1
00:00:03,340 --> 00:00:06,940
Josh:
So remember all those big billion dollar paychecks multi-billion

2
00:00:06,940 --> 00:00:09,820
Josh:
dollar paychecks that we've been talking about zuck issuing to these top

3
00:00:09,820 --> 00:00:14,640
Josh:
tier ai engineers ejaz well the good news and the bad news is that we know where

4
00:00:14,640 --> 00:00:18,120
Josh:
the money's been going now because the meta team has released their first major

5
00:00:18,120 --> 00:00:22,680
Josh:
product and it's a little freaking scary it's a little dystopian we're going

6
00:00:22,680 --> 00:00:25,980
Josh:
to get into exactly what it is how it all works but before that there is a second

7
00:00:25,980 --> 00:00:29,320
Josh:
character in this story that is taking a seemingly opposite approach.

8
00:00:29,460 --> 00:00:32,800
Josh:
And that's OpenAI, who has dedicated all of their engineering talent towards

9
00:00:32,800 --> 00:00:34,140
Josh:
a new feature called Pulse.

10
00:00:34,500 --> 00:00:38,820
Josh:
Now, Pulse proactively assumes what you are going to want and delivers you value

11
00:00:38,820 --> 00:00:41,580
Josh:
and thinks for you literally overnight in order to make your life better.

12
00:00:41,660 --> 00:00:44,180
Josh:
So what we're seeing here, as these companies release more features,

13
00:00:44,320 --> 00:00:47,880
Josh:
is that they're actually taking two very separate approaches that are seemingly

14
00:00:47,880 --> 00:00:51,520
Josh:
one kind of optimistic, one a little more dystopian than each other.

15
00:00:51,600 --> 00:00:53,860
Josh:
EJS, can you first walk us through the meta news, what they announced,

16
00:00:54,100 --> 00:00:58,040
Josh:
how it works, what they've been up to with all of this multi multi multi multi

17
00:00:58,040 --> 00:00:59,460
Josh:
billion dollars worth of talent.

18
00:00:59,460 --> 00:01:02,740
Ejaaz:
Okay so the product is appropriately

19
00:01:02,740 --> 00:01:05,700
Ejaaz:
called vibes which implies

20
00:01:05,700 --> 00:01:10,300
Ejaaz:
that it's something uh good that's gonna kind of like comfort you maybe even

21
00:01:10,300 --> 00:01:15,280
Ejaaz:
sedate you josh and i use that term very specifically because they've basically

22
00:01:15,280 --> 00:01:23,500
Ejaaz:
created ai tiktok ai ig reels um in the words of alexander wang the head of

23
00:01:23,500 --> 00:01:25,440
Ejaaz:
Meta's super intelligence lab.

24
00:01:25,560 --> 00:01:32,840
Ejaaz:
He goes, excited to share Vibes, a new feed in the Meta AI app for short form AI generated videos.

25
00:01:33,100 --> 00:01:36,240
Ejaaz:
And rather than speak about what it is, I'm going to show you a few videos.

26
00:01:36,380 --> 00:01:39,800
Ejaaz:
You've got an astronaut riding a bike here. You've got a bear and a dinosaur

27
00:01:39,800 --> 00:01:44,340
Ejaaz:
fighting underwater. You've got a cat announcing the news. You've got people running on clouds.

28
00:01:44,520 --> 00:01:48,080
Ejaaz:
And before I start shitting on this, Josh, I'm going to say that it's actually

29
00:01:48,080 --> 00:01:53,020
Ejaaz:
really high quality um the graphics the kind of way that you can seamlessly

30
00:01:53,020 --> 00:01:59,600
Ejaaz:
edit and remix a bunch of these different contents and videos it's actually pretty cool but i don't

31
00:01:59,990 --> 00:02:02,970
Ejaaz:
really agree with the thesis which is basically

32
00:02:02,970 --> 00:02:06,690
Ejaaz:
how can i get a bunch of people addicted to watching short form content as if

33
00:02:06,690 --> 00:02:11,810
Ejaaz:
they weren't already addicted as well already you can now basically take any

34
00:02:11,810 --> 00:02:15,630
Ejaaz:
kind of idea that you might have spend a few seconds running out a prompt and

35
00:02:15,630 --> 00:02:19,430
Ejaaz:
create a new video that just endlessly loops addicting you to watch it over

36
00:02:19,430 --> 00:02:22,370
Ejaaz:
and over again can you walk us through

37
00:02:22,370 --> 00:02:26,490
Josh:
Like exactly how this works is is it the user that's kind of prompting a model

38
00:02:26,490 --> 00:02:30,210
Josh:
and then posting it to a separate timeline is that kind of what this is So maybe

39
00:02:30,210 --> 00:02:34,590
Josh:
like an Instagram-Facebook hybrid, but purely AI-generated? Yeah.

40
00:02:34,750 --> 00:02:40,450
Ejaaz:
So if you imagine you're on your Instagram app and you click add a post or create

41
00:02:40,450 --> 00:02:44,870
Ejaaz:
a post, you will now see an option to create an AI-generated video.

42
00:02:45,050 --> 00:02:48,510
Ejaaz:
And when you tap that, it'll come up with a little prompt box,

43
00:02:48,610 --> 00:02:52,470
Ejaaz:
similar to if you've ever interacted with ChatGPT or even typed anything in Google.

44
00:02:52,550 --> 00:02:56,990
Ejaaz:
And you can just describe what you want to see in a video.

45
00:02:56,990 --> 00:03:03,850
Ejaaz:
So it could be, in this case, a cow wearing sunglasses mooing at you enthusiastically,

46
00:03:03,850 --> 00:03:11,010
Ejaaz:
or it could be a bird flying underwater. It could be pretty much anything you want to make up.

47
00:03:11,210 --> 00:03:14,750
Josh:
Okay, this was shocking when I saw this.

48
00:03:14,970 --> 00:03:17,150
Josh:
I guess it would be an understatement, it was my perception.

49
00:03:17,310 --> 00:03:20,690
Josh:
But before we get into my perception, your perception, because I'm curious what

50
00:03:20,690 --> 00:03:24,530
Josh:
you think, Ejaz, what did the public think of this? How did they take this news?

51
00:03:24,530 --> 00:03:27,550
Ejaaz:
Okay so there are two camps uh

52
00:03:27,550 --> 00:03:30,590
Ejaaz:
one is louder than the other so camp

53
00:03:30,590 --> 00:03:33,710
Ejaaz:
number one unsurprising uh is summarized

54
00:03:33,710 --> 00:03:36,750
Ejaaz:
by lindy man over here always a good reminder

55
00:03:36,750 --> 00:03:39,730
Ejaaz:
that the people making these products are not using them

56
00:03:39,730 --> 00:03:43,090
Ejaaz:
zuck is spending his time doing mma training running

57
00:03:43,090 --> 00:03:46,610
Ejaaz:
a business reading but why buying hawaiian

58
00:03:46,610 --> 00:03:49,370
Ejaaz:
real estate and talking to people this is meant for

59
00:03:49,370 --> 00:03:52,410
Ejaaz:
you not him he sells you plastic and

60
00:03:52,410 --> 00:03:55,510
Ejaaz:
rubber so he can buy wood and stone the point that this

61
00:03:55,510 --> 00:03:58,230
Ejaaz:
man is making is basically he's selling you a

62
00:03:58,230 --> 00:04:01,030
Ejaaz:
product that is going to waste your hours waste your

63
00:04:01,030 --> 00:04:03,950
Ejaaz:
time suck your attention up into a vacuum

64
00:04:03,950 --> 00:04:07,150
Ejaaz:
that pays betters bills pays their advertisers out

65
00:04:07,150 --> 00:04:10,370
Ejaaz:
whilst you eventually are the product

66
00:04:10,370 --> 00:04:13,290
Ejaaz:
now the other camp is displayed

67
00:04:13,290 --> 00:04:16,730
Ejaaz:
by Arthur Rock over here where he goes not surprised

68
00:04:16,730 --> 00:04:19,890
Ejaaz:
by the overwhelming d-cell takes on this it

69
00:04:19,890 --> 00:04:22,930
Ejaaz:
is the same myopic snobbery of inventions past

70
00:04:22,930 --> 00:04:27,690
Ejaaz:
david hockney called the photograph looking at the world from the point of view

71
00:04:27,690 --> 00:04:33,130
Ejaaz:
of a paralyzed cyclops basically this person is saying you guys are being extremely

72
00:04:33,130 --> 00:04:37,850
Ejaaz:
short-sighted this is the new innovation that everyone's been hoping for and

73
00:04:37,850 --> 00:04:40,090
Ejaaz:
it'll end up being a very productive thing

74
00:04:41,030 --> 00:04:45,170
Ejaaz:
Before I want you to retaliate to this, Josh, I can see his point of view.

75
00:04:45,350 --> 00:04:49,350
Ejaaz:
I can see his point of view. If this is a tool that is used for good,

76
00:04:49,630 --> 00:04:54,810
Ejaaz:
imagine being able to educate on a very abstract concept that is so hard to

77
00:04:54,810 --> 00:04:58,910
Ejaaz:
communicate over words or just by speaking into a microphone, much that we do.

78
00:04:58,910 --> 00:05:02,490
Ejaaz:
If there was a way that we could visually kind of illustrate a complex topic

79
00:05:02,490 --> 00:05:05,350
Ejaaz:
and teach many others, I could see that being really productive.

80
00:05:05,350 --> 00:05:08,290
Ejaaz:
But I don't think that that's what it's going to be used for the majority.

81
00:05:08,690 --> 00:05:12,870
Ejaaz:
I think it's going to be used for mass consumption. Whatever gets you views, whatever gets you paid.

82
00:05:13,330 --> 00:05:17,850
Ejaaz:
Josh, I want to pass the ball to you to give the first opinionated take on this, please.

83
00:05:18,090 --> 00:05:23,530
Josh:
Okay. So like I said, like we've been saying for months, they have spent an

84
00:05:23,530 --> 00:05:26,470
Josh:
astronomical amount of money on hiring talent.

85
00:05:26,630 --> 00:05:29,870
Josh:
They have extracted the best engineers in the world.

86
00:05:30,170 --> 00:05:36,190
Ejaaz:
$23 billion, which includes a $15 billion investment to acquire Alexander Wang,

87
00:05:36,550 --> 00:05:37,710
Ejaaz:
the guy that heads up this product.

88
00:05:37,930 --> 00:05:40,790
Josh:
They spent $15 billion for the guy that just posted that tweet.

89
00:05:41,090 --> 00:05:45,490
Josh:
So that's the context there. They have spent an ungodly amount of money designing

90
00:05:45,490 --> 00:05:49,270
Josh:
and developing this team, which you would hope that they are working to design

91
00:05:49,270 --> 00:05:50,270
Josh:
this really impressive software.

92
00:05:50,270 --> 00:05:53,770
Josh:
And before I get into the actual commentary, I do want to provide a little context

93
00:05:53,770 --> 00:05:57,090
Josh:
to the people who kind of aren't aware, or this was something I was thinking

94
00:05:57,090 --> 00:06:00,990
Josh:
about as I was reading the news, is that when you think about Meta and previously Facebook,

95
00:06:01,630 --> 00:06:08,030
Josh:
the Facebook team spent a lot of years and a lot of talent on developing a way to better serve ads.

96
00:06:08,230 --> 00:06:11,090
Josh:
And that resulted in a very addicted algorithm. them.

97
00:06:11,470 --> 00:06:15,090
Josh:
So they spent God knows how many engineers, millions of hours,

98
00:06:15,390 --> 00:06:16,610
Josh:
top talent out of universities.

99
00:06:16,930 --> 00:06:20,650
Josh:
They sucked them in. They paid them with a lot of incentives and these wonderful

100
00:06:20,650 --> 00:06:22,930
Josh:
offices that they get to work in, free lunches.

101
00:06:23,050 --> 00:06:26,190
Josh:
And they said, hey, your job is to figure out how to create a better timeline

102
00:06:26,190 --> 00:06:28,590
Josh:
to get people more hooked to sell better ads.

103
00:06:28,770 --> 00:06:31,510
Josh:
And that is a little bit of a dystopian way of seeing things.

104
00:06:31,630 --> 00:06:34,510
Josh:
So when I heard that they were spending all this money to hire all this talent,

105
00:06:34,710 --> 00:06:37,310
Josh:
this new AI initiative for a new company named Meta, I was like,

106
00:06:37,370 --> 00:06:40,970
Josh:
okay, we got a chance here. and the very first product they announce is.

107
00:06:41,500 --> 00:06:46,500
Josh:
How to get people addicted to a timeline for the AI age, where it is AI version

108
00:06:46,500 --> 00:06:51,680
Josh:
of this timeline, except it's generated completely through an AI model instead of human generated.

109
00:06:51,900 --> 00:06:54,640
Josh:
And the scary thing for me is that AI generated content, EJ,

110
00:06:54,700 --> 00:06:55,820
Josh:
as we've talked about this in the past,

111
00:06:56,260 --> 00:06:59,860
Josh:
it can get very good very quickly, where if you have access to a front facing

112
00:06:59,860 --> 00:07:03,860
Josh:
camera on someone's smartphone, and it's watching your eyes dilate as you watch

113
00:07:03,860 --> 00:07:07,140
Josh:
more interesting videos or lower and it detects every retention rate,

114
00:07:07,280 --> 00:07:08,500
Josh:
it can slowly iterate over time.

115
00:07:08,660 --> 00:07:11,720
Josh:
And with TikTok, when you're scrolling through these timelines, there's a a

116
00:07:11,720 --> 00:07:14,540
Josh:
fixed set of content that you can get through so maybe the

117
00:07:14,540 --> 00:07:17,360
Josh:
algorithm is choosing between five top videos that are interesting to

118
00:07:17,360 --> 00:07:20,580
Josh:
you with an ai generated timeline there

119
00:07:20,580 --> 00:07:23,520
Josh:
is no limit to the amount of videos

120
00:07:23,520 --> 00:07:26,840
Josh:
it's very open-ended so that n number that they could choose from becomes infinite which

121
00:07:26,840 --> 00:07:29,880
Josh:
means you can iterate over time until it becomes maximally addicting

122
00:07:29,880 --> 00:07:32,640
Josh:
for the specific user and generate content on

123
00:07:32,640 --> 00:07:36,060
Josh:
the fly so that is my dystopian approach it is very discouraging to

124
00:07:36,060 --> 00:07:39,220
Josh:
see this like i understand where they're coming from but

125
00:07:39,220 --> 00:07:42,280
Josh:
like my god as the first product to offer without any

126
00:07:42,280 --> 00:07:45,060
Josh:
good explanation like i'd love to hear from

127
00:07:45,060 --> 00:07:47,780
Josh:
them why i think they really just drop this

128
00:07:47,780 --> 00:07:51,040
Josh:
and like hey here's this thing we've been working on here's our first cool project um

129
00:07:51,040 --> 00:07:55,220
Josh:
without explaining the the vision and the why behind it and i think that's equally

130
00:07:55,220 --> 00:07:58,520
Josh:
as important because sure this is great technology it's nothing novel we've

131
00:07:58,520 --> 00:08:02,560
Josh:
seen vo3 make great videos way better than these so it's not like this is novel

132
00:08:02,560 --> 00:08:05,920
Josh:
technology they've just applied a timeline to it for the first time um but i'm

133
00:08:05,920 --> 00:08:08,240
Josh:
curious if you agree or disagree just what are your thoughts on,

134
00:08:08,760 --> 00:08:10,540
Josh:
where do you stand with this okay.

135
00:08:10,540 --> 00:08:16,400
Ejaaz:
So i'm conflicted i agree with a lot of what you just said but i want to point

136
00:08:16,400 --> 00:08:21,660
Ejaaz:
a few things out i want to argue with you a bit josh feel free to jump in whenever okay um

137
00:08:22,050 --> 00:08:28,830
Ejaaz:
Number one, I think we might be sitting on our own ethical horses right now.

138
00:08:29,570 --> 00:08:31,870
Ejaaz:
Do we apply the same kind of critique

139
00:08:31,870 --> 00:08:37,390
Ejaaz:
to TikTok as much or social media in general? I'm not sure that we do.

140
00:08:37,630 --> 00:08:41,930
Ejaaz:
I think that because you and I are so invested in AI and we want to see it kind

141
00:08:41,930 --> 00:08:44,510
Ejaaz:
of change the world, we expect that anyone,

142
00:08:44,730 --> 00:08:47,970
Ejaaz:
particularly someone that's spending hundreds of billions of dollars on compute

143
00:08:47,970 --> 00:08:53,190
Ejaaz:
to scale the best AI models in the world to use it for good or whatever we define as good.

144
00:08:54,050 --> 00:08:58,810
Ejaaz:
I don't think Zuck's ever explicitly stated that he was going to do that, right?

145
00:08:59,090 --> 00:09:02,670
Ejaaz:
He never claimed that he wanted to build the best AI model to make you a genius.

146
00:09:02,910 --> 00:09:07,750
Ejaaz:
He said Facebook or Meta specifically's mission and goal has always been to

147
00:09:07,750 --> 00:09:10,330
Ejaaz:
bring people together, to create a better social environment,

148
00:09:10,470 --> 00:09:13,870
Ejaaz:
to help them connect, blah, blah, blah, blah, blah. and he said that Meta's

149
00:09:13,870 --> 00:09:17,450
Ejaaz:
chops have been cut in building consumer apps.

150
00:09:17,670 --> 00:09:21,090
Ejaaz:
That's their bread and butter. He said that on the Doha Cash interview that

151
00:09:21,090 --> 00:09:22,650
Ejaaz:
he was on a couple of weeks back, right?

152
00:09:22,990 --> 00:09:27,890
Ejaaz:
And he's kind of delivering on that promise, I guess. And maybe we just don't

153
00:09:27,890 --> 00:09:29,390
Ejaaz:
like how it's materialized.

154
00:09:29,590 --> 00:09:33,690
Ejaaz:
I agree with you that he should have sold the vision a bit more because Zuck

155
00:09:33,690 --> 00:09:38,010
Ejaaz:
has never explicitly come out and said, hey, we want to increase eyeballs on our app.

156
00:09:38,230 --> 00:09:41,210
Ejaaz:
We need to increase shareholder value. And therefore, that's what we're trying

157
00:09:41,210 --> 00:09:44,810
Ejaaz:
to do here. he's always sold a more like, we're going to connect people online

158
00:09:44,810 --> 00:09:46,670
Ejaaz:
and that's going to bring us closer together.

159
00:09:46,750 --> 00:09:49,410
Ejaaz:
I want to hear what his pitch is.

160
00:09:49,590 --> 00:09:53,390
Ejaaz:
I'm kind of curious to hear from him versus Alexander Wang.

161
00:09:53,690 --> 00:09:58,890
Ejaaz:
The other thing that I'll say is I do think they're going to spin this from

162
00:09:58,890 --> 00:10:00,730
Ejaaz:
an educational point of view as well.

163
00:10:00,850 --> 00:10:06,430
Ejaaz:
I think they're going to use examples of people using this to kind of explain

164
00:10:06,430 --> 00:10:10,770
Ejaaz:
really complex topics, to illustrate new inventions or whatever that might be,

165
00:10:10,850 --> 00:10:14,550
Ejaaz:
or even to ideate like kind of new film concepts or video concepts to entertain.

166
00:10:14,730 --> 00:10:19,010
Ejaaz:
And I don't think net net, that's going to really result in too much of a bad thing.

167
00:10:19,310 --> 00:10:25,990
Ejaaz:
But I think you and I have been primed on TikTok too much. The other thing that is in your camp, Josh,

168
00:10:26,250 --> 00:10:32,450
Ejaaz:
is I think Zuck has seen a growing trend of Gen Z people watching absolute slop,

169
00:10:32,710 --> 00:10:34,730
Ejaaz:
not AI slop, just slop on TikTok.

170
00:10:34,730 --> 00:10:38,930
Ejaaz:
If you watch what some of the kids are digesting, like what was that skibbity

171
00:10:38,930 --> 00:10:43,190
Ejaaz:
toilet trend that was going nuts that basically all the kids could talk about

172
00:10:43,190 --> 00:10:45,990
Ejaaz:
or baby shark and all that kind of stuff.

173
00:10:46,230 --> 00:10:49,250
Ejaaz:
I think he's seen that and he knows that there are a bunch of kids that are

174
00:10:49,250 --> 00:10:52,430
Ejaaz:
entering the workforce or whatever that like this kind of stuff and he's just

175
00:10:52,430 --> 00:10:56,310
Ejaaz:
engineering a product that they're going to get hooked on and I'm very bearish on that.

176
00:10:57,030 --> 00:10:59,790
Josh:
Very well could be. I guess we'll leave it at that and we'll see.

177
00:10:59,990 --> 00:11:02,750
Josh:
I'm very much looking forward to the messaging that they're going to have around

178
00:11:02,750 --> 00:11:05,290
Josh:
this going forward and how they plan to use the technology.

179
00:11:05,470 --> 00:11:09,190
Josh:
But maybe you're right. Maybe we're just being a little too optimistic on our

180
00:11:09,190 --> 00:11:10,630
Josh:
expectations for Meta and the team.

181
00:11:10,810 --> 00:11:14,570
Ejaaz:
But this isn't unlike Meta or Zuck's nature, right?

182
00:11:14,750 --> 00:11:16,250
Ejaaz:
Like literally last week, they

183
00:11:16,250 --> 00:11:20,650
Ejaaz:
released a new AI dating app with two features that will help you find

184
00:11:21,080 --> 00:11:24,300
Ejaaz:
the love of your life, or, you know, just a girlfriend or a boyfriend or whoever

185
00:11:24,300 --> 00:11:25,200
Ejaaz:
you want to hang out with.

186
00:11:25,760 --> 00:11:29,600
Ejaaz:
And I'm referencing, this post summarizes actually very well,

187
00:11:29,780 --> 00:11:34,840
Ejaaz:
that the new AI assistant lets you search for a Brooklyn tech bro who would

188
00:11:34,840 --> 00:11:37,120
Ejaaz:
go to EDM concerts with me.

189
00:11:37,980 --> 00:11:41,960
Ejaaz:
And this one hit pretty close to home because, you know, we live in Brooklyn

190
00:11:41,960 --> 00:11:44,180
Ejaaz:
and it very much fulfills the stereotype here.

191
00:11:44,340 --> 00:11:49,060
Ejaaz:
So the point I'm making is Zuck and Meta haven't really kind of blindsided us

192
00:11:49,060 --> 00:11:52,120
Ejaaz:
or surprised us. They've kind of been making moves like this all along.

193
00:11:52,320 --> 00:11:56,300
Ejaaz:
And maybe this is the next step of a path that they just want to go down.

194
00:11:56,840 --> 00:11:59,800
Josh:
Yeah, I guess. Well, time will tell. I'm looking forward to future messaging

195
00:11:59,800 --> 00:12:01,460
Josh:
from $15 billion Alex Wang.

196
00:12:01,640 --> 00:12:04,940
Josh:
Hopefully he has something a little more inspiring to share with us.

197
00:12:05,120 --> 00:12:09,360
Josh:
But on the inspiring front, on the optimistic front, on the hopeful front,

198
00:12:09,580 --> 00:12:13,420
Josh:
we have OpenAI Ejaz, who is now doing something proactive.

199
00:12:13,640 --> 00:12:16,780
Josh:
I think this is the first time where an AI is actually thinking on behalf of

200
00:12:16,780 --> 00:12:19,580
Josh:
you without being prompted. Can you explain to us what's going on with this new pulse feature.

201
00:12:19,580 --> 00:12:25,360
Ejaaz:
Okay what if i told you josh that the episode that we're filming tomorrow the

202
00:12:25,360 --> 00:12:34,060
Ejaaz:
ai roundup was actually formed by open ai's latest product that they released a few days ago that'd

203
00:12:34,060 --> 00:12:37,360
Josh:
Be pretty stoked because uh that would save us a lot of time in preparation yep.

204
00:12:37,360 --> 00:12:44,380
Ejaaz:
I am talking about open ai pulse and in sam's words pulse works for you overnight

205
00:12:44,380 --> 00:12:48,200
Ejaaz:
and keeps thinking about your interests, your connected data,

206
00:12:48,600 --> 00:12:50,200
Ejaaz:
your recent chats, and more.

207
00:12:50,440 --> 00:12:55,420
Ejaaz:
Every morning, you get a custom-generated set of stuff you might be interested in.

208
00:12:55,520 --> 00:12:59,680
Ejaaz:
It performs well if you tell ChatGPT more about what's important to you.

209
00:12:59,780 --> 00:13:04,340
Ejaaz:
In regular chat, you could mention, I'd like to go visit Bora Bora someday,

210
00:13:04,360 --> 00:13:08,060
Ejaaz:
or my kid is six months old and I'm interested in developmental milestones.

211
00:13:08,300 --> 00:13:13,060
Ejaaz:
And in the future, you might get useful updates. And to kind of summarize what

212
00:13:13,060 --> 00:13:17,260
Ejaaz:
this product does, Josh, it's kind of a digital you.

213
00:13:18,070 --> 00:13:21,650
Ejaaz:
Thinks and gets better as you sleep, literally.

214
00:13:22,230 --> 00:13:25,830
Ejaaz:
So whilst you and I are asleep, it's thinking about what would Josh want to

215
00:13:25,830 --> 00:13:27,250
Ejaaz:
hear about tomorrow morning?

216
00:13:27,410 --> 00:13:31,210
Ejaaz:
What was the task Josh was trying to figure out at 12 p.m.

217
00:13:31,350 --> 00:13:34,850
Ejaaz:
Today that he didn't end up figuring out? How can I help him achieve that goal

218
00:13:34,850 --> 00:13:38,170
Ejaaz:
faster and quicker than his deadline in a week's time?

219
00:13:38,430 --> 00:13:41,450
Ejaaz:
It's meant to be a digital version of you that's meant to help and aid you,

220
00:13:41,810 --> 00:13:45,330
Ejaaz:
very much like a personalized AI that you and I have been so,

221
00:13:45,470 --> 00:13:49,030
Ejaaz:
so excited about kind of using and interacting with.

222
00:13:49,250 --> 00:13:52,330
Ejaaz:
This was a feature that they released for pro users.

223
00:13:52,630 --> 00:13:57,450
Ejaaz:
And in the words of Fiji Simo, the CEO of product, she goes,

224
00:13:57,670 --> 00:14:00,010
Ejaaz:
AI should do more than just answer questions.

225
00:14:00,030 --> 00:14:03,790
Ejaaz:
It should anticipate your needs and help reach your goals.

226
00:14:04,190 --> 00:14:09,850
Ejaaz:
And I kind of like this idea of instead of pitching you an AI agent that can

227
00:14:09,850 --> 00:14:10,910
Ejaaz:
solve all your problems,

228
00:14:11,090 --> 00:14:16,150
Ejaaz:
they're taking the step before that, which is let me help understand what you're

229
00:14:16,150 --> 00:14:20,450
Ejaaz:
trying to achieve and help you in small ways, which is give you up the information

230
00:14:20,450 --> 00:14:23,310
Ejaaz:
that you didn't even know you needed to help reach your goals.

231
00:14:23,430 --> 00:14:28,010
Josh:
This is super exciting. This is, it feels like a very natural extension to OpenAI's memory feature.

232
00:14:28,170 --> 00:14:30,410
Josh:
I mean, if you remember, they were the first ones, I believe,

233
00:14:30,490 --> 00:14:33,470
Josh:
to roll out memory and it became kind of the moat that they had.

234
00:14:33,610 --> 00:14:36,510
Josh:
And I love that this is an attempt to build on top of the moat.

235
00:14:36,570 --> 00:14:43,130
Josh:
And I also love that that came from Fiji, who is the CEO of product of OpenAI, like the a semi-CEO.

236
00:14:43,190 --> 00:14:46,570
Josh:
But it's cool that they're actually, they're shipping these really cool features

237
00:14:46,570 --> 00:14:49,290
Josh:
that are now building on top of that mode that they've created.

238
00:14:49,510 --> 00:14:53,230
Josh:
And again, I mean, I use ChatGPT mostly because of that memory mode.

239
00:14:53,310 --> 00:14:55,810
Josh:
I think it's really powerful. It makes a big deal. And I would love,

240
00:14:56,570 --> 00:14:59,890
Josh:
for AI to be proactive. And by all means, I very much think this is the first

241
00:14:59,890 --> 00:15:03,270
Josh:
time an AI is doing this, where it is actually ingesting your memory,

242
00:15:03,670 --> 00:15:04,910
Josh:
sitting on it, thinking about it.

243
00:15:05,150 --> 00:15:08,930
Josh:
They are applying the absolute best models they have overnight when you are

244
00:15:08,930 --> 00:15:12,530
Josh:
not online, and then surfacing the most helpful things based on their absolute

245
00:15:12,530 --> 00:15:14,130
Josh:
best models the following morning.

246
00:15:14,350 --> 00:15:16,830
Josh:
And that to me seems really helpful.

247
00:15:17,150 --> 00:15:20,270
Josh:
Like, okay, maybe it's in a way, it's like, a lot of times someone will tell

248
00:15:20,270 --> 00:15:22,870
Josh:
you like, oh, just go sleep on it. You'll come to an answer in the morning.

249
00:15:23,350 --> 00:15:25,650
Josh:
AI is very much. It's sleeping on it. It's sleeping on it for you,

250
00:15:25,810 --> 00:15:28,430
Josh:
on your behalf. And it is surfacing the best answer. So to me,

251
00:15:28,510 --> 00:15:30,230
Josh:
this is a really cool feature. I'm a big fan.

252
00:15:30,450 --> 00:15:34,230
Ejaaz:
Do you want to know something cool, Josh? So I have a pro subscription and I've

253
00:15:34,230 --> 00:15:37,070
Ejaaz:
been using it in the four days since it's been released.

254
00:15:37,250 --> 00:15:38,470
Josh:
What has it been like? What do you get?

255
00:15:38,630 --> 00:15:43,590
Ejaaz:
It started off kind of like a newsletter, your own personalized newsletter on

256
00:15:43,590 --> 00:15:46,850
Ejaaz:
your life, Josh, on all your interests that you wanted to hear about.

257
00:15:47,350 --> 00:15:50,410
Ejaaz:
And the newsletter would tell me things that happened

258
00:15:50,410 --> 00:15:53,130
Ejaaz:
over the last couple of days okay this is the this

259
00:15:53,130 --> 00:15:56,330
Ejaaz:
is the progress i'm really excited to tell you about this okay so step one is

260
00:15:56,330 --> 00:15:59,070
Ejaaz:
newsletter debrief on the things that have happened the

261
00:15:59,070 --> 00:16:02,570
Ejaaz:
last few days it that's cool but nothing super

262
00:16:02,570 --> 00:16:05,710
Ejaaz:
novel day two it started

263
00:16:05,710 --> 00:16:09,370
Ejaaz:
telling me those things that happened over the last couple of days but also

264
00:16:09,370 --> 00:16:14,590
Ejaaz:
started giving me ideas of what could happen in the future it was like okay

265
00:16:14,590 --> 00:16:18,230
Ejaaz:
oh have you ever thought about like you know maybe this AI agent could end up

266
00:16:18,230 --> 00:16:23,190
Ejaaz:
actually feasibly doing this because of this advancement that NVIDIA did with this chip.

267
00:16:23,270 --> 00:16:26,950
Ejaaz:
And I'm like, huh, you know what? I never really thought about that.

268
00:16:27,110 --> 00:16:28,730
Ejaaz:
That's actually really interesting.

269
00:16:29,070 --> 00:16:35,250
Ejaaz:
And then yesterday, Josh, the prompt that it gave me was a list of things that

270
00:16:35,250 --> 00:16:37,110
Ejaaz:
was the latest news in AI.

271
00:16:37,410 --> 00:16:42,890
Ejaaz:
And it said, by the way, have you considered talking about this angle when you

272
00:16:42,890 --> 00:16:45,770
Ejaaz:
film the episode on Wednesday? Yeah.

273
00:16:46,480 --> 00:16:50,980
Ejaaz:
And I was like, how, wait, how do you know I'm filming an episode on Wednesday?

274
00:16:51,140 --> 00:16:55,920
Ejaaz:
And then I realized I'd had a separate conversation with ChatGPT earlier on

275
00:16:55,920 --> 00:16:59,380
Ejaaz:
that day, trying to brainstorm ideas of how I would frame an argument.

276
00:16:59,520 --> 00:17:04,680
Ejaaz:
And it had brought in a new news topic that had broke that morning and synthesized

277
00:17:04,680 --> 00:17:07,560
Ejaaz:
it with that goal that I was going to film on Wednesday.

278
00:17:07,640 --> 00:17:10,420
Ejaaz:
And it was said, by the way, you should include this in that and frame it this way.

279
00:17:10,520 --> 00:17:14,880
Ejaaz:
So it's already kind of getting me to think ahead of my life in ways that I

280
00:17:14,880 --> 00:17:18,220
Ejaaz:
actually hadn't considered. or maybe I would only think about on that day.

281
00:17:18,380 --> 00:17:21,220
Ejaaz:
So it's kind of like helping my future self in a weird way.

282
00:17:21,400 --> 00:17:26,360
Josh:
That is very cool. And it seems like you are not alone in your happy sentiment

283
00:17:26,360 --> 00:17:28,760
Josh:
around it because Kevin, I mean, we have some public commentary.

284
00:17:29,300 --> 00:17:31,820
Josh:
People also seem to be pretty stoked that this is happening, right?

285
00:17:32,280 --> 00:17:36,820
Ejaaz:
Yeah, so Kevin Rose has a tweet here where he goes, ChatGPT Pulse is one of

286
00:17:36,820 --> 00:17:39,360
Ejaaz:
the most impactful features OpenAI has released.

287
00:17:39,540 --> 00:17:43,240
Ejaaz:
It's an agent that continuously researches on your behalf, building on topics

288
00:17:43,240 --> 00:17:47,220
Ejaaz:
from your recent conversations. i'm having a hard time wrapping my head around

289
00:17:47,220 --> 00:17:52,800
Ejaaz:
this paradigm because it's truly unlike anything we've seen before and i guess

290
00:17:52,800 --> 00:17:53,960
Ejaaz:
that i kind of resonate with

291
00:17:53,960 --> 00:17:56,780
Josh:
That so i guess in the search to figure out how this works uh we

292
00:17:56,780 --> 00:17:59,840
Josh:
stumbled upon this post by nathan dashes and he

293
00:17:59,840 --> 00:18:02,960
Josh:
just said yeah just had an interesting experience with gpt plus last night

294
00:18:02,960 --> 00:18:05,700
Josh:
i asked gpt a question and it gave me an answer that was

295
00:18:05,700 --> 00:18:09,020
Josh:
just so so this morning the first card impulse was three

296
00:18:09,020 --> 00:18:12,280
Josh:
times better to the same question um and

297
00:18:12,280 --> 00:18:15,040
Josh:
then you could see there's a response from adam fry who is the lead of product

298
00:18:15,040 --> 00:18:17,940
Josh:
at open ai and he said well overnight we use our best

299
00:18:17,940 --> 00:18:20,940
Josh:
models and also try to think the broader intent

300
00:18:20,940 --> 00:18:23,740
Josh:
of the question you ask to see if chat gpt can

301
00:18:23,740 --> 00:18:27,880
Josh:
be even more helpful so what's really cool is what they're doing is i just i

302
00:18:27,880 --> 00:18:30,640
Josh:
also have to wonder if they're using this as a test bed for new models without

303
00:18:30,640 --> 00:18:34,140
Josh:
telling you they said they're absolute best models i wonder if that's a teaser

304
00:18:34,140 --> 00:18:37,120
Josh:
i don't know i'm just speculating so i wonder if this this creates this new

305
00:18:37,120 --> 00:18:42,120
Josh:
meta of engaging with the ai where you can actually prompt which topics to think about overnight.

306
00:18:42,480 --> 00:18:46,360
Josh:
And if you don't feel satisfied with the answer, maybe you could just subtly

307
00:18:46,360 --> 00:18:49,460
Josh:
hint it to request to think harder on it overnight.

308
00:18:49,720 --> 00:18:55,920
Ejaaz:
Yeah, I think you kind of nailed it, Josh. So there's this paragraph in Fiji's

309
00:18:55,920 --> 00:19:00,460
Ejaaz:
announcement post where she goes, our reasoning models, and she's referring

310
00:19:00,460 --> 00:19:01,740
Ejaaz:
to OpenAI's latest models,

311
00:19:02,260 --> 00:19:06,080
Ejaaz:
were built to spend more time thinking before they answer, allowing them to

312
00:19:06,080 --> 00:19:11,080
Ejaaz:
work through complex tasks in science, coding, math at near PhD level.

313
00:19:11,180 --> 00:19:12,940
Ejaaz:
And this is something you and I have spoken about, Josh.

314
00:19:13,060 --> 00:19:17,260
Ejaaz:
We keep talking about these models breaking benchmarks and being so, so smart.

315
00:19:17,460 --> 00:19:22,080
Ejaaz:
But most of our listeners, including you and I, don't really get to see the

316
00:19:22,080 --> 00:19:23,680
Ejaaz:
fruits of this labor, right?

317
00:19:23,760 --> 00:19:27,400
Ejaaz:
We're like, okay, we're not doing PhD level coding. So what? Who cares?

318
00:19:27,680 --> 00:19:33,000
Ejaaz:
And Fiji goes, Pulse takes that intelligence and applies it proactively to your

319
00:19:33,440 --> 00:19:36,480
Ejaaz:
personal everyday life helping you make progress

320
00:19:36,480 --> 00:19:39,440
Ejaaz:
in the things that matter even when you don't think so it's almost

321
00:19:39,440 --> 00:19:42,300
Ejaaz:
as if they're saying these models were already available and

322
00:19:42,300 --> 00:19:45,160
Ejaaz:
intelligent and we're surfacing this feature this new

323
00:19:45,160 --> 00:19:49,160
Ejaaz:
product for you so that you can see it happen to you in real time literally

324
00:19:49,160 --> 00:19:53,600
Ejaaz:
change the way you work and interact with ai over time and i just want to give

325
00:19:53,600 --> 00:19:57,560
Ejaaz:
my quick take an opinion on this because we just covered like the doom of vision

326
00:19:57,560 --> 00:20:02,460
Ejaaz:
of like meta's uh scrolling a TikTok app, and then we have suddenly this.

327
00:20:03,000 --> 00:20:07,800
Ejaaz:
This is way more optimistic. I can see myself using this every day and actually

328
00:20:07,800 --> 00:20:09,920
Ejaaz:
growing to really depend and love it.

329
00:20:10,340 --> 00:20:13,380
Ejaaz:
Obviously, the Duma side of that is like, if I'm dependent on this and OpenAI

330
00:20:13,380 --> 00:20:16,320
Ejaaz:
has all my data, could they use that for different things?

331
00:20:16,440 --> 00:20:21,080
Ejaaz:
But the optimistic side is, do I care if it's making me more of a productive human being?

332
00:20:21,220 --> 00:20:25,800
Ejaaz:
If I can do 10x more things than I could originally do, if it can anticipate

333
00:20:25,800 --> 00:20:29,500
Ejaaz:
problems before they even happen to me, that's something that I'm willing not

334
00:20:29,500 --> 00:20:34,540
Ejaaz:
only to pay 200 bucks for, but maybe even a thousand bucks for if it ends up doing that, right?

335
00:20:35,180 --> 00:20:39,560
Ejaaz:
And then the stark contrast to point out the obvious is this seems to be a product

336
00:20:39,560 --> 00:20:43,680
Ejaaz:
that is sucking your attention to make you more productive, to make you smarter.

337
00:20:44,060 --> 00:20:48,360
Ejaaz:
Whereas Meta's thing seems to be like, you know, let me help you consume a bunch

338
00:20:48,360 --> 00:20:51,160
Ejaaz:
of video content and waste your time on that kind of stuff.

339
00:20:51,340 --> 00:20:54,960
Ejaaz:
That's the obvious. Do you have the same kind of take, Josh, or anything different?

340
00:20:54,960 --> 00:20:59,960
Josh:
Yeah, what you said reminded me of this war that I've been at with all of the

341
00:20:59,960 --> 00:21:03,280
Josh:
large model companies, which is against the singular text box.

342
00:21:03,500 --> 00:21:07,100
Josh:
Like the fact that you're presented with the world's knowledge and a single

343
00:21:07,100 --> 00:21:09,620
Josh:
text box to engage with it really drives me crazy.

344
00:21:10,400 --> 00:21:13,300
Josh:
Frustration to no end and what i love about

345
00:21:13,300 --> 00:21:16,140
Josh:
the open ai approach is well it's just it's proactive it

346
00:21:16,140 --> 00:21:19,720
Josh:
serves you things that you wouldn't necessarily require

347
00:21:19,720 --> 00:21:22,780
Josh:
that amount of knowledge for but it takes the large knowledge that's unused and

348
00:21:22,780 --> 00:21:25,420
Josh:
it applies it to something that's valuable for you and it serves it to you and i think

349
00:21:25,420 --> 00:21:28,260
Josh:
that's great and i mean meta meta just has

350
00:21:28,260 --> 00:21:31,240
Josh:
this this kind of like strange slightly less

351
00:21:31,240 --> 00:21:34,240
Josh:
inspiring approach and sure ai video it's not

352
00:21:34,240 --> 00:21:37,380
Josh:
that great now it's going to get fantastic i'm sure ai video

353
00:21:37,380 --> 00:21:40,520
Josh:
will i mean eventually become more interesting than

354
00:21:40,520 --> 00:21:44,180
Josh:
tiktok and it will become a very interesting valuable

355
00:21:44,180 --> 00:21:47,180
Josh:
platform it's just on a personal note leaves

356
00:21:47,180 --> 00:21:51,160
Josh:
me uninspired to see all this talent going towards that instead of something

357
00:21:51,160 --> 00:21:56,740
Josh:
like open ai so for today for this moment in time i am a little upset with meta

358
00:21:56,740 --> 00:22:01,400
Josh:
pretty happy with open ai we'll see how this changes over time but two very

359
00:22:01,400 --> 00:22:05,700
Josh:
interesting noteworthy product releases that had happened just very recently here?

360
00:22:05,820 --> 00:22:09,600
Ejaaz:
I think it is a tale of two different cities.

361
00:22:10,480 --> 00:22:14,480
Ejaaz:
Facebook and Meta is doing its own thing. It is a social media platform.

362
00:22:14,740 --> 00:22:17,940
Ejaaz:
They never claim to change the world and make you the most intelligent person

363
00:22:17,940 --> 00:22:19,800
Ejaaz:
ever. They're there to kind of connect people.

364
00:22:20,080 --> 00:22:24,860
Ejaaz:
And I'm ironically lacking the argument for connection with their new product.

365
00:22:25,140 --> 00:22:29,160
Ejaaz:
Addiction, I get. Connection, less so. And with the OpenAI product,

366
00:22:29,780 --> 00:22:34,940
Ejaaz:
Sam has always kind of pitched this vision of AGI, super intelligence,

367
00:22:35,160 --> 00:22:37,880
Ejaaz:
like we're going to make you smarter and it's going to make you way more productive

368
00:22:37,880 --> 00:22:40,200
Ejaaz:
and humans are going to prosper, UBI, all this kind of stuff.

369
00:22:40,460 --> 00:22:42,300
Ejaaz:
And I can see this as a step towards that.

370
00:22:42,620 --> 00:22:46,200
Ejaaz:
The thing that I'm most excited about in the OpenAI stuff is this seems like

371
00:22:46,200 --> 00:22:51,160
Ejaaz:
the premeditated step before launching your own personal agent that can do way

372
00:22:51,160 --> 00:22:55,640
Ejaaz:
more things than just talk to you. It can buy stuff for you and build stuff for you.

373
00:22:55,820 --> 00:22:59,400
Ejaaz:
And, you know, they release something this morning um that will allow you to

374
00:22:59,400 --> 00:23:02,580
Ejaaz:
like pay or buy uh anything which anyway is super cool

375
00:23:03,270 --> 00:23:07,370
Ejaaz:
Josh, we have a really exciting episode that we're filming on Thursday,

376
00:23:07,370 --> 00:23:12,490
Ejaaz:
which is with the guys that built this product, which built OpenAI Pulse. The Pulse product.

377
00:23:12,610 --> 00:23:13,390
Josh:
The one that we like today.

378
00:23:13,850 --> 00:23:17,410
Ejaaz:
The Pulse, the optimistic outlook. I'm highlighting the sentence.

379
00:23:17,410 --> 00:23:18,170
Josh:
Talking to the optimist.

380
00:23:18,570 --> 00:23:22,770
Ejaaz:
Yeah, I'm highlighting the end of Sam's announcement post where he goes,

381
00:23:23,090 --> 00:23:26,790
Ejaaz:
huge congrats to Christina and Samir and the team for building this.

382
00:23:26,910 --> 00:23:31,450
Ejaaz:
Christina and Samir are coming on our show on Thursday. We are recording that

383
00:23:31,450 --> 00:23:36,130
Ejaaz:
episode and we're going to dig deep into AI personalization and agents.

384
00:23:36,350 --> 00:23:39,010
Ejaaz:
And we cannot wait to show this to you guys.

385
00:23:39,470 --> 00:23:43,410
Ejaaz:
But that comes to the end of our episode. Unless Josh, you have any other hot takes?

386
00:23:43,610 --> 00:23:47,830
Josh:
That's everything. So yeah, if you guys are excited about the Christina and

387
00:23:47,830 --> 00:23:49,130
Josh:
Samir episode, please tune in.

388
00:23:49,230 --> 00:23:52,130
Josh:
That is like pretty huge for us to actually talk to the people who are responsible

389
00:23:52,130 --> 00:23:55,250
Josh:
for creating the news that we were discussing on the show. Super exciting.

390
00:23:55,590 --> 00:23:59,430
Josh:
One task for you in the comment section is to share which side you are on.

391
00:23:59,550 --> 00:24:00,670
Josh:
Are we being delusional?

392
00:24:01,290 --> 00:24:04,050
Josh:
Is it the pessimists that are correct? Is it the people who are building AI

393
00:24:04,050 --> 00:24:06,530
Josh:
Slap? Are you team meta? Are you team open AI today?

394
00:24:06,750 --> 00:24:08,210
Josh:
It's all right. You're allowed to change your opinion next week,

395
00:24:08,290 --> 00:24:09,370
Josh:
but for today, where do you stand?

396
00:24:09,730 --> 00:24:12,890
Josh:
And as always, if you enjoyed, please share it with a friend who would also

397
00:24:12,890 --> 00:24:14,490
Josh:
enjoy an episode from Limitless.

398
00:24:14,770 --> 00:24:17,770
Josh:
We appreciate your support always. Don't forget to like, subscribe, all the good things.

399
00:24:17,910 --> 00:24:21,550
Josh:
And we will see you again in the next one for a roundup of all of the other

400
00:24:21,550 --> 00:24:22,950
Josh:
stuff that we didn't get to talk about today.

401
00:24:23,270 --> 00:24:27,170
Josh:
So stay tuned. We have a lot coming up and we will talk to you all in the next one.