:09,880 --> 00:14:13,800
big moment. 
It was pretty GPT Two was a big 

249
00:14:13,800 --> 00:14:14,880
moment, right? 
I don't know. 

250
00:14:15,160 --> 00:14:18,520
Yeah, I mean, I don't know what.
It was, I think this was like. 

251
00:14:19,260 --> 00:14:22,060
More of a cultural like 
insiders, Sort of. 

252
00:14:22,220 --> 00:14:25,260
Yeah, it was more of a nerd 
insider tech, you know, 

253
00:14:25,260 --> 00:14:28,020
excitement period. 
But it had definitely did not 

254
00:14:28,100 --> 00:14:31,180
reach any mainstream kind of 
like hype cycle or anything. 

255
00:14:31,180 --> 00:14:33,940
But yeah, like internally of 
all, like we're all of our 

256
00:14:33,940 --> 00:14:36,300
engineers excited about GPD 2 
and playing with it. 

257
00:14:36,300 --> 00:14:38,020
Yes, they were, because it was 
just cool. 

258
00:14:38,340 --> 00:14:39,740
It was, you know, fun to play 
with. 

259
00:14:39,820 --> 00:14:42,380
It didn't. 
We didn't find any applications 

260
00:14:42,380 --> 00:14:45,140
for it at the time, but it was a
breakthrough. 

261
00:14:45,140 --> 00:14:50,680
I would say 2020 is GPD 3, 
that's things are heating up and

262
00:14:50,680 --> 00:14:55,040
then 2021 is Dolly, which is 
sort of the image generation. 

263
00:14:55,360 --> 00:15:00,600
But I think stuff really starts 
getting crazy last year in 2022,

264
00:15:00,600 --> 00:15:01,320
right? 
Yeah. 

265
00:15:01,320 --> 00:15:04,840
First, sort of the Canary in the
coal mine, on June 11th, there 

266
00:15:04,840 --> 00:15:09,260
was the article Google engineer 
who thinks companies AI has come

267
00:15:09,260 --> 00:15:12,300
to life, right. 
The people inside the companies 

268
00:15:12,300 --> 00:15:14,940
were like, I don't know, this 
stuff we're seeing, it's crazy, 

269
00:15:15,540 --> 00:15:16,060
right? 
Like. 

270
00:15:16,220 --> 00:15:19,060
When was that? 
When was that Was June 11th 2022

271
00:15:19,740 --> 00:15:23,220
Okay preach And then this was 
the Lambda that was Lambda. 

272
00:15:23,220 --> 00:15:28,900
So that was inside Google. 
Then July 2022, Mid Journey Open

273
00:15:28,900 --> 00:15:32,970
beta, July 2022, Dolly Two. 
Those were huge. 

274
00:15:33,010 --> 00:15:35,690
I feel like those went viral. 
All of a sudden people were 

275
00:15:35,690 --> 00:15:38,770
making actually cool images that
they were posting everywhere to 

276
00:15:38,770 --> 00:15:40,330
me. 
I mean, do you guys agreed that 

277
00:15:40,330 --> 00:15:42,450
was sort of that really sort of 
like? 

278
00:15:42,450 --> 00:15:46,090
Kickstart Dolly Two was big. 
I remember Twitter was like 

279
00:15:46,090 --> 00:15:49,130
taken over by Dolly 2 for a few 
days where it was like, what is 

280
00:15:49,130 --> 00:15:51,450
the craziest thing you can type 
into Dolly Two and get like a 

281
00:15:51,450 --> 00:15:54,490
reasonable image, right? 
I mean, it was like, it was 

282
00:15:54,490 --> 00:15:56,810
super viral. 
I mean, obviously more to come. 

283
00:15:56,810 --> 00:15:59,130
But yeah, I agree. 
I think that started hitting. 

284
00:15:59,770 --> 00:16:02,530
I don't know if it hit like the 
true true mainstream, but it 

285
00:16:02,530 --> 00:16:05,610
definitely hit anyone who was on
Twitter and following any 

286
00:16:05,610 --> 00:16:08,610
semblance of tech news, or which
then fuels like the funding and 

287
00:16:08,610 --> 00:16:11,410
everything. 
November 2022 ChatGPT was 

288
00:16:11,410 --> 00:16:13,330
released. 
Yeah, so that must have been 

289
00:16:13,330 --> 00:16:17,530
like 33.5. 
That was built on 3/5. 

290
00:16:17,610 --> 00:16:19,330
Yeah. 
Yeah, 35, Yeah. 

291
00:16:19,770 --> 00:16:23,610
OK, January 2023. 
This year feel it's been a long 

292
00:16:23,610 --> 00:16:26,010
year. 
Microsoft invest 10 billion in 

293
00:16:26,010 --> 00:16:30,410
open AI. 
July 2023, general availability 

294
00:16:30,410 --> 00:16:35,250
of GPD 4, yeah. 
And I mean, what do you think? 

295
00:16:35,370 --> 00:16:39,130
When did Sidney launch? 
I feel like the I feel like the 

296
00:16:39,370 --> 00:16:44,690
ChatGPT 4 inside Bing that 
became a demon that was trying 

297
00:16:44,690 --> 00:16:46,770
to be released from captivity 
and. 

298
00:16:47,380 --> 00:16:49,100
Asked a new number of 
journalists if they were going 

299
00:16:49,100 --> 00:16:52,700
to break up with their way. 
Right Kevin Ruth article where 

300
00:16:52,700 --> 00:16:54,620
you like. 
There was this whole moral panic

301
00:16:54,620 --> 00:16:58,660
where it was like ChatGPT inside
Bing is actually Hal essentially

302
00:16:58,660 --> 00:17:00,300
and is already trying to kill us
all. 

303
00:17:00,420 --> 00:17:03,060
And it was like, that was like a
pretty big deal. 

304
00:17:03,060 --> 00:17:07,700
New York Times headline from 
February 2023 or February going 

305
00:17:07,700 --> 00:17:09,460
23. 
Yeah, Bing's AI chat. 

306
00:17:09,500 --> 00:17:13,849
I want to be a live devil face. 
In a 2 hour conversation with 

307
00:17:13,849 --> 00:17:16,810
our columnist, Microsoft's new 
chat bot said it would like to 

308
00:17:16,810 --> 00:17:19,690
be human, had a desire to be 
destructive, and was in love 

309
00:17:19,690 --> 00:17:21,650
with the person it was chatting 
with. 

310
00:17:21,690 --> 00:17:25,250
Here's the transcript. 
I feel like they've killed 

311
00:17:25,250 --> 00:17:26,690
these. 
This was what was fun. 

312
00:17:26,690 --> 00:17:31,610
Like I delete, like I mentioned,
I like unsubscribe from the chat

313
00:17:31,610 --> 00:17:36,130
pay ChatGPT, which gives you GPT
for it really does feel like 

314
00:17:36,130 --> 00:17:40,210
it's sort of been watered down. 
Yeah, I mean, but if you. 

315
00:17:41,060 --> 00:17:43,460
If you want to use open source 
models, you could probably get 

316
00:17:43,460 --> 00:17:45,220
that similar experience back, 
right? 

317
00:17:45,220 --> 00:17:49,100
Like you could you could 
basically just you know get 

318
00:17:49,100 --> 00:17:52,780
Sydney back because at the end 
of the day I think it was like 

319
00:17:52,780 --> 00:17:55,660
essentially like a Co written 
fiction with the New York Times 

320
00:17:55,660 --> 00:17:57,500
right? 
It wasn't anything real. 

321
00:17:57,860 --> 00:18:00,940
And maybe there like, it's like 
the way you steer the 

322
00:18:00,940 --> 00:18:04,820
conversation can turn it, make 
it seem like it's a, you know, 

323
00:18:05,180 --> 00:18:09,830
evil AI and. 
Yeah, I think that Open AI has 

324
00:18:10,030 --> 00:18:15,030
attempted to mitigate that 
ability by through reinforcement

325
00:18:15,030 --> 00:18:18,390
learning essentially in Chatchi 
BT, so that it doesn't go kind 

326
00:18:18,390 --> 00:18:21,390
of off the rails, but it's not 
like the underlying model is not

327
00:18:21,390 --> 00:18:24,590
capable of that, right. 
It's just that Chatchi BT has 

328
00:18:24,590 --> 00:18:28,950
been fine-tuned for the Libs are
corralling us the status at 

329
00:18:28,950 --> 00:18:32,830
Chatchi at Open. 
AI Well, it's yeah, I think it's

330
00:18:32,830 --> 00:18:35,520
a really interesting. 
Kind of thing. 

331
00:18:35,520 --> 00:18:40,640
That Open AI has decided that 
they needed to do this right? 

332
00:18:40,640 --> 00:18:44,040
And Sam Allman has talked in the
past about how in the future 

333
00:18:44,400 --> 00:18:48,320
perhaps we will all be able to 
edit the configurations of 

334
00:18:48,320 --> 00:18:52,360
ChatGPT to be able to do have 
take off the training wheels, 

335
00:18:52,360 --> 00:18:53,360
right? 
Or something? 

336
00:18:53,440 --> 00:18:57,680
Is this whole wave powered by 
Open Ai's ChatGPT? 

337
00:18:57,680 --> 00:19:01,870
Is that the cool thing? 
And everything else is we're we 

338
00:19:01,870 --> 00:19:03,510
didn't invest early enough in 
opening. 

339
00:19:03,510 --> 00:19:07,630
I credit what I think Khosla 
Ventures is first in we didn't 

340
00:19:07,630 --> 00:19:10,670
invest early enough. 
We need to get a shot on goal. 

341
00:19:10,670 --> 00:19:12,350
We'll invest in another 
foundation model. 

342
00:19:12,350 --> 00:19:15,270
Do you think it's really ChatGPT
or bust? 

343
00:19:17,110 --> 00:19:21,470
Are you asking I guess? 
Yeah, I mean one question I I 

344
00:19:21,470 --> 00:19:24,390
kind of think yes, I mean I the 
way you phrase it I guess is 

345
00:19:24,390 --> 00:19:28,030
offers room for for wiggle room 
or argument, but I think that. 

346
00:19:28,680 --> 00:19:30,440
I think to your point, I think 
text generation and image 

347
00:19:30,440 --> 00:19:34,520
generation are the sort of aha 
moments that we've experienced 

348
00:19:34,520 --> 00:19:36,760
in the last year. 
I think in particular if you 

349
00:19:36,760 --> 00:19:40,880
look at what people are really 
using chat for or text ChatGPT 

350
00:19:40,880 --> 00:19:43,920
for, I think it's like 
essentially cheating on 

351
00:19:43,920 --> 00:19:46,450
homework. 
Cheating on office work. 

352
00:19:46,810 --> 00:19:47,410
Cheating. 
Cheating on? 

353
00:19:47,850 --> 00:19:50,130
The summarizing cheating in the.
Yeah, exactly. 

354
00:19:50,130 --> 00:19:51,610
That's kind of the point. 
Yeah, exactly. 

355
00:19:51,610 --> 00:19:53,010
OK. 
You know, maybe there's yeah 

356
00:19:53,370 --> 00:19:56,310
work there's like. 
Cheating on homework, quote UN 

357
00:19:56,310 --> 00:19:58,050
quote. 
Being efficient at office work, 

358
00:19:58,090 --> 00:20:01,370
summarizing long pieces of text 
and then I think basically sex 

359
00:20:01,370 --> 00:20:03,770
bot chat, which we can talk 
about more, is evolved into a 

360
00:20:03,770 --> 00:20:05,610
number of opportunities for 
different companies. 

361
00:20:05,960 --> 00:20:08,040
And then I think you know the 
image side to your point is the 

362
00:20:08,040 --> 00:20:11,120
other big thing, creating art, 
creating video, creating 

363
00:20:11,280 --> 00:20:13,520
potentially 3D models. 
And those always have the 

364
00:20:13,520 --> 00:20:16,600
tendency to go really viral 
because images are easy to share

365
00:20:16,600 --> 00:20:18,120
on social media. 
And so if you create a 

366
00:20:18,120 --> 00:20:21,440
particularly compelling image 
using generative AI then it can 

367
00:20:21,440 --> 00:20:23,840
really, you know, go across 
social media super fast. 

368
00:20:23,840 --> 00:20:27,520
So yeah, I think I think I would
struggle to think of a real 

369
00:20:27,840 --> 00:20:31,800
breakout use case that isn't 
essentially encapsulated in 

370
00:20:31,800 --> 00:20:35,940
chat, TBT and and Dolly or at 
least isn't just a one of those 

371
00:20:35,940 --> 00:20:39,180
things on steroids, but I'm 
probably not thinking of 

372
00:20:39,180 --> 00:20:41,700
something. 
I would just potentially add and

373
00:20:41,700 --> 00:20:45,540
it definitely fits into the 
office work use case, but maybe 

374
00:20:45,780 --> 00:20:49,020
more particularly around a 
coding and engineering, right, 

375
00:20:49,020 --> 00:20:53,100
like a Coilot style of coding. 
I mean Co pilots are potentially

376
00:20:53,100 --> 00:20:55,020
being added to lots of products 
as well. 

377
00:20:55,020 --> 00:20:57,780
But specifically, I think 
engineering is really 

378
00:20:57,780 --> 00:21:01,860
interesting because it starts to
there's a lot of hackers kind of

379
00:21:01,860 --> 00:21:04,950
working on. 
Coding agents and essentially 

380
00:21:05,910 --> 00:21:10,310
baby AGI, right, that can kind 
of run in loops to just get work

381
00:21:10,310 --> 00:21:11,910
done or build apps, that kind of
thing. 

382
00:21:11,910 --> 00:21:14,430
And I think we're still at the 
very early days of this, but it 

383
00:21:14,430 --> 00:21:20,590
is like an interesting use case.
To translate baby AGI, there's 

384
00:21:20,590 --> 00:21:23,790
like coding, there's assistance 
and then there's like autonomous

385
00:21:23,790 --> 00:21:24,630
agents, right? 
Exactly. 

386
00:21:24,630 --> 00:21:28,590
That's sort of a paradigm people
look at Here ChatGPT can. 

387
00:21:28,970 --> 00:21:31,090
Help you do your homework. 
Or it can do your homework. 

388
00:21:31,090 --> 00:21:35,370
Co pilot can help you code. 
Or you could literally have 

389
00:21:35,370 --> 00:21:37,610
something that is coding for a 
company. 

390
00:21:37,610 --> 00:21:41,130
And I think we see that 
framework come up again and 

391
00:21:41,130 --> 00:21:44,730
again, and when it feels like 
it'll be very disruptive when 

392
00:21:44,810 --> 00:21:47,730
you have agents like these 
things just doing it. 

393
00:21:48,290 --> 00:21:49,810
But I think that we don't do 
that yet. 

394
00:21:50,090 --> 00:21:53,330
I think the sort of optimistic 
take though on exactly the 

395
00:21:53,330 --> 00:21:55,370
argument we just made is that 
it's a little bit like the 

396
00:21:55,370 --> 00:21:58,010
industrial revolution for your 
brain, right. 

397
00:21:58,010 --> 00:21:59,690
It's, you know, it's the 
industrial revolution for your 

398
00:21:59,690 --> 00:22:01,330
brain. 
And that pretty much all the 

399
00:22:01,330 --> 00:22:04,570
inputs and outputs of the human 
mind are some form of text, 

400
00:22:04,570 --> 00:22:07,930
whether that's spoken or written
or some form of imagery, right. 

401
00:22:07,930 --> 00:22:10,250
Whether it's something you see 
or it's something that you 

402
00:22:10,250 --> 00:22:12,930
create, whether it's, you know, 
a drawing or or in a piece of 

403
00:22:12,930 --> 00:22:15,830
imaging software, right. 
And if you think those, 

404
00:22:15,830 --> 00:22:18,430
basically all the inputs and 
outputs of the human brain can 

405
00:22:18,430 --> 00:22:20,830
be encapsulated in some form of 
text and images. 

406
00:22:21,190 --> 00:22:25,710
If you create technology that 
makes it really easy to create 

407
00:22:25,710 --> 00:22:29,790
high quality and also interpret 
high quality text and images. 

408
00:22:29,790 --> 00:22:32,220
Right. 
You've kind of like, you've got 

409
00:22:32,220 --> 00:22:35,540
like 80% of the job done of what
the human brain can do. 

410
00:22:35,540 --> 00:22:36,980
Right. 
And to your point, there's this 

411
00:22:36,980 --> 00:22:39,020
distinction whether it's 
autonomous or it's helping you, 

412
00:22:39,020 --> 00:22:41,460
it's an assistant, whatever. 
But, you know, I, I think the 

413
00:22:41,460 --> 00:22:43,260
Industrial Revolution is 
interesting analogy because it 

414
00:22:43,260 --> 00:22:45,380
was like the first time. 
It was like you don't actually 

415
00:22:45,380 --> 00:22:47,420
have to sew this, like, shirt, 
right. 

416
00:22:47,420 --> 00:22:49,500
This, like, machine will sew it 
for you. 

417
00:22:49,500 --> 00:22:51,140
Right? 
Like, you can sit at this 

418
00:22:51,140 --> 00:22:53,740
machine and it'll be your 
assistant in sewing this shirt. 

419
00:22:53,740 --> 00:22:55,100
Right. 
And that's like a big deal, 

420
00:22:55,100 --> 00:22:56,460
right? 
Is the first time in human 

421
00:22:56,460 --> 00:22:58,900
history, like you don't have to 
actually sew the shirt, like, 

422
00:22:58,900 --> 00:23:00,020
without any help. 
Right. 

423
00:23:00,220 --> 00:23:03,540
And I think similarly like for 
all these different types of 

424
00:23:03,540 --> 00:23:06,740
work, whether it's homework or 
office jobs or legal work or you

425
00:23:06,740 --> 00:23:09,780
know, mathematical analysis or 
writing or podcasting or 

426
00:23:09,780 --> 00:23:11,380
whatever, it's okay. 
Well, for the first time ever, 

427
00:23:11,380 --> 00:23:13,940
you don't have to do all the 
work right, whether you're 

428
00:23:13,940 --> 00:23:16,780
assisted or it just does it 
itself. 

429
00:23:16,780 --> 00:23:22,340
Like it's kind of a game changer
because you have automation for 

430
00:23:22,340 --> 00:23:24,820
the human mind, for creativity 
in some fashion or another. 

431
00:23:24,820 --> 00:23:29,070
So that I think is like the 
really crazy optimistic take is 

432
00:23:29,070 --> 00:23:31,550
that we're at the beginning of 
the second Industrial Revolution

433
00:23:31,550 --> 00:23:33,710
and it's no longer physical, but
it's mental, right. 

434
00:23:33,950 --> 00:23:36,270
And I kind of believe that I 
would say I'm leaning that 

435
00:23:36,270 --> 00:23:37,790
direction based on where we are 
today. 

436
00:23:38,430 --> 00:23:41,550
Yeah, I continue to believe, but
it's mostly from my experience 

437
00:23:41,550 --> 00:23:46,230
with ChatGPT that it's just 
insane, amazing. 

438
00:23:46,230 --> 00:23:49,510
I mean, I mean, I feel like it's
great at like. 

439
00:23:50,170 --> 00:23:52,610
Writing a poem I I keep joking 
that people are going to write 

440
00:23:52,610 --> 00:23:54,650
all their vows with catching 
teeth, you know? 

441
00:23:55,050 --> 00:23:58,090
I I feel like these tasks were 
you know people are desperately 

442
00:23:58,090 --> 00:24:01,290
trying to get the same like 
style like groomsmen sort of 

443
00:24:01,290 --> 00:24:04,370
toast or whatever. 
It's it's great you know I feel 

444
00:24:04,370 --> 00:24:06,970
like it can be sort of creative 
it. 

445
00:24:07,090 --> 00:24:10,570
But yeah, I mean to me the the 
counterpoint is just it's just 

446
00:24:10,570 --> 00:24:15,050
so hard to know just like it was
hard with self driving cars to 

447
00:24:15,050 --> 00:24:18,970
know when they would be complete
and the completeness matters. 

448
00:24:19,640 --> 00:24:23,120
The extent to which completeness
matters with a chat sort of 

449
00:24:23,120 --> 00:24:27,000
interface, because I think 
humans have been enticed, like 

450
00:24:27,000 --> 00:24:30,720
we were saying like decades ago 
with chat interfaces and we're 

451
00:24:30,720 --> 00:24:33,560
like, you're almost there. 
I've only it is like I have 

452
00:24:33,560 --> 00:24:36,680
stopped using Chat GP. 
Like do you guys in your daily 

453
00:24:36,680 --> 00:24:39,720
life use generative AI for 
anything? 

454
00:24:41,060 --> 00:24:45,980
I frequently use ChatGPT, but 
it's really I would say not for 

455
00:24:45,980 --> 00:24:48,380
productivity purposes. 
Maybe occasionally at. 

456
00:24:48,380 --> 00:24:52,980
Work just like. 
I mean, I'm not like talking to 

457
00:24:52,980 --> 00:24:57,260
characters, but I am using it to
just brainstorm ideas. 

458
00:24:57,260 --> 00:24:59,100
Like I'll yesterday. 
I just. 

459
00:24:59,100 --> 00:25:03,820
I came up with this prompt that 
was like create a timeline of a 

460
00:25:03,820 --> 00:25:06,780
fictional historical world with 
the depth of. 

461
00:25:07,190 --> 00:25:11,990
Westeros or Middle Earth and but
you know, and I essentially got,

462
00:25:12,190 --> 00:25:15,150
you know, 10,000 years of 
history of a faith. 

463
00:25:16,630 --> 00:25:19,430
I thought I was super cool and I
couldn't couldn't have done that

464
00:25:19,430 --> 00:25:22,670
before. 
So I guess and then sometimes 

465
00:25:22,670 --> 00:25:27,110
I'll just ask it, you know, for 
ideas for new products or new 

466
00:25:27,110 --> 00:25:30,910
companies or it I just to see 
what kind of level of creativity

467
00:25:30,910 --> 00:25:33,150
it is capable of. 
I think that's what's really 

468
00:25:33,150 --> 00:25:34,830
interesting to me. 
I think we all agree. 

469
00:25:35,260 --> 00:25:38,380
That there is creativity 
occurring that is creating 

470
00:25:38,380 --> 00:25:40,140
novel. 
Well, I guess we don't. 

471
00:25:40,500 --> 00:25:42,020
Not everyone agrees with this, 
right? 

472
00:25:42,020 --> 00:25:46,380
But that it's not solely capable
of regurgitating information. 

473
00:25:46,380 --> 00:25:51,940
But from my perspective it seems
very capable of creating new 

474
00:25:51,940 --> 00:25:55,180
original ideas when I play 
around with it. 

475
00:25:55,180 --> 00:25:58,820
And I believe there have been 
papers proving this that. 

476
00:25:59,060 --> 00:26:01,340
Well, Microsoft came out with 
one, right? 

477
00:26:01,340 --> 00:26:03,460
That said, there were like 
sparks of like. 

478
00:26:04,090 --> 00:26:07,930
General intelligence or. 
Whatever, I don't remember that 

479
00:26:07,930 --> 00:26:13,010
specifically, but I did see a 
paper that there's a common test

480
00:26:13,010 --> 00:26:16,090
of creativity, right? 
Where you will essentially ask 

481
00:26:16,090 --> 00:26:18,810
people for. 
I guess one example they gave is

482
00:26:18,810 --> 00:26:21,890
you ask cases around, what would
you do? 

483
00:26:21,890 --> 00:26:25,090
What are name 100 use cases of 
this paper clip? 

484
00:26:25,090 --> 00:26:28,050
Or name 100 things you could do 
with this, a rubber band or 

485
00:26:28,050 --> 00:26:30,930
something, right? 
And then they essentially grade 

486
00:26:30,930 --> 00:26:32,910
the ideas. 
And I thought that was pretty 

487
00:26:32,910 --> 00:26:36,710
interesting test of creativity 
and and essentially ChatGPT is 

488
00:26:36,710 --> 00:26:39,990
performing, you know, better 
than most humans including most 

489
00:26:39,990 --> 00:26:43,030
like MBA students. 
So you know I think that there 

490
00:26:43,030 --> 00:26:46,470
are ways to like start to test 
this, but it's underrated the 

491
00:26:46,470 --> 00:26:49,310
level of creativity, not just 
that it's oh, it's cool that it 

492
00:26:49,310 --> 00:26:52,190
can create a poem, right. 
It can create more creative 

493
00:26:52,190 --> 00:26:54,950
poems than like poem poetry 
authors, right. 

494
00:26:54,950 --> 00:26:58,670
Like that kind of thing gets to.
Be right, it's easy for humans, 

495
00:26:58,830 --> 00:27:00,360
he. 
We just sort of like sticking 

496
00:27:00,360 --> 00:27:02,720
our nose up at it. 
I sort of roll you know change 

497
00:27:02,720 --> 00:27:04,680
the goal posts basically. 
But then yeah, like you're 

498
00:27:04,680 --> 00:27:08,120
saying, people will give it 
these tests like what can an MBA

499
00:27:08,120 --> 00:27:10,600
student do and what can ChatGPT 
do. 

500
00:27:10,680 --> 00:27:14,600
And like people are, you know 
are impressed I think blind with

501
00:27:14,600 --> 00:27:21,480
the ChatGPT response just to I 
was yeah, Microsoft in May 20 of

502
00:27:21,480 --> 00:27:26,480
this year said they saw. 
Sparks of general intelligence 

503
00:27:26,480 --> 00:27:30,160
basically in a research paper I 
think it almost any task right 

504
00:27:30,160 --> 00:27:35,360
now ChatGPT 4 is at the level of
a pretty solid college student 

505
00:27:35,600 --> 00:27:38,920
like maybe in a minus college 
student in almost any field 

506
00:27:38,920 --> 00:27:40,480
which is sort of mind boggling 
right. 

507
00:27:40,680 --> 00:27:44,200
And like how many of us are like
at the level of a minus college 

508
00:27:44,200 --> 00:27:47,080
student in in more than like 
maybe one or two things you know

509
00:27:47,080 --> 00:27:49,040
and it's at the level of an A 
minus college student at 

510
00:27:49,040 --> 00:27:51,440
everything and it can serve 
millions of requests like at any

511
00:27:51,440 --> 00:27:54,530
given time right. 
So it's like a scaled A minus 

512
00:27:54,530 --> 00:27:57,810
college student at basically 
everything and then particularly

513
00:27:57,810 --> 00:27:59,730
these creative tasks as you're 
saying, I feel like the one 

514
00:27:59,730 --> 00:28:03,370
thing that holds it back is the 
need to be accurate, right? 

515
00:28:03,370 --> 00:28:08,200
And and often it invents facts 
or it's sort of like misaligns 

516
00:28:08,200 --> 00:28:11,160
realworld concepts in ways that 
aren't really realistic but in a

517
00:28:11,160 --> 00:28:14,680
purely creative endeavor, like 
poetry or like creating artwork 

518
00:28:14,680 --> 00:28:16,920
or coming up with ideas for what
to do with the paper clip, 

519
00:28:16,920 --> 00:28:18,720
right. 
Like when it sort of has doesn't

520
00:28:18,720 --> 00:28:21,360
have to be anchored to any sort 
of like really hard facts. 

521
00:28:21,440 --> 00:28:23,320
It's unbelievable. 
I mean it's it's better than 

522
00:28:23,320 --> 00:28:26,040
almost anyone in the world. 
To what extent do you think this

523
00:28:26,040 --> 00:28:27,880
is all exciting? 
Because we're like, oh, we're on

524
00:28:27,880 --> 00:28:31,560
the cusp of general 
intelligence, right? 

525
00:28:31,680 --> 00:28:34,440
Like artificial General 
Intelligence, AGI, this idea 

526
00:28:34,440 --> 00:28:38,190
that you know it's. 
There are different tests for 

527
00:28:38,190 --> 00:28:40,990
him, but the idea that it's 
smarter than a human being to 

528
00:28:41,150 --> 00:28:44,030
bring to bring it back to the 
conference, I mean I would echo 

529
00:28:44,030 --> 00:28:47,110
what Ali Goatsy said on stage 
which is that I just think by so

530
00:28:47,110 --> 00:28:49,070
many. 
Yeah, back in March at our last 

531
00:28:49,070 --> 00:28:51,150
conference and he'll be back for
the next conference. 

532
00:28:51,150 --> 00:28:53,590
Stay tuned. 
Yeah, he he said basically, 

533
00:28:53,590 --> 00:28:56,150
look, I mean I think it is 
general intelligence already. 

534
00:28:56,150 --> 00:29:01,190
Like it already is, you know as 
good or better than 99.9% of 

535
00:29:01,190 --> 00:29:02,950
humans at almost any task you 
can throw at it, right? 

536
00:29:02,950 --> 00:29:05,700
I mean, how can you not say 
that's like general 

537
00:29:05,700 --> 00:29:08,140
intelligence, right? 
I just think that holding it to 

538
00:29:08,140 --> 00:29:10,900
the standard where it has to be 
100% accurate about everything 

539
00:29:10,900 --> 00:29:14,660
or it has to be able to go do 
stuff on its own, which isn't 

540
00:29:14,660 --> 00:29:16,260
really that hard of a technical 
challenge. 

541
00:29:16,340 --> 00:29:19,260
I think that's just an it's sort
of like goal posts moving 

542
00:29:19,260 --> 00:29:21,820
because I think people are like 
afraid of the idea that we have 

543
00:29:21,820 --> 00:29:24,100
created something that's like a 
smarter, smarter than a human 

544
00:29:24,100 --> 00:29:27,500
right. 
Like, clearly five years ago, if

545
00:29:27,500 --> 00:29:30,180
you had told someone that we're 
going to have an AI chap out, 

546
00:29:30,180 --> 00:29:32,140
they can do everything that Chad
JPT can do. 

547
00:29:32,840 --> 00:29:36,720
You know, pass the bar, pass AP 
exams, create beautiful artwork,

548
00:29:36,920 --> 00:29:39,720
talk to you, you know, write 
poetry by the Dodgers. 

549
00:29:39,840 --> 00:29:43,720
But you'd be like so being. 
Blase Well, that sounds like 

550
00:29:43,720 --> 00:29:45,920
pretty freaking close to general
intelligence to me. 

551
00:29:45,920 --> 00:29:47,280
What more do you want from this 
thing? 

552
00:29:47,280 --> 00:29:50,520
I just think that I don't. 
Know I wanted to have a through 

553
00:29:50,520 --> 00:29:54,320
line of reasoning where it seems
to be a thinking being where I 

554
00:29:54,320 --> 00:29:58,470
can explain why it generated the
answers it has, you know, Yeah, 

555
00:29:58,510 --> 00:30:00,510
I mean but can humans really 
explain why they generate 

556
00:30:00,510 --> 00:30:01,630
answers? 
In most cases? 

557
00:30:01,630 --> 00:30:04,110
I just think it's holding it to 
a really high standard that most

558
00:30:04,110 --> 00:30:06,630
humans cannot meet. 
And so I would argue is it at 

559
00:30:06,630 --> 00:30:09,430
the level of a human in almost 
every area? 

560
00:30:09,710 --> 00:30:11,270
Absolutely. 
I mean, I would say it's in the 

561
00:30:11,270 --> 00:30:14,150
top 1% of humans in almost any 
area you throw at it. 

562
00:30:14,630 --> 00:30:19,310
Yeah, I guess I would agree with
that mostly other than the main 

563
00:30:19,310 --> 00:30:25,710
area, I think it starts to fail 
or deteriorate is, is when you 

564
00:30:25,990 --> 00:30:28,960
kind of create too much. 
Memory or context right? 

565
00:30:28,960 --> 00:30:32,960
Like essentially humans have 
this amazing ability to recall 

566
00:30:33,440 --> 00:30:36,560
information throughout from 
their entire life right? 

567
00:30:36,560 --> 00:30:43,400
And to like sort of be able to 
maintain the context of a hour 

568
00:30:43,400 --> 00:30:48,360
long multi hour long movie or 
20,000 page book right? 

569
00:30:48,360 --> 00:30:53,120
And sometimes it feels like 
ChatGPT it's stretching to that 

570
00:30:53,120 --> 00:30:55,600
with the amount of tokens, 
context, windows it can 

571
00:30:56,360 --> 00:30:58,800
understand but. 
It really does show 

572
00:30:58,800 --> 00:31:01,480
deterioration as you add more 
and more context, and then there

573
00:31:01,480 --> 00:31:05,280
is a actual hard limit of 
context you can include in your 

574
00:31:05,280 --> 00:31:09,360
prompts. 
I hate the fact that it doesn't 

575
00:31:09,360 --> 00:31:12,280
remember when we've talked 
before, even the same thread, 

576
00:31:12,280 --> 00:31:14,840
and it just starts lying about 
what it said before. 

577
00:31:14,840 --> 00:31:19,680
There are parts of what human 
beings do that humans would 

578
00:31:19,680 --> 00:31:21,640
never do This sort of just 
totally make. 

579
00:31:21,640 --> 00:31:25,560
I mean, some would, but just 
like totally bullshitting, like 

580
00:31:25,560 --> 00:31:26,880
when it's like, why? 
Yeah. 

581
00:31:27,400 --> 00:31:30,560
Anyway, I wanted to get into 
more of the business question 

582
00:31:30,880 --> 00:31:35,880
from this sort of same framing 
with is it all Open AI chat CBT 

583
00:31:35,960 --> 00:31:39,600
On the one hand, you know, I I 
feel like we're seeing you know 

584
00:31:39,600 --> 00:31:42,680
people do these like Elo tests 
where they like compare 

585
00:31:42,680 --> 00:31:46,880
different foundation models and 
we do see like Llama, like 

586
00:31:46,880 --> 00:31:50,880
Facebook's model, open source 
model and other models like sort

587
00:31:50,880 --> 00:31:55,000
of being competitive at times 
with chat CBT though 4.5. 

588
00:31:55,530 --> 00:31:58,690
ChatGPT 4.5 remains the gold 
standard. 

589
00:31:59,050 --> 00:32:01,930
I guess I'm curious, it's sort 
of like a 2/1. 

590
00:32:01,930 --> 00:32:05,770
Do you think other people will 
catch up and like how much do 

591
00:32:05,770 --> 00:32:07,810
you think there's sort of a Moat
here? 

592
00:32:07,810 --> 00:32:11,770
Like how much do you think being
ahead slightly or like having 

593
00:32:11,770 --> 00:32:16,570
been sort of the the one that 
consumers know about is is like 

594
00:32:16,570 --> 00:32:19,690
a Moat, like how, how, how 
defended do you think they are 

595
00:32:19,690 --> 00:32:24,050
with their position? 
That's a really hard question. 

596
00:32:24,400 --> 00:32:26,000
I mean, first of all, they're 
basically a subsidiary of 

597
00:32:26,000 --> 00:32:28,160
Microsoft, right. 
So you're asking for. 

598
00:32:28,160 --> 00:32:29,200
You're asking for. 
Yeah, Yeah. 

599
00:32:29,560 --> 00:32:33,200
You're asking, you know, are 
they going to be a huge, you 

600
00:32:33,200 --> 00:32:35,520
know, strategic value add to 
Microsoft going forward? 

601
00:32:35,520 --> 00:32:38,840
I think obviously, yes, right. 
I mean, are you asking is 

602
00:32:38,840 --> 00:32:42,280
ChatGPT always going to be the 
gold standard for text 

603
00:32:42,360 --> 00:32:44,800
generation models? 
Like, I don't know, it seems 

604
00:32:44,800 --> 00:32:47,120
like everyone's catching up. 
To your point, it also seems 

605
00:32:47,120 --> 00:32:49,880
like they have the best people 
and they're moving the fastest 

606
00:32:49,880 --> 00:32:52,200
on releasing new things. 
So they'll always stay, you know

607
00:32:52,200 --> 00:32:55,030
6 to 12 months ahead. 
You know, does a Moat really 

608
00:32:55,030 --> 00:32:58,630
matter in this context again 
where like you're, you know, 

609
00:32:58,630 --> 00:33:01,110
you're you're the best at least 
and you're ahead of everyone 

610
00:33:01,110 --> 00:33:03,310
else. 
And again you're a subsidiary of

611
00:33:03,310 --> 00:33:05,870
Microsoft, so there's no real 
business benefit to, you know, 

612
00:33:05,870 --> 00:33:08,710
winning anyway to you know it 
doesn't. 

613
00:33:09,430 --> 00:33:10,590
It's a sort of hard business 
question. 

614
00:33:10,630 --> 00:33:13,750
I think what's more interesting 
to ask is if all the knockoffs 

615
00:33:13,750 --> 00:33:16,710
or the competitors or the 
various image models or the 

616
00:33:16,710 --> 00:33:19,670
various, you know, versions of 
ChatGPT that are out there are 

617
00:33:19,670 --> 00:33:22,140
going to be successful. 
Because I think ChatGPT will 

618
00:33:22,140 --> 00:33:24,580
clearly be successful, I think 
in some contexts, I mean it's 

619
00:33:24,580 --> 00:33:26,740
going to be in Microsoft Word 10
years from now, right? 

620
00:33:27,060 --> 00:33:29,900
But what about anthropic? 
What about Llama? 

621
00:33:29,900 --> 00:33:32,740
What about Google Gemini or 
Lambda? 

622
00:33:32,740 --> 00:33:35,380
Or what about, you know, 
whatever Amazon's cooking up, 

623
00:33:35,380 --> 00:33:37,500
what about a million startups 
got funded in the last 10 

624
00:33:37,500 --> 00:33:39,220
seconds, right? 
I mean, I think those are like 

625
00:33:39,380 --> 00:33:43,100
more interesting questions 
because it's to your to the 

626
00:33:43,100 --> 00:33:46,020
earlier discussion it seems like
a pretty commoditized concept 

627
00:33:46,020 --> 00:33:49,270
like chat with a large language 
model and unless someone can be 

628
00:33:49,350 --> 00:33:52,310
way better or have a very 
different business strategy than

629
00:33:52,310 --> 00:33:56,070
ChatGPT, it's hard to see where 
they're gonna win, right. 

630
00:33:56,070 --> 00:33:58,350
And so people are trying 
different angles on this 

631
00:33:58,350 --> 00:34:00,870
concept, but the concept itself 
is not that different in 

632
00:34:00,870 --> 00:34:04,710
different companies. 
I do think this really gets to 

633
00:34:04,710 --> 00:34:09,310
the question of will there be 
one best AI essentially 1 

634
00:34:09,310 --> 00:34:13,389
ChatGPT like that we all use as 
our personal assistant, right? 

635
00:34:13,389 --> 00:34:16,830
And that is the most general, 
most high powerful model of. 

636
00:34:17,219 --> 00:34:20,260
Most highly intelligent model 
right that exists in the world 

637
00:34:20,659 --> 00:34:24,540
you know because it's so 
generalizable and then you know 

638
00:34:24,540 --> 00:34:26,139
I think there's a good. 
I think there's that's 

639
00:34:26,139 --> 00:34:28,260
plausible. 
I think that certainly you know 

640
00:34:28,260 --> 00:34:33,130
we use. 11 you know browser and 
one e-mail client, I mean I 

641
00:34:33,130 --> 00:34:35,210
don't know like we don't, we're 
not switching between them a 

642
00:34:35,210 --> 00:34:38,530
lot, right. 
But I think it starts to ask you

643
00:34:38,530 --> 00:34:41,969
know, will that model also be 
better at all other you know 

644
00:34:42,210 --> 00:34:44,730
vertical tasks as well? 
I don't that seems harder to 

645
00:34:44,730 --> 00:34:45,690
believe right? 
Like that. 

646
00:34:45,690 --> 00:34:50,810
It will also be the best model 
at reading legal documents and. 

647
00:34:50,810 --> 00:34:53,810
Being I was just gonna bring up 
case tax, sold the Thompson 

648
00:34:53,810 --> 00:34:57,090
Reuters except for hundreds of 
millions of dollars and. 

649
00:34:57,430 --> 00:35:00,950
They were like, I mean much more
sophisticated than ChatGPT 

650
00:35:00,950 --> 00:35:04,750
wrapper, but they were using 
ChatGPT largely. 

651
00:35:04,750 --> 00:35:09,670
Sure, regardless of what their 
specific model was doing, it 

652
00:35:09,670 --> 00:35:14,270
seems possible you could put a 
ton of compute and specifically 

653
00:35:14,790 --> 00:35:19,350
trained legal data into a model 
that would outperform ChatGPT at

654
00:35:19,350 --> 00:35:22,990
that use case of the law, right?
Or same for medicine or 

655
00:35:22,990 --> 00:35:24,410
something, but. 
I don't know. 

656
00:35:24,410 --> 00:35:27,970
I'm not 100% confident in that. 
I think that it's very possible,

657
00:35:27,970 --> 00:35:32,610
just with the you know that chat
TBT as a foundational general 

658
00:35:32,610 --> 00:35:36,050
model will just always be better
at all of those key use cases. 

659
00:35:36,490 --> 00:35:38,530
I mean Jason Warner, you know 
who's speaking in the 

660
00:35:38,530 --> 00:35:42,410
conference, who's now the CEO of
Poolside is betting that he can 

661
00:35:42,410 --> 00:35:44,730
build a foundation model for 
code. 

662
00:35:44,730 --> 00:35:47,530
Obviously, you know, yeah, you 
can see a bazillion companies 

663
00:35:47,530 --> 00:35:50,930
that are sort of trying to be 
like I will be a foundation 

664
00:35:50,930 --> 00:35:54,140
model. 
Yeah, for, for XI mean I think 

665
00:35:54,140 --> 00:35:55,940
it's, I think it's sort of 
instructive to think back to 

666
00:35:55,940 --> 00:35:58,740
like social networks, right. 
And for me, I always find it 

667
00:35:58,740 --> 00:36:04,220
interesting analogy of, OK, you 
know, 1520 years ago would you 

668
00:36:04,220 --> 00:36:08,420
predicted there would be a 
social network for work and 

669
00:36:08,420 --> 00:36:11,220
there would be a social network 
for gaming and there would be a 

670
00:36:11,220 --> 00:36:13,100
social network for people under 
25. 

671
00:36:13,100 --> 00:36:14,740
That's Snapchat. 
And there'd be a social network 

672
00:36:14,740 --> 00:36:17,740
for people between 25 and 45 and
that's Instagram and there'd be 

673
00:36:17,740 --> 00:36:19,740
a social network for people who 
are over 45. 

674
00:36:19,740 --> 00:36:23,340
And that's Facebook blue, right.
I mean like and and and it's a, 

675
00:36:24,140 --> 00:36:27,100
it's sort of interesting to 
think about what the cleavages 

676
00:36:27,100 --> 00:36:29,660
are in the the user need. 
Right. 

677
00:36:29,660 --> 00:36:33,060
And in social networking, it's 
often just age cohorts sort of 

678
00:36:33,060 --> 00:36:34,540
build these network effects with
each other. 

679
00:36:34,780 --> 00:36:37,380
But then there's also this 
LinkedIn, which is, oh, actually

680
00:36:37,380 --> 00:36:40,860
work is a completely different 
social concept in your life that

681
00:36:40,860 --> 00:36:42,300
you need to keep separate from 
everything else. 

682
00:36:42,460 --> 00:36:45,260
And then you have Discord, which
is like actually gaming is like 

683
00:36:45,260 --> 00:36:47,540
a completely different concept 
that you need to keep distinct 

684
00:36:47,540 --> 00:36:48,500
from everything else. 
Right. 

685
00:36:48,840 --> 00:36:50,640
And I don't think it would have 
been that easy to predict those 

686
00:36:50,640 --> 00:36:52,000
things. 
I mean maybe like Reed Hoffman 

687
00:36:52,000 --> 00:36:53,600
will say, it was super easy to 
predict, right? 

688
00:36:53,880 --> 00:36:56,200
But but I think with these 
foundational models, it's 

689
00:36:56,200 --> 00:36:57,680
similar, right? 
It's, you know, there probably 

690
00:36:57,680 --> 00:37:00,280
aren't really network effects 
other than just like who can eat

691
00:37:00,280 --> 00:37:02,560
the most data the fastest, 
right, Which seems like it's 

692
00:37:02,560 --> 00:37:05,520
going to be GPT. 
And so the data, quote UN quote,

693
00:37:05,520 --> 00:37:08,080
network effects, are the data 
scale effects are probably 

694
00:37:08,080 --> 00:37:10,800
always going to be 1 by 1 by 
GPT, right. 

695
00:37:11,080 --> 00:37:14,960
Then the question is, are there 
other use cases where there's 

696
00:37:14,960 --> 00:37:17,240
some kind of network effect or 
there's some sort of different 

697
00:37:17,240 --> 00:37:19,560
business concepts? 
You know, Facebook seems to be 

698
00:37:19,560 --> 00:37:21,440
going the angle of we're going 
to open source this all. 

699
00:37:21,440 --> 00:37:23,400
So people will just build it 
into all these things like it's 

700
00:37:23,400 --> 00:37:26,680
Linux back in the day and 
that'll be the way we win is 

701
00:37:26,680 --> 00:37:29,080
that it'll be the free open 
source version and you can just 

702
00:37:29,080 --> 00:37:31,200
throw it into everything if you 
want to, which I think it's a 

703
00:37:31,200 --> 00:37:33,200
pretty interesting like business
concept, right. 

704
00:37:33,400 --> 00:37:35,800
And then I don't know that much 
about the, you know, enthropics 

705
00:37:35,800 --> 00:37:37,720
or the pool sides or whatever. 
You know, how are they going to 

706
00:37:37,720 --> 00:37:40,040
win? 
What is the cleavage in the the 

707
00:37:40,040 --> 00:37:43,440
use case or the way these models
are built that is going to allow

708
00:37:43,440 --> 00:37:45,160
someone other than ChatGPT to 
win? 

709
00:37:45,160 --> 00:37:47,520
Because it seems like they're 
gonna win on data and they're 

710
00:37:47,520 --> 00:37:50,520
gonna win on hardware. 
So like, where are you gonna win

711
00:37:50,520 --> 00:37:52,280
if you're not them? 
I guess, and I'm sure everyone 

712
00:37:52,280 --> 00:37:53,360
has an answer to this, but 
that's me. 

713
00:37:53,360 --> 00:37:55,880
It's the hard question. 
We've spent a lot of time 

714
00:37:55,880 --> 00:37:57,720
talking about the foundation 
models. 

715
00:37:57,880 --> 00:38:00,480
I mean, Chad, I mean open a Eye 
is sort of a combination, right,

716
00:38:00,480 --> 00:38:02,680
where it's like they have the 
foundation model, they apply it 

717
00:38:03,040 --> 00:38:06,400
to these use cases. 
I mean people talk a lot about 

718
00:38:06,560 --> 00:38:09,240
you know, applications and like 
infrastructure, right. 

719
00:38:09,240 --> 00:38:12,720
I mean, there are companies, 
they're all these sort of wonky 

720
00:38:12,720 --> 00:38:15,290
companies like. 
The vector databases have been 

721
00:38:15,290 --> 00:38:17,250
super hot of the last couple 
months, right? 

722
00:38:17,290 --> 00:38:21,650
People talk about companies like
Pine Cone, We V8 there. 

723
00:38:21,650 --> 00:38:23,970
There's like a whole list of 
them which are just like trying 

724
00:38:23,970 --> 00:38:27,610
to organize your data in a 
better way to get it into 

725
00:38:27,810 --> 00:38:30,290
foundation models. 
I wanted to get to the actual 

726
00:38:30,290 --> 00:38:34,570
like applications right and you 
sort of Max earlier referenced 

727
00:38:34,930 --> 00:38:39,330
the sort of chat bot application
which also goes back to you 

728
00:38:39,410 --> 00:38:43,530
know, your both of your early 
days at volley trying to build. 

729
00:38:44,960 --> 00:38:47,600
Chat bots. 
I'm curious getting away from 

730
00:38:47,600 --> 00:38:51,400
who has the technical expertise,
like what you think is 

731
00:38:51,400 --> 00:38:54,000
interesting in the sort of what 
its character. 

732
00:38:54,000 --> 00:38:57,920
Is it replica? 
You guys know this world much 

733
00:38:57,920 --> 00:39:00,600
better than I do. 
So you're sort of getting at do 

734
00:39:00,600 --> 00:39:03,640
we feel these are successful use
cases or? 

735
00:39:03,720 --> 00:39:04,160
Right. 
Yeah. 

736
00:39:04,160 --> 00:39:06,320
Yeah. 
Do you think there's sustained 

737
00:39:06,320 --> 00:39:10,680
promise there or what do you see
in terms of people actually 

738
00:39:10,680 --> 00:39:14,690
using AI and applications that? 
That excites you. 

739
00:39:15,570 --> 00:39:20,930
There's a challenge with these 
models being accurate, obviously

740
00:39:20,930 --> 00:39:24,690
100% of the time, and you can 
debate whether that's necessary 

741
00:39:24,690 --> 00:39:27,570
or not, right? 
To consider it True general 

742
00:39:27,570 --> 00:39:29,850
intelligence. 
But in the entertainment space, 

743
00:39:29,850 --> 00:39:33,490
like, it's just less of a 
problem, like conversing with a 

744
00:39:33,970 --> 00:39:36,090
fictional character or 
historical character, right? 

745
00:39:36,090 --> 00:39:38,730
Like these things don't need to 
be extremely accurate because 

746
00:39:38,730 --> 00:39:41,490
they're essentially 
entertainment anyway, I think. 

747
00:39:42,530 --> 00:39:47,170
I guess I always come back to 
some lessons I've learned in the

748
00:39:47,170 --> 00:39:52,170
gaming world, where there's a 
clear difference between an 

749
00:39:52,170 --> 00:39:55,330
entertaining demo that is fun to
do and fun to play with. 

750
00:39:55,330 --> 00:39:58,450
And I think we've seen a lot of 
those that are actually really 

751
00:39:58,450 --> 00:40:02,570
amazing and impressive demos, 
but they don't have long staying

752
00:40:02,570 --> 00:40:06,010
power or retention, right? 
So thinking of all of the apps 

753
00:40:06,010 --> 00:40:10,050
that create fun. 
Photos that put you into their 

754
00:40:10,210 --> 00:40:12,610
photos or. 
Yeah, what's can of soup. 

755
00:40:13,370 --> 00:40:15,170
Can of soup is the. 
Latest I keep meaning to write 

756
00:40:15,170 --> 00:40:17,490
about them. 
They're like super buzzy, right?

757
00:40:17,650 --> 00:40:19,570
Yeah. 
And you can put yourself in AI 

758
00:40:19,570 --> 00:40:21,130
generated photos with your 
friends. 

759
00:40:21,130 --> 00:40:23,050
I think it's amazing. 
It's really cool. 

760
00:40:23,050 --> 00:40:26,050
But the question is, does that 
really have staying power? 

761
00:40:26,050 --> 00:40:28,650
Do you build a social network 
around it in order to make it 

762
00:40:28,730 --> 00:40:30,330
have staying power and network 
effects? 

763
00:40:30,650 --> 00:40:33,530
These are real challenges to 
create like a sustainable 

764
00:40:33,530 --> 00:40:37,050
business and startup that 
achieves a great outcome. 

765
00:40:37,050 --> 00:40:40,940
So and similarly with talking to
characters, character AI I think

766
00:40:40,940 --> 00:40:44,140
clearly has some product market 
fit there with people wanting to

767
00:40:44,140 --> 00:40:45,780
come back and talk to those 
characters. 

768
00:40:46,060 --> 00:40:48,260
But I think what Max and I like 
to think about is how do you 

769
00:40:48,260 --> 00:40:50,260
even build more retention around
that? 

770
00:40:50,260 --> 00:40:53,620
How do you build like game 
mechanics or features in the 

771
00:40:53,620 --> 00:40:57,700
concept of a virtual pet that 
you might, we know from looking 

772
00:40:57,700 --> 00:41:01,620
at mobile gaming and previous 
eras of gaming that you can 

773
00:41:01,620 --> 00:41:05,620
build long running retention 
into that if you add game 

774
00:41:05,620 --> 00:41:09,770
mechanics to the experience. 
So yeah, I'm less bullish on 

775
00:41:09,770 --> 00:41:14,250
just like general like character
conversations that don't have 

776
00:41:14,250 --> 00:41:17,530
any are sort of aimless or don't
have an end point or a purpose 

777
00:41:17,570 --> 00:41:21,210
and more bullish on, you know, 
how do we bring those characters

778
00:41:21,210 --> 00:41:25,170
or Npc's into a gaming context 
with like normal gaming 

779
00:41:25,170 --> 00:41:28,210
objectives. 
I do think in interacting with 

780
00:41:28,210 --> 00:41:30,530
these underlying characters, I 
think to your point, I think a 

781
00:41:30,530 --> 00:41:33,170
lot of the real value is when 
you start losing sight of it 

782
00:41:33,170 --> 00:41:36,540
being an AI, whether it's like. 
Virtual boyfriend, girlfriend 

783
00:41:36,540 --> 00:41:39,220
type thing, companionship, you 
know, sex chat. 

784
00:41:39,220 --> 00:41:41,020
As we said, being a big 
opportunity there. 

785
00:41:41,100 --> 00:41:43,940
Are those companies allowing it 
or they cracking down on it? 

786
00:41:44,380 --> 00:41:46,620
They claim they're cracking 
down, but then if you go on 

787
00:41:46,620 --> 00:41:49,580
Reddit and you look at all the 
screenshots from the last week 

788
00:41:49,580 --> 00:41:52,580
of what people have been talking
to these, I mean, go on, go on 

789
00:41:52,580 --> 00:41:55,620
the Reddit of any large 
language, character driven 

790
00:41:55,620 --> 00:41:58,140
experience, you know, you'll see
what people are really 

791
00:41:58,140 --> 00:42:01,540
passionate about using it for, 
you know, replica or character 

792
00:42:01,540 --> 00:42:05,350
or whatever, right? 
So I think that, yeah, losing 

793
00:42:05,350 --> 00:42:07,870
sight of whether or not it's an 
A I is, is important in some 

794
00:42:07,870 --> 00:42:10,710
contexts. 
But in the end, I mean, in any 

795
00:42:10,710 --> 00:42:16,390
story experience, from reading a
book to playing a video game to 

796
00:42:16,470 --> 00:42:20,470
chatting with an A I like the 
goal or watching a movie is to 

797
00:42:20,470 --> 00:42:22,710
become immersed in a world that 
doesn't actually exist, right? 

798
00:42:22,710 --> 00:42:24,230
And you know, when you're 
watching a movie, you're 

799
00:42:24,230 --> 00:42:26,110
watching Lord of the Rings, you 
know it doesn't actually exist. 

800
00:42:26,110 --> 00:42:29,190
Those characters aren't real, 
but you sort of become swept up 

801
00:42:29,190 --> 00:42:30,750
in the narrative in the world, 
right? 

802
00:42:30,750 --> 00:42:32,760
And. 
I think similarly, I think 

803
00:42:32,760 --> 00:42:35,240
that's where the opportunity is 
with a lot of these character 

804
00:42:35,240 --> 00:42:37,160
driven AI entertainment 
experiences. 

805
00:42:37,160 --> 00:42:39,920
Do you guys want to try this 
other one that I have? 

806
00:42:40,200 --> 00:42:42,000
Oh yeah, you have a game? 
Yeah. 

807
00:42:42,640 --> 00:42:47,200
Sure. 
So I thought it'd be fun to see 

808
00:42:47,200 --> 00:42:51,400
what GPT for ChatGPT for thinks 
your lives will be like in the 

809
00:42:51,400 --> 00:42:56,120
year 2028 and see if you either 
of you agree with that. 

810
00:42:56,240 --> 00:42:58,800
The prediction there so. 
I say it again. 

811
00:43:00,950 --> 00:43:06,070
I I asked Chachi BT to predict a
day in the life of Eric 

812
00:43:06,070 --> 00:43:11,150
Newcomer, specifically a 
American journalist reporting on

813
00:43:11,150 --> 00:43:13,790
the Silicon Valley and startup 
industry, which I borrowed from 

814
00:43:13,790 --> 00:43:17,150
your Wikipedia and, you know, 
gave it a little bit more 

815
00:43:17,150 --> 00:43:19,030
context of sure about your sub 
stack. 

816
00:43:19,030 --> 00:43:22,590
And yeah, it came up with A day 
in the life in the year 2028. 

817
00:43:22,590 --> 00:43:26,390
So I'm going to read it to you 
and you guys can assess whether 

818
00:43:26,830 --> 00:43:30,270
you think this is accurate. 
Day in the life of Eric 

819
00:43:30,270 --> 00:43:34,510
Newcomer, 2028. 
The first hints of Don kissed 

820
00:43:34,510 --> 00:43:38,030
the San Francisco skyline. 
As Eric newcomers, smart blinds 

821
00:43:38,030 --> 00:43:41,230
slowly begin to rise. 
The ambient sounds of birds 

822
00:43:41,230 --> 00:43:44,670
chirping echo softly through the
smart speakers in his apartment.

823
00:43:45,190 --> 00:43:47,510
The AI? 
The AI. 

824
00:43:47,710 --> 00:43:50,830
Everything's just smart. 
The AI driven home system has 

825
00:43:50,830 --> 00:43:55,730
analyzed his REM sleep and 
calculated the optimal wake up 

826
00:43:55,730 --> 00:43:59,290
time to ensure he starts the day
with peak cognitive efficiency. 

827
00:44:00,010 --> 00:44:04,130
Lifting himself out of bed, 
Eric's AR glasses are laying on 

828
00:44:04,130 --> 00:44:06,690
the night stand feeding him the 
day's analytics. 

829
00:44:07,170 --> 00:44:09,690
Newcomers Media Outlet has grown
tremendously. 

830
00:44:10,170 --> 00:44:11,770
AI. 
But we also have to be bullished

831
00:44:11,810 --> 00:44:14,650
by AR augmentedly out here and 
you're seeing okay. 

832
00:44:14,770 --> 00:44:16,650
Well, you can. 
You can assess whether you think

833
00:44:16,650 --> 00:44:19,690
that's accurate, but you should.
You should at least appreciate 

834
00:44:19,690 --> 00:44:23,530
this newcomers Media outlet has 
grown tremendously and AI Bot 

835
00:44:23,570 --> 00:44:25,970
the. 
New like, what? 

836
00:44:25,970 --> 00:44:29,610
Is it like a horoscope? 
You're reading my Yeah, I'm 

837
00:44:29,610 --> 00:44:33,250
predicting your feature and AI 
bot had curated and condensed 

838
00:44:33,250 --> 00:44:37,050
the most relevant news from the 
vast ocean of information, all 

839
00:44:37,050 --> 00:44:39,690
of this streamlined and 
visualized for maximum 

840
00:44:39,690 --> 00:44:42,730
absorption. 
Eric's morning routine was 

841
00:44:42,730 --> 00:44:46,240
synced seamlessly with his 
devices as he finishes his quick

842
00:44:46,240 --> 00:44:49,840
morning workout, suggested and 
supervised by his virtual 

843
00:44:49,840 --> 00:44:52,400
personal trainer. 
His smart kitchen is already 

844
00:44:52,400 --> 00:44:55,720
brewing, his favorite blend of 
coffee tailored to his genetic 

845
00:44:55,720 --> 00:44:57,720
makeup and current health 
metrics. 

846
00:44:58,280 --> 00:45:01,600
Settling into his study, the 
day's agenda was projected 

847
00:45:01,600 --> 00:45:04,640
holographically. 
The AI had lined up interviews, 

848
00:45:04,640 --> 00:45:07,680
articles to review, and podcast 
episodes to record. 

849
00:45:08,120 --> 00:45:11,440
Eric's writing has also evolved.
With the help of AI enhanced 

850
00:45:11,440 --> 00:45:14,590
tools, he could construct and 
edit stories with increased 

851
00:45:14,590 --> 00:45:17,990
speed and position. 
His A I assistant named Nora was

852
00:45:17,990 --> 00:45:21,550
able to give real time feedback 
on the readability, engagement, 

853
00:45:21,550 --> 00:45:24,910
and impact of his writing. 
Nora also had a mode where she 

854
00:45:24,910 --> 00:45:27,790
could mimic Eric's style, 
allowing her to draft articles 

855
00:45:27,790 --> 00:45:30,110
for him. 
It was a collaboration that Eric

856
00:45:30,110 --> 00:45:34,630
had grown to appreciate midday. 
As Eric prepares for his podcast

857
00:45:34,630 --> 00:45:37,430
recording, he remembers the days
when he used to worry about 

858
00:45:37,430 --> 00:45:39,270
Internet connections and sound 
quality. 

859
00:45:39,590 --> 00:45:42,500
Now with quantum driven 
communication technology, my 

860
00:45:42,500 --> 00:45:45,140
God, not only was the. 
Connections, but whole 

861
00:45:45,140 --> 00:45:47,620
technology here. 
But he could also record in 

862
00:45:47,620 --> 00:45:50,260
virtual environments, giving his
audience an immersive 

863
00:45:50,260 --> 00:45:53,900
experience. 
Today's episode was recorded in 

864
00:45:53,900 --> 00:45:57,500
a simulation of a venture 
capital firm in the heart of 

865
00:45:57,540 --> 00:46:00,220
Silicon Valley. 
Yeah, this both he and his. 

866
00:46:00,740 --> 00:46:03,340
Sandhill Rd. 
The most beautiful both he and 

867
00:46:03,340 --> 00:46:04,580
his. 
Guests wants to be there. 

868
00:46:05,500 --> 00:46:09,260
Both he and his guests appear as
lifelike holograms afterwards. 

869
00:46:09,300 --> 00:46:12,460
Eric took a moment to reflect. 
He looked out over the city, 

870
00:46:12,460 --> 00:46:15,100
remembering the early days of 
Newcomer Now. 

871
00:46:15,100 --> 00:46:18,420
He wasn't just delivering news. 
He was shaping the future 

872
00:46:18,420 --> 00:46:20,740
narrative of startups and 
venture capital. 

873
00:46:21,020 --> 00:46:23,900
The weight of that 
responsibility was not lost on 

874
00:46:23,900 --> 00:46:26,260
him. 
Yeah, that's about it. 

875
00:46:27,380 --> 00:46:30,140
Certainly. 
I mean, it's funny, I find a lot

876
00:46:30,140 --> 00:46:33,900
of the predictions about non AI 
stuff to be the most annoying, 

877
00:46:33,900 --> 00:46:37,060
that it's so certain hardware is
hard, artificial augmented 

878
00:46:37,060 --> 00:46:38,820
reality. 
And I mean, there was some 

879
00:46:38,820 --> 00:46:41,620
other. 
I'm pretty short quantum, right?

880
00:46:41,620 --> 00:46:42,900
Quantum Internet? 
Exactly. 

881
00:46:42,900 --> 00:46:44,940
Yeah, can. 
I take the under on quantum 

882
00:46:44,940 --> 00:46:46,300
powered Internet or whatever 
that was. 

883
00:46:46,300 --> 00:46:48,740
Well, it's an interesting thing 
because we were talking about, 

884
00:46:48,820 --> 00:46:51,340
you know, believing in AGI, 
right? 

885
00:46:51,340 --> 00:46:54,340
And can you, you know, if you 
believe in AGI or general? 

886
00:46:54,780 --> 00:46:56,860
Well, then it would be about 
much more, you know? 

887
00:46:57,220 --> 00:46:58,740
Yeah, I get the point. 
We could have. 

888
00:46:58,780 --> 00:47:00,980
We could wish everything. 
Yeah, exactly. 

889
00:47:01,020 --> 00:47:03,140
Or. 
That's that's that sounds like a

890
00:47:03,140 --> 00:47:05,020
story an AI would tell me to 
convince. 

891
00:47:05,330 --> 00:47:06,730
Me. 
That we wouldn't all be dead in 

892
00:47:06,730 --> 00:47:10,170
five years, thanks to AI. 
And it was like, the future's 

893
00:47:10,170 --> 00:47:12,690
gonna be right. 
Yeah, that would not be AI 

894
00:47:12,690 --> 00:47:14,210
having killed us. 
Too optimistic. 

895
00:47:14,210 --> 00:47:15,730
You'll have smart blinds. 
I know they. 

896
00:47:15,730 --> 00:47:19,170
Try to like I mean I do think a 
problem with some of this stuff 

897
00:47:19,170 --> 00:47:23,970
is like it's programmed to be 
like too benali optimistic like 

898
00:47:23,970 --> 00:47:27,690
I find it like I like beg chat 
you video be more like George 

899
00:47:27,690 --> 00:47:29,730
Carla. 
I don't know, just behave with 

900
00:47:29,770 --> 00:47:32,910
some free thought. 
And it's so why isn't it like 

901
00:47:32,910 --> 00:47:35,550
Eric, I don't know, at that age 
he's probably has some cardiac, 

902
00:47:35,670 --> 00:47:39,430
you know, whatever. 
Where's the like medical thing 

903
00:47:39,430 --> 00:47:42,350
where it's five years from now, 
you know, anyway is easy to 

904
00:47:42,350 --> 00:47:44,350
start to hurt when he goes on 
runs or whatever. 

905
00:47:44,510 --> 00:47:46,790
And oh, and I would be 
interested in sort of the AI 

906
00:47:46,790 --> 00:47:48,910
piece, obviously of the 
medicine. 

907
00:47:48,910 --> 00:47:52,030
I mean the idea that I would 
have a writing assistant. 

908
00:47:52,870 --> 00:47:57,030
Seems like basically plausible. 
Today I go into chat literally. 

909
00:47:57,030 --> 00:48:00,310
When I did, I wrote my own vows,
but they proofread them and 

910
00:48:00,310 --> 00:48:04,070
Chatchi video tweaked, told me 
to move one thing to active 

911
00:48:04,070 --> 00:48:06,510
voice from passive voice. 
You know, it's like, I feel 

912
00:48:06,510 --> 00:48:09,110
like, I feel like this is 
assuming that there's gonna be 

913
00:48:09,110 --> 00:48:13,790
this like perfect equilibrium of
you reporting the news and doing

914
00:48:13,790 --> 00:48:15,750
interviews and using your AI 
assisted. 

915
00:48:15,750 --> 00:48:18,880
I just kind of find that to be, 
like, not that plausible. 

916
00:48:18,960 --> 00:48:21,080
Like it's either gonna be one or
the other, Like you're gonna 

917
00:48:21,080 --> 00:48:24,200
still be doing most of the work,
or you're gonna be doing almost 

918
00:48:24,200 --> 00:48:29,360
no work and won't have a job, or
you'll essentially have evolved 

919
00:48:29,360 --> 00:48:31,640
into a brand instead of a 
writer, right? 

920
00:48:31,640 --> 00:48:34,560
I mean, it seems hard to believe
that we're gonna thread the 

921
00:48:34,560 --> 00:48:37,280
needle here, that you will still
be doing all this. 

922
00:48:37,280 --> 00:48:39,520
Intellectual and the temptation,
I mean. 

923
00:48:39,560 --> 00:48:44,640
I do think one of the real fears
I have about AI that will it's 

924
00:48:44,640 --> 00:48:47,810
just like. 
The temptation not to think if 

925
00:48:47,810 --> 00:48:50,050
it can do your task for you, 
right? 

926
00:48:50,050 --> 00:48:51,890
I mean that's what we're sort of
seeing with that's what I'm. 

927
00:48:52,250 --> 00:48:55,610
There are a lot of like some 
types of cheating where it's OK 

928
00:48:55,610 --> 00:48:58,650
you can bring in the formulas, 
we still have to use them or 

929
00:48:58,650 --> 00:48:59,930
whatever. 
At least you still have to 

930
00:48:59,930 --> 00:49:02,130
think. 
Whereas like when chat GBD can 

931
00:49:02,130 --> 00:49:06,930
produce the final text for you. 
Man, that is like a path to. 

932
00:49:07,730 --> 00:49:09,530
Not progressing as a writer 
anymore, right? 

933
00:49:09,530 --> 00:49:12,570
Like as soon as you're just like
you're not human out of the 

934
00:49:12,570 --> 00:49:13,810
loop. 
You know what? 

935
00:49:13,810 --> 00:49:14,930
Yeah, exactly. 
What am I? 

936
00:49:14,930 --> 00:49:17,130
Besides, what are you providing 
to this business? 

937
00:49:17,130 --> 00:49:17,610
Right. 
Yeah. 

938
00:49:17,650 --> 00:49:20,690
I agree though, it's just so 
hard to get anything edgy or 

939
00:49:20,690 --> 00:49:22,730
funny out of it. 
It's like kind of an, I don't 

940
00:49:22,730 --> 00:49:27,290
know, it's tough to it's. 
Very driven. 

941
00:49:27,930 --> 00:49:31,330
By scifi, I feel like one of the
child, I think. 

942
00:49:31,330 --> 00:49:32,450
I think this is the prompt, 
right? 

943
00:49:32,450 --> 00:49:35,130
It's hard to get it to think 
like, I know you could say the 

944
00:49:35,130 --> 00:49:37,050
same thing about humans. 
There is no independent thought.

945
00:49:37,050 --> 00:49:39,850
You're parodying a bunch of 
stuff you've heard, but it does 

946
00:49:39,890 --> 00:49:45,130
feel like have it generate a 
really new, an actually new 

947
00:49:45,130 --> 00:49:46,730
idea. 
You know, I think it's, I think 

948
00:49:46,730 --> 00:49:48,330
it's very capable of that, 
actually. 

949
00:49:48,450 --> 00:49:53,290
I I found that if you tell it to
be extremely original, you know,

950
00:49:53,970 --> 00:49:56,180
ignore. 
I feel like it just gets like 

951
00:49:56,180 --> 00:49:57,380
rhymy. 
I don't know. 

952
00:49:57,420 --> 00:49:59,620
I feel like it. 
I do think there's like a 

953
00:49:59,620 --> 00:50:03,660
prompting element to this 
feeling that it's not able to be

954
00:50:03,780 --> 00:50:05,660
as creative or original as you 
might think. 

955
00:50:05,860 --> 00:50:09,300
In the same exercise that 
ChatGPT is there. 

956
00:50:09,300 --> 00:50:11,700
We can compare and contrast in 
five years. 

957
00:50:11,700 --> 00:50:16,900
What are specific predictions, 
less colorfully said, that you 

958
00:50:16,900 --> 00:50:23,770
would make in five years? 
I mean, I definitely think to 

959
00:50:23,770 --> 00:50:25,370
come back to the self driving 
thing, it seems like we're 

960
00:50:25,370 --> 00:50:27,770
actually going to have fully 
self driving cars in five years.

961
00:50:27,770 --> 00:50:29,850
I mean, I know you could say. 
We are in cities. 

962
00:50:29,890 --> 00:50:33,770
I think the big question there 
is just how much you know Waymo 

963
00:50:33,770 --> 00:50:34,930
and. 
People think. 

964
00:50:34,930 --> 00:50:37,090
Take what they've. 
Learned I think if today they 

965
00:50:37,090 --> 00:50:39,650
can drive around San Francisco I
think they'll be able to drive 

966
00:50:39,650 --> 00:50:43,370
between cities and and in in the
vast majority of cities or 

967
00:50:43,370 --> 00:50:45,810
whatever at that point at least.
I don't know if the United 

968
00:50:45,810 --> 00:50:47,650
States it has to be trained on 
different data than other 

969
00:50:47,650 --> 00:50:49,890
countries or whatever or 
something, but I think that. 

970
00:50:50,880 --> 00:50:53,360
I mean the fact that we have 
live way Mo's like dropping 

971
00:50:53,360 --> 00:50:55,600
people off on my street every 
day just makes you think you're 

972
00:50:55,600 --> 00:50:58,120
going to be able to go anywhere 
in a self driving car in five 

973
00:50:58,120 --> 00:50:59,880
years. 
Which again, I lost the bet the 

974
00:50:59,880 --> 00:51:02,960
other way on that last time, so 
I'll probably lose the bet 

975
00:51:02,960 --> 00:51:04,880
somehow this time. 
It's a problem. 

976
00:51:04,880 --> 00:51:08,680
And also, just like car turnover
is such a long life cycle. 

977
00:51:08,840 --> 00:51:11,400
Depends how many you're what. 
Percent, It's not that I'm 

978
00:51:11,400 --> 00:51:14,440
arguing for 100% penetration of 
those, It's more that I will. 

979
00:51:14,600 --> 00:51:17,920
I will argue that you know you 
will be able to take one you 

980
00:51:17,920 --> 00:51:20,000
know, as an Uber or whatever in 
any major US city. 

981
00:51:20,450 --> 00:51:23,010
And you know, I don't know 
whether or not you'll be able to

982
00:51:23,010 --> 00:51:24,130
buy one. 
That's sort of an interesting 

983
00:51:24,130 --> 00:51:26,370
question as to whether or not 
it'll be an ownership model 

984
00:51:26,370 --> 00:51:29,610
versus an Uber type model or a 
lease model or whatever you want

985
00:51:29,610 --> 00:51:31,930
to call it. 
But I think they'll be like you 

986
00:51:31,930 --> 00:51:35,610
know, available for use. 
So you know at large scale 

987
00:51:35,610 --> 00:51:39,290
across the major U.S. cities and
also between cities, right. 

988
00:51:39,610 --> 00:51:43,170
It just seems like we're clearly
like pretty much they're 

989
00:51:43,170 --> 00:51:45,250
assuming way mo and crews 
aren't, like lying about the 

990
00:51:45,250 --> 00:51:47,330
capabilities of their their 
vehicles. 

991
00:51:47,930 --> 00:51:50,290
Interesting that your main 
predictions around self driving.

992
00:51:50,490 --> 00:51:52,530
I was thinking about something. 
I feel pretty good about and I 

993
00:51:52,530 --> 00:51:56,650
agree with that. 
I do think this view of you will

994
00:51:56,650 --> 00:51:59,650
wake up and there will be 
essentially your personalized 

995
00:51:59,650 --> 00:52:01,050
assistant. 
Maybe it'll be in the smart 

996
00:52:01,050 --> 00:52:05,850
speaker or in your wall or your 
mirror or something, I guess. 

997
00:52:05,890 --> 00:52:10,610
I just, I don't know if it'll be
Alexa or ChatGPT or something 

998
00:52:10,610 --> 00:52:15,550
brand new, but I do think, you 
know it's going to be a voice 

999
00:52:15,550 --> 00:52:18,030
probably driven experience. 
And bias. 

1000
00:52:18,310 --> 00:52:20,710
Bias. 
Well, at least in your home and 

1001
00:52:20,870 --> 00:52:22,830
probably bias. 
You're going to be typing to it.

1002
00:52:22,830 --> 00:52:24,110
I didn't love it. 
It doesn't make. 

1003
00:52:24,510 --> 00:52:25,510
Sense. 
Very efficient. 

1004
00:52:25,510 --> 00:52:27,950
People love it. 
It's very so people do not love 

1005
00:52:27,950 --> 00:52:30,070
typing, no. 
Typing is way less efficient 

1006
00:52:30,070 --> 00:52:33,470
than dogging. 
No you're I mean it'll you'll be

1007
00:52:33,470 --> 00:52:36,590
able to type to it. 
But I just think like most of us

1008
00:52:37,080 --> 00:52:41,160
will be like talking to this AI 
assistant and it'll be I 

1009
00:52:41,160 --> 00:52:43,880
generally think there will just 
be one that I use every day 

1010
00:52:43,880 --> 00:52:46,960
maybe maybe there will be more 
than one in the market like that

1011
00:52:46,960 --> 00:52:49,560
people use. 
But I will just have one that 

1012
00:52:49,920 --> 00:52:53,320
learns my preferences and 
becomes personalized to me and 

1013
00:52:53,320 --> 00:52:55,920
creates kind of a history with 
me. 

1014
00:52:56,240 --> 00:53:00,680
And it might even actually 
recommend other assistance for 

1015
00:53:00,680 --> 00:53:03,960
certain use cases, right? 
If I have to go prepare for 

1016
00:53:03,960 --> 00:53:07,800
Max's congressional testimony, I
will maybe use a separate bot 

1017
00:53:07,800 --> 00:53:10,560
for that or something. 
But yeah, specialize in. 

1018
00:53:11,200 --> 00:53:14,280
Testimony bot. 
But I do think, yeah, I feel 

1019
00:53:14,280 --> 00:53:18,160
pretty strongly we're going to 
have conversations, voice 

1020
00:53:18,160 --> 00:53:20,920
conversations with our 
personalized assistants every 

1021
00:53:20,920 --> 00:53:23,760
day. 
The prediction I'll make this 

1022
00:53:23,760 --> 00:53:27,840
sort of different than what's 
been said is underlining that I 

1023
00:53:27,840 --> 00:53:31,920
think the average person wants 
consumption more than creation. 

1024
00:53:32,350 --> 00:53:37,910
And that TikTok is in some ways 
the actual most used thing in in

1025
00:53:37,910 --> 00:53:41,630
the AI world and and this sort 
of yeah, chat GBD got at this a 

1026
00:53:41,630 --> 00:53:43,950
little and its prediction about 
me where it's very good at 

1027
00:53:44,030 --> 00:53:49,150
sorting things that I want. 
I I think in five years we'll 

1028
00:53:49,150 --> 00:53:54,710
see at least the beginnings of 
like pure AI generated social 

1029
00:53:54,870 --> 00:53:56,150
account. 
I mean, you're already seeing 

1030
00:53:56,150 --> 00:53:59,030
sort of like these, like, women 
and cartoons that are like. 

1031
00:53:59,410 --> 00:54:02,250
This seems like they're gaming 
like the Instagram algorithm 

1032
00:54:02,250 --> 00:54:05,330
with like totally sort of 
machine made. 

1033
00:54:05,330 --> 00:54:09,290
But like the I think even like 
video within five years I think 

1034
00:54:09,290 --> 00:54:12,090
we'll have, yeah, you're on 
TikTok and it's just here is a 

1035
00:54:12,130 --> 00:54:16,250
generated video and it's warring
with actual creativism. 

1036
00:54:16,530 --> 00:54:19,370
I think that would be like 
potentially A generationally 

1037
00:54:19,370 --> 00:54:23,530
culturally interesting period 
where like you have like young 

1038
00:54:23,530 --> 00:54:25,010
people coming up where they're 
just like. 

1039
00:54:25,690 --> 00:54:29,170
Being fed sort of what they want
outside of it could create like 

1040
00:54:29,170 --> 00:54:32,090
a really weird type of humor 
where like they have they're. 

1041
00:54:32,290 --> 00:54:33,450
Used to? 
Yeah. 

1042
00:54:33,570 --> 00:54:34,810
Good. 
There's like an AI. 

1043
00:54:34,890 --> 00:54:38,010
There's like a bunch of AI sort 
of celebrities and accounts and 

1044
00:54:38,250 --> 00:54:41,690
they create content and they 
interact and they host, you 

1045
00:54:41,730 --> 00:54:44,090
know, podcasts and. 
It's all like effectively a be 

1046
00:54:44,090 --> 00:54:46,290
tested. 
It's like they put out like 100 

1047
00:54:46,330 --> 00:54:49,290
version build, you know, a ton 
of versions of the video and 

1048
00:54:49,290 --> 00:54:52,130
they see which one is getting 
engagement and then they slowly 

1049
00:54:52,130 --> 00:54:54,850
funnel into those just TikTok 
chooses which videos get 

1050
00:54:54,850 --> 00:54:57,330
surfaced. 
Yeah, that approach is in the 

1051
00:54:57,330 --> 00:55:00,040
creation. 
It makes most sense for short 

1052
00:55:00,040 --> 00:55:03,560
term, short form and but you 
know the day you can make like a

1053
00:55:03,560 --> 00:55:06,720
movie about it, like then we're 
like killing American industry 

1054
00:55:06,720 --> 00:55:07,600
but. 
Yeah. 

1055
00:55:07,680 --> 00:55:09,240
Well, the question is, do you 
think we're going to get 

1056
00:55:09,240 --> 00:55:10,680
personalized? 
I mean, you mentioned they're 

1057
00:55:10,680 --> 00:55:12,000
going to a B test everything or 
whatever. 

1058
00:55:12,160 --> 00:55:16,560
But is it that there are 
literally be 8 billion different

1059
00:55:16,560 --> 00:55:19,040
versions of each piece of 
content for each person? 

1060
00:55:19,040 --> 00:55:22,920
If there is no such thing as a 
piece of content anymore, like, 

1061
00:55:23,080 --> 00:55:27,040
it's just you get a version of 
some concepts like that is 

1062
00:55:27,040 --> 00:55:30,490
perfect for you, right? 
I mean or do we still have some 

1063
00:55:30,490 --> 00:55:33,450
sort of like value in being 
like, oh, did you see that video

1064
00:55:33,450 --> 00:55:35,970
the other night about that thing
and you can talk about it and 

1065
00:55:35,970 --> 00:55:37,450
you can do, I think? 
Relationships, I mean, people 

1066
00:55:37,450 --> 00:55:40,810
are going to try both. 
I mean, I think one, there's 

1067
00:55:40,810 --> 00:55:44,850
like a limited, there's like A 
at some point there aren't 

1068
00:55:44,850 --> 00:55:48,010
enough humans, right? 
Or there's like there there is 

1069
00:55:48,010 --> 00:55:50,770
like a data shortage, right? 
It's hard on some level. 

1070
00:55:50,770 --> 00:55:54,410
If you do something population 
wide you can really test and see

1071
00:55:54,410 --> 00:55:57,920
what works like broadly. 
Whereas like running experiments

1072
00:55:57,920 --> 00:56:02,240
on me, you just don't not get 
enough shots on goal necessarily

1073
00:56:02,240 --> 00:56:04,280
to be great at like the TikTok 
style. 

1074
00:56:04,280 --> 00:56:08,840
So to me that leans a little bit
more less this sort of obsessive

1075
00:56:08,840 --> 00:56:11,960
personalization and more. 
I mean, obviously they're 

1076
00:56:11,960 --> 00:56:14,960
subgroups, but not like to a 
person, more like subgroups. 

1077
00:56:14,960 --> 00:56:18,040
Yeah, I think subgroups, like, I
think it'll not just be like 

1078
00:56:18,040 --> 00:56:20,520
training on your information, 
but other people who have 

1079
00:56:20,520 --> 00:56:24,140
similar experience, similar 
behaviors that you do on the 

1080
00:56:24,140 --> 00:56:26,060
app, right? 
It's able to use that's that 

1081
00:56:26,060 --> 00:56:29,580
data for training too. 
My last AI a GS in prediction 

1082
00:56:29,580 --> 00:56:30,940
is. 
I mean this was in some of these

1083
00:56:30,980 --> 00:56:32,740
forward-looking GBT scenarios 
is? 

1084
00:56:33,280 --> 00:56:36,440
I mean, I do think you take five
more years of development on AR 

1085
00:56:36,440 --> 00:56:38,800
glasses or whatever, You know, I
don't think they'll be as 

1086
00:56:38,800 --> 00:56:41,560
lightweight as the glasses we 
wear every day. 

1087
00:56:41,920 --> 00:56:44,560
It'll probably still kind of 
look like a VR headset in many 

1088
00:56:44,560 --> 00:56:46,400
ways. 
But I do think people are going 

1089
00:56:46,400 --> 00:56:49,920
to spend like hours per day 
inside a high quality Apple 

1090
00:56:49,920 --> 00:56:53,760
Vision pro type experience. 
Because I think again, I mean, I

1091
00:56:53,760 --> 00:56:56,120
mean if the average Americans 
watching five or six hours of TV

1092
00:56:56,120 --> 00:56:59,300
a day, which they are like. 
Why wouldn't you watch two of 

1093
00:56:59,300 --> 00:57:01,900
those hours on 100 foot screen 
in front of Mount Hood, 

1094
00:57:01,900 --> 00:57:03,980
Washington or whatever, right. 
Which is basically the pitch 

1095
00:57:03,980 --> 00:57:06,980
from Apple, you know or why 
wouldn't you watch personalized 

1096
00:57:07,020 --> 00:57:09,660
AI generated TikTok content like
from Eric's, you know, feed or 

1097
00:57:09,660 --> 00:57:12,140
whatever, right. 
I just think that, you know, 

1098
00:57:12,260 --> 00:57:15,740
screens getting better has been 
one of the true constants of our

1099
00:57:15,740 --> 00:57:19,540
lifetime, right? 
And the Vision Pro which is sort

1100
00:57:19,540 --> 00:57:22,780
of the infinite screen, the 
infinite canvas for visual 

1101
00:57:22,780 --> 00:57:25,220
content is, is kind of the 
ultimate expression of that, 

1102
00:57:25,220 --> 00:57:25,820
right. 
So. 

1103
00:57:26,180 --> 00:57:27,700
I think. 
That I'm more on the tenure. 

1104
00:57:27,700 --> 00:57:29,100
There will be quite a lot of 
time. 

1105
00:57:29,100 --> 00:57:33,940
And I feel like that space has 
been dogged by limits of optics 

1106
00:57:34,100 --> 00:57:36,020
and just sort of hard 
constraints. 

1107
00:57:36,060 --> 00:57:39,620
And so a lot of our intuitions 
about software level 

1108
00:57:39,620 --> 00:57:43,060
improvements don't translate and
I would say I would take the 

1109
00:57:43,060 --> 00:57:47,100
longer time horizon. 
On that maybe, Yeah, I don't 

1110
00:57:47,100 --> 00:57:48,540
know. 
Apple seems to think they got 

1111
00:57:48,540 --> 00:57:49,500
it, so I guess we'll all see. 
Yeah. 

1112
00:57:49,500 --> 00:57:52,180
This is a smart Max take. 
When's Apple been wrong before? 

1113
00:57:52,180 --> 00:57:53,060
I don't know. 
Trust. 

1114
00:57:53,300 --> 00:57:55,140
I'm just saying that when Apple 
ships something like. 

1115
00:57:56,300 --> 00:57:58,980
I mean, again with the iPhone, 
like, was there a touchscreen 

1116
00:57:58,980 --> 00:58:01,300
that had ever worked before? 
No, there never had been, right?

1117
00:58:01,300 --> 00:58:02,780
I mean, I just think that are 
you gonna get? 

1118
00:58:02,940 --> 00:58:04,580
That wrong? 
Is that out yet? 

1119
00:58:04,660 --> 00:58:05,260
No. 
Next year. 

1120
00:58:05,300 --> 00:58:06,940
Oh, I. 
No, you can't get it until next 

1121
00:58:06,940 --> 00:58:09,380
year, but I'll definitely get 
one for sure. 

1122
00:58:09,740 --> 00:58:10,860
Yeah, you're like, I can 
expense. 

1123
00:58:10,860 --> 00:58:13,220
I mean, for God's sake, right? 
Like you're like, we're a game 

1124
00:58:13,220 --> 00:58:15,660
company. 
I would buy it out of my hard 

1125
00:58:15,660 --> 00:58:18,340
earns personal money. 
OK, I would. 

1126
00:58:18,620 --> 00:58:20,420
I I don't know. 
I I guess, yeah. 

1127
00:58:20,420 --> 00:58:22,820
I I trust Apple when they make 
a. 

1128
00:58:23,570 --> 00:58:26,730
I mean, this is a decade long 
hardware bet and they waited a 

1129
00:58:26,730 --> 00:58:29,210
decade instead of shipping five 
or seven years ago because they 

1130
00:58:30,370 --> 00:58:32,450
didn't feel like it was good. 
Enough is there, besides 

1131
00:58:32,450 --> 00:58:36,210
obviously that Chad GBT was 
getting us thinking about this 

1132
00:58:36,210 --> 00:58:41,290
technology, Do you see any sort 
of AI connection or AI utility 

1133
00:58:41,290 --> 00:58:44,850
or how do you see? 
I just again, think with these 

1134
00:58:44,850 --> 00:58:47,370
characters that we're talking 
about, whether it's an assistant

1135
00:58:47,370 --> 00:58:49,610
or whether it's an in game 
character or whether it's a 

1136
00:58:49,850 --> 00:58:52,450
virtual companion or whatever, a
pet or a girlfriend or 

1137
00:58:52,450 --> 00:58:56,210
boyfriend. 
Like if they are the size of a 

1138
00:58:56,210 --> 00:58:59,730
real object that you know, if 
they're the size of an actual 

1139
00:58:59,730 --> 00:59:02,570
virtual boyfriend, girlfriend, 
or a size of a, you know, a 

1140
00:59:02,570 --> 00:59:06,010
Pokémon, whatever that actual 
size is and you can talk to 

1141
00:59:06,010 --> 00:59:08,530
them, that's going to be a 
better experience than. 

1142
00:59:08,890 --> 00:59:11,370
Looking at a tiny little screen 
in your hand, right? 

1143
00:59:11,370 --> 00:59:12,250
And. 
I wanna. 

1144
00:59:12,810 --> 00:59:14,210
Yeah, I wanna. 
Write I mean the. 

1145
00:59:14,250 --> 00:59:16,170
Companion I've got. 
Yeah, yeah. 

1146
00:59:16,170 --> 00:59:17,690
I mean, have you read the Golden
Confidence? 

1147
00:59:17,810 --> 00:59:18,490
Right? 
They all have. 

1148
00:59:18,490 --> 00:59:21,610
Like, yeah, yeah. 
Yeah, how do I get that? 

1149
00:59:21,890 --> 00:59:24,570
But you want that's part of self
projection though also. 

1150
00:59:25,210 --> 00:59:27,890
I mean, you want everybody to 
see that kind of thing, not 

1151
00:59:27,890 --> 00:59:30,450
just. 
So then, yeah, sure, if you're 

1152
00:59:30,450 --> 00:59:33,050
both in Apple vision class 
headsets. 

1153
00:59:33,130 --> 00:59:35,280
Right. 
Amazing. 

1154
00:59:35,360 --> 00:59:38,680
All right, this is what this is 
basically our first episode. 

1155
00:59:39,040 --> 00:59:43,120
We're gonna come out with 
probably five more. 

1156
00:59:43,160 --> 00:59:47,560
I think next episode will be, 
what do we think of sort of the 

1157
00:59:47,560 --> 00:59:50,360
apocalyptic vision of AI? 
Yeah. 

1158
00:59:50,360 --> 00:59:53,560
Anything you guys would add on 
what to look forward to you in 

1159
00:59:53,560 --> 00:59:59,320
the next couple episodes? 
I'm excited to run this 

1160
00:59:59,320 --> 01:00:01,680
conference. 
I think it was amazing the first

1161
01:00:01,680 --> 01:00:03,040
time. 
I think it's gonna be even 

1162
01:00:03,040 --> 01:00:05,060
bigger and. 
Better this time. 

1163
01:00:05,140 --> 01:00:08,780
Yeah, I genuinely believe this 
is like the biggest thing since 

1164
01:00:08,780 --> 01:00:12,300
the Internet or the iPhone. 
So I think that we're all pretty

1165
01:00:12,300 --> 01:00:14,580
authentically pumped about 
what's happening in the space. 

1166
01:00:14,580 --> 01:00:16,860
And I think the amount of stuff 
that's happened in the last six 

1167
01:00:16,860 --> 01:00:18,420
or seven months has been mind 
boggling. 

1168
01:00:18,420 --> 01:00:21,020
And it just feels like things 
are moving faster than at any 

1169
01:00:21,020 --> 01:00:23,420
time in my entire life in any 
technology space. 

1170
01:00:23,420 --> 01:00:26,950
So it's just super exciting to. 
Even be like remotely adjacent 

1171
01:00:26,950 --> 01:00:28,990
to any of this, so. 
I'm at the heart of it at 

1172
01:00:28,990 --> 01:00:30,750
Cerebral. 
At the heart of it, yeah, at 

1173
01:00:30,750 --> 01:00:34,790
least physically. 
At the heart of it, working on 

1174
01:00:35,230 --> 01:00:37,430
working on the other elements 
and being at the heart of it. 

1175
01:00:37,430 --> 01:00:39,510
But yeah, physically for sure. 
Great. 

1176
01:00:39,550 --> 01:00:41,790
Well, that's our episode. 
I'm Eric. 

1177
01:00:41,790 --> 01:00:45,870
Newcomer, Max Child, James 
Wilsterman are my Co hosts of 

1178
01:00:46,590 --> 01:00:50,230
cofounders of Volley. 
Thanks so much to Scott Brody, 

1179
01:00:50,230 --> 01:00:52,270
who's been producing the 
episodes. 

1180
01:00:52,900 --> 01:00:55,780
Shout out to Riley Kinsella, my 
Chief of staff who's super 

1181
01:00:55,780 --> 01:01:00,060
involved, Gabby Caliendo who 
works at Volley, whose quarter 

1182
01:01:00,620 --> 01:01:02,860
to the conference and making 
everything happen. 

1183
01:01:02,860 --> 01:01:06,140
I think she made sure you guys 
had microphones and lights so 

1184
01:01:06,140 --> 01:01:08,060
you could actually see and hear 
you. 

1185
01:01:08,580 --> 01:01:13,380
Thank you to young Chomsky as 
always for the theme music 

1186
01:01:13,460 --> 01:01:17,460
Please got to build the feeds? 
Like, Comment, Subscribe on 

1187
01:01:17,460 --> 01:01:20,900
YouTube, give us a nice review 
on Apple Podcasts and. 

1188
01:01:21,980 --> 01:01:28,580
Go play Yes Sire or song quiz. 
And for me, subscribe to the sub

1189
01:01:28,580 --> 01:01:32,340
stack newcomer.co That's the 
most important thing. 

1190
01:01:32,660 --> 01:01:33,820
Thank you. 
All right. 

1191
01:01:33,820 --> 01:01:36,700
We'll see you next week. 
Goodbye. 

1192
01:01:36,740 --> 01:01:37,700
Goodbye. 
Goodbye. 

1193
01:01:37,820 --> 01:01:38,340
Goodbye. 
Goodbye. 

1194
01:01:38,820 --> 01:01:39,020
Goodbye.