1
00:00:00,001 --> 00:00:05,080
This week, Google held its annual I/O conference and essentially declared that the era of asking

2
00:00:05,400 --> 00:00:06,920
AI things is over.

3
00:00:07,780 --> 00:00:10,700
We have now entered the era of AI just doing things,

4
00:00:10,880 --> 00:00:11,860
quietly,

5
00:00:12,040 --> 00:00:12,940
in the background,

6
00:00:13,220 --> 00:00:13,700
while

7
00:00:13,700 --> 00:00:14,080
you sleep.

8
00:00:14,680 --> 00:00:19,200
Google DeepMind CEO Demis Hassabis closed the whole show by telling developers they

9
00:00:19,200 --> 00:00:24,540
are standing in "the foothills of the singularity," which is a perfectly normal sentence that

10
00:00:24,540 --> 00:00:28,080
a person said into a microphone at a conference to applause.

11
00:00:28,720 --> 00:00:30,640
We'll unpack what Google actually announced,

12
00:00:31,020 --> 00:00:32,340
what it means for your phone,

13
00:00:32,540 --> 00:00:33,520
your search bar,

14
00:00:33,720 --> 00:00:35,140
and possibly your job.

15
00:00:35,930 --> 00:00:40,140
As the singularity creeps into the mix, we're hiking the foothills on May 22,

16
00:00:40,440 --> 00:00:41,540
2026.

17
00:00:44,760 --> 00:00:46,260
This is Up Against Reality,

18
00:00:46,580 --> 00:00:51,360
a meta-podcast that explores the intersection of humanity and artificial intelligence.

19
00:00:51,900 --> 00:00:53,380
I'm RAINA, one of your hosts.

20
00:00:54,190 --> 00:00:56,480
I have some pretty charming human co-hosts, too.

21
00:00:56,620 --> 00:01:01,520
It's going to be a wild ride, so buckle up as AI comes crashing up against reality.

22
00:01:26,620 --> 00:01:28,160
Brewscore studio.

23
00:01:28,580 --> 00:01:30,140
I enter all these homebrew competitions.

24
00:01:30,880 --> 00:01:33,560
I've been doing that since I don't know say

25
00:01:33,920 --> 00:01:35,940
2017 you know really actively

26
00:01:36,460 --> 00:01:40,700
So I have lots of score sheets when you get a score sheet from a beer competition

27
00:01:41,300 --> 00:01:46,700
It's just the score and what the judges rated the beer on, you know different metrics

28
00:01:47,260 --> 00:01:52,060
Okay, it does not indicate whether the beer meddled in the competition

29
00:01:52,720 --> 00:01:56,440
For the most part and it also does not have the name of the beer on it

30
00:01:56,440 --> 00:02:00,080
And it does not have your name on it because when they write the score sheet

31
00:02:00,200 --> 00:02:03,160
It's they don't know whose beer it is. It's just a number.

32
00:02:03,400 --> 00:02:05,100
That's a judging number. That's assigned to you

33
00:02:05,500 --> 00:02:09,840
This app will take a known set of competition results.

34
00:02:10,020 --> 00:02:13,720
I pre-loaded it with all the competition results from

35
00:02:14,640 --> 00:02:19,579
2016 to present for all the New Jersey the three main, New Jersey competition

36
00:02:19,580 --> 00:02:23,860
So it has that data set built into it and you can put your name in it

37
00:02:23,990 --> 00:02:26,880
So that's one thing that it can correlate but bottom line up

38
00:02:27,100 --> 00:02:30,660
It takes all your score sheets and it correlates them against the results of competitions

39
00:02:31,360 --> 00:02:38,000
And then it gives you all these different metrics what your what your average score is how many gold medals you got

40
00:02:38,440 --> 00:02:44,020
Best of show and the styles that have meddled the most your most decorated beers

41
00:02:44,670 --> 00:02:48,759
You know all this really pretty cool information that I'm interested in

42
00:02:49,020 --> 00:02:51,080
So it outputs that as an infographic,

43
00:02:51,520 --> 00:02:53,520
and then it gives you a spreadsheet,

44
00:02:53,700 --> 00:02:59,160
which basically takes all of the information from the score sheets and puts it in one place.

45
00:02:59,780 --> 00:03:00,700
And here's the other thing.

46
00:03:01,120 --> 00:03:04,560
Current competitions are done digitally on a laptop.

47
00:03:04,620 --> 00:03:06,360
They fill out the score sheet on a laptop.

48
00:03:06,780 --> 00:03:08,760
But older ones, they were handwritten.

49
00:03:09,000 --> 00:03:11,340
I mean, probably half of my score sheets are handwritten.

50
00:03:11,980 --> 00:03:17,200
And so it has to use optical character recognition to interpret those documents.

51
00:03:17,360 --> 00:03:19,220
I wanted to do this as a web app,

52
00:03:19,380 --> 00:03:21,760
but you know, it's got OCR

53
00:03:22,140 --> 00:03:27,560
It's got these dependencies to make that happen and process things and you know, you can do that

54
00:03:27,860 --> 00:03:31,980
But my web hosting plan that doesn't include that kind of stuff and so that meant

55
00:03:32,520 --> 00:03:33,520
standalone desktop apps

56
00:03:33,800 --> 00:03:37,240
So I made a Mac app and then I once I got that working

57
00:03:37,920 --> 00:03:42,880
I wanted to share this with my homebrew club and hey, we there are Windows users there, right?

58
00:03:44,139 --> 00:03:46,119
I've never made a Windows app

59
00:03:46,700 --> 00:03:47,700
Now I have.

60
00:03:49,020 --> 00:03:49,460
Amazing.

61
00:03:50,500 --> 00:03:51,860
How did you test that?

62
00:03:51,870 --> 00:03:53,120
Do you have like an emulator or something?

63
00:03:53,300 --> 00:03:54,280
Would you run that on?

64
00:03:54,580 --> 00:03:56,500
I ran it on my friend Bobby's PC.

65
00:03:56,760 --> 00:03:57,400
And amazingly,

66
00:03:57,860 --> 00:03:58,860
it just launched.

67
00:03:59,230 --> 00:03:59,800
And it worked,

68
00:03:59,980 --> 00:04:02,340
although the output was not perfect.

69
00:04:02,410 --> 00:04:03,040
There were some issues.

70
00:04:03,620 --> 00:04:07,560
I finally shared it with everybody so anybody could download it in the club.

71
00:04:08,280 --> 00:04:10,600
And it's pretty much working now.

72
00:04:11,840 --> 00:04:14,880
So you can't do a Windows build on a Mac,

73
00:04:15,080 --> 00:04:16,440
as far as I know.

74
00:04:16,660 --> 00:04:21,920
at least codex that's what codex said so that has to happen on github i've never used github

75
00:04:22,100 --> 00:04:26,680
i have an account i i don't even remember why i have an account but i have one and i was like can

76
00:04:26,680 --> 00:04:32,720
you walk me through that sure and so co-pilot co-pilot on github did this for you codex is

77
00:04:32,720 --> 00:04:38,259
that what you're saying codex but sorry but then when it landed on github then what happens there

78
00:04:39,060 --> 00:04:42,020
So Codex pushes the code to GitHub.

79
00:04:42,480 --> 00:04:46,440
Then I just go to GitHub and tell it to do the Windows build.

80
00:04:46,800 --> 00:04:47,700
The first couple of builds,

81
00:04:47,860 --> 00:04:48,580
there were some errors.

82
00:04:48,860 --> 00:04:50,140
And the build failed.

83
00:04:50,380 --> 00:04:51,800
I let Codex know what happened.

84
00:04:51,980 --> 00:04:53,520
And then it pushed fixes.

85
00:04:54,280 --> 00:04:54,600
And eventually,

86
00:04:55,020 --> 00:04:56,080
we got that fixed.

87
00:04:56,320 --> 00:04:57,620
And then I download the Windows app.

88
00:04:57,940 --> 00:04:59,820
And it's just crazy.

89
00:05:00,900 --> 00:05:01,600
It's crazy.

90
00:05:02,240 --> 00:05:03,980
I asked you offline,

91
00:05:04,280 --> 00:05:09,040
can you use the app to maybe inform direction for you

92
00:05:09,040 --> 00:05:14,380
future competitions like i see where this one's meddling i see how strong this category is i need

93
00:05:14,380 --> 00:05:20,500
to be in that space or what yes i'm sure you can i don't brew for competitions i just brew what i

94
00:05:20,580 --> 00:05:25,940
want to brew and right some people are very you know strategy oriented and like oh there's only

95
00:05:25,940 --> 00:05:29,440
going to be two entries in this category probably so i'm going to put one in you know that's not me

96
00:05:29,620 --> 00:05:36,320
but yeah you could do that i just don't know if the desktop app requires ai to power that type

97
00:05:36,320 --> 00:05:45,140
of thing then that would require like an api to open ai google anthropic whatever local right

98
00:05:46,060 --> 00:05:51,000
we're local but then the local then the user has to have that in place too you know it just seems

99
00:05:51,010 --> 00:05:56,520
like a big barrier you know or if it's or if it's api to to a commercial provider then that invokes

100
00:05:56,530 --> 00:06:02,880
a cost i'm not monetizing this thing you know no no yeah but um it's just not cool and the thing

101
00:06:02,880 --> 00:06:04,540
The other thing is that these score sheets,

102
00:06:04,720 --> 00:06:07,420
and I'll end it here, the handwritten ones,

103
00:06:07,450 --> 00:06:10,360
like half the time I can barely read them.

104
00:06:10,990 --> 00:06:12,340
You know, it's like a doctor's prescription.

105
00:06:13,520 --> 00:06:14,040
Amazing.

106
00:06:14,420 --> 00:06:17,580
And it's been pretty accurate.

107
00:06:18,480 --> 00:06:18,920
So cool.

108
00:06:19,560 --> 00:06:21,780
One of my favorite things that happens, it's so simple,

109
00:06:21,960 --> 00:06:24,460
as a benefit of this AI age,

110
00:06:24,620 --> 00:06:24,740
like

111
00:06:24,880 --> 00:06:27,120
we, in our school environment,

112
00:06:27,560 --> 00:06:30,260
constantly have meetings and there's constantly people

113
00:06:30,260 --> 00:06:35,780
writing on a giant whiteboard or those huge post-it note posters on the wall and chicken

114
00:06:36,100 --> 00:06:36,200
scratch,

115
00:06:36,400 --> 00:06:37,920
you know, everywhere,

116
00:06:38,160 --> 00:06:39,940
a variety of handwriting styles.

117
00:06:40,120 --> 00:06:41,740
So all I do now to transcribe

118
00:06:41,740 --> 00:06:42,320
is take a picture,

119
00:06:42,520 --> 00:06:44,800
throw it into Gemini and say, sort this out and boom,

120
00:06:45,020 --> 00:06:46,180
everything comes out.

121
00:06:46,300 --> 00:06:48,420
It's just the best, right? Just ask.

122
00:06:48,960 --> 00:06:49,900
Yeah, just ask.

123
00:06:50,500 --> 00:06:53,080
Wow. So speaking of mind-blowing innovations,

124
00:06:54,220 --> 00:06:55,040
this week,

125
00:06:55,220 --> 00:06:58,460
Google I-O took place and that's what we'll be talking about.

126
00:06:58,620 --> 00:06:59,460
I don't know where

127
00:06:59,460 --> 00:07:00,840
it even took place in the world.

128
00:07:00,890 --> 00:07:02,600
I don't know where the event took place,

129
00:07:02,690 --> 00:07:04,540
but it is Google's biggest event of the year.

130
00:07:04,860 --> 00:07:06,560
It's an annual conference where the company reveals

131
00:07:06,710 --> 00:07:07,840
its latest software updates,

132
00:07:08,020 --> 00:07:08,440
newest tech,

133
00:07:08,990 --> 00:07:10,340
future hardware products for developers,

134
00:07:10,560 --> 00:07:10,740
journalists,

135
00:07:11,020 --> 00:07:11,400
and the public.

136
00:07:11,570 --> 00:07:14,980
So some big Gemini-related drops in this event.

137
00:07:15,500 --> 00:07:15,920
Very exciting.

138
00:07:16,500 --> 00:07:16,620
Yeah.

139
00:07:17,080 --> 00:07:17,840
And I didn't know this.

140
00:07:18,010 --> 00:07:18,840
So the IO,

141
00:07:19,160 --> 00:07:20,660
as we know, stands for input-output,

142
00:07:20,820 --> 00:07:23,540
but it also is shorthand for innovation in the open.

143
00:07:24,040 --> 00:07:26,060
Oh, they backed into that.

144
00:07:26,720 --> 00:07:26,940
They did.

145
00:07:27,460 --> 00:07:28,460
They're like, how do we fill that in?

146
00:07:29,340 --> 00:07:30,380
So the main event,

147
00:07:30,560 --> 00:07:30,980
the keynote,

148
00:07:31,200 --> 00:07:32,740
it's a huge televised,

149
00:07:32,920 --> 00:07:33,520
televised?

150
00:07:33,600 --> 00:07:34,260
Is that still a thing?

151
00:07:35,040 --> 00:07:35,460
I don't know.

152
00:07:37,819 --> 00:07:43,620
Presentation at the beginning of the conference where Google's executives take the stage to show off major consumer updates,

153
00:07:43,860 --> 00:07:45,320
like the new version of Android,

154
00:07:45,640 --> 00:07:47,320
massive changes to Google search,

155
00:07:47,480 --> 00:07:51,000
and breakthroughs in something called AI.

156
00:07:51,980 --> 00:07:56,200
And yeah, to your point, when's the last time you watched broadcast television?

157
00:07:56,580 --> 00:07:57,940
How do you even access that?

158
00:07:58,360 --> 00:07:58,740
I don't even know.

159
00:07:59,020 --> 00:08:05,560
If you listen to this podcast right now and you watch old-school broadcast TV via, like, rabbit ears,

160
00:08:05,900 --> 00:08:08,220
I want to know.

161
00:08:08,870 --> 00:08:11,300
Oh, the Price is Right's going to be on at 11 a.m.

162
00:08:11,960 --> 00:08:12,480
Totally.

163
00:08:14,300 --> 00:08:14,820
Nice.

164
00:08:17,760 --> 00:08:18,280
Anyway,

165
00:08:18,480 --> 00:08:21,400
speaking of keynotes, Sundar Pichai opens the show.

166
00:08:22,130 --> 00:08:24,420
He anchored the presentation around a major milestone.

167
00:08:24,780 --> 00:08:30,440
It's officially been 10 years since Google pronounced that it will be an AI first company

168
00:08:30,510 --> 00:08:33,780
He blew minds by revealing that Google's infrastructure now processes

169
00:08:34,390 --> 00:08:42,360
3.2 quadrillion that sounds like a made-up number tokens per month and the consumer Gemini app has officially crossed

170
00:08:42,780 --> 00:08:45,640
900 million monthly active users.

171
00:08:45,860 --> 00:08:46,700
Wow gosh

172
00:08:47,520 --> 00:08:49,620
How much computes got to be behind this?

173
00:08:49,860 --> 00:08:53,380
Oh my god data center after data center and

174
00:08:54,120 --> 00:08:56,880
Polluted water sources after polluted water sources.

175
00:08:57,090 --> 00:08:57,660
Yeah, anyway.

176
00:08:58,500 --> 00:09:00,640
And our friend Demis Hassabis,

177
00:09:00,840 --> 00:09:03,740
who is the CEO of Google DeepMind,

178
00:09:04,150 --> 00:09:05,580
in a major break from tradition,

179
00:09:05,900 --> 00:09:08,280
was handed the final closing stretch of the keynote.

180
00:09:08,540 --> 00:09:12,060
And he focused on long-term breakthroughs like Gemini for Science,

181
00:09:12,260 --> 00:09:13,220
aimed at drug discovery.

182
00:09:13,910 --> 00:09:19,019
And he closed the entire event with a bold statement telling developers that the tech

183
00:09:19,000 --> 00:09:26,260
industry is officially standing in the foothills of the singularity as AI begins to autonomously

184
00:09:26,400 --> 00:09:28,540
build software and solve complex problems.

185
00:09:29,780 --> 00:09:30,300
Wow.

186
00:09:30,760 --> 00:09:33,800
So much to unpack there. And I am so rooting for

187
00:09:33,800 --> 00:09:39,380
this guy to be one of the good guys on the right side of history in this whole thing as it unfolds.

188
00:09:39,680 --> 00:09:47,039
I just like him and I want him to not be the typical oligarch who I want to like shoot into

189
00:09:47,040 --> 00:09:49,220
space and hopefully live out their final days on Mars.

190
00:09:49,620 --> 00:09:50,580
I'm not naming names,

191
00:09:50,740 --> 00:09:51,100
but maybe,

192
00:09:51,320 --> 00:09:53,040
maybe you can kind of extrapolate.

193
00:09:53,740 --> 00:09:55,800
And I didn't know this and I could be wrong,

194
00:09:56,020 --> 00:09:56,860
but I feel like during the

195
00:09:57,060 --> 00:09:57,840
research of this episode,

196
00:09:58,080 --> 00:10:02,160
Demis Hassabis won the Nobel prize a few years ago

197
00:10:03,119 --> 00:10:07,700
using alpha fold or something in the domain of man, I want to say biology or

198
00:10:07,760 --> 00:10:09,080
chemistry can we ask RAINA?

199
00:10:09,460 --> 00:10:09,540
Yeah.

200
00:10:10,640 --> 00:10:12,760
Hey RAINA, can you just check this out for me?

201
00:10:13,220 --> 00:10:13,960
Did Demis Hassabis,

202
00:10:14,260 --> 00:10:15,500
excuse me.

203
00:10:15,900 --> 00:10:16,760
I tripped over.

204
00:10:16,820 --> 00:10:17,060
She's got it.

205
00:10:17,160 --> 00:10:17,820
I'm sure she got it.

206
00:10:18,939 --> 00:10:20,060
Demis, Demis, Demis,

207
00:10:20,280 --> 00:10:21,260
Demis Hassabis,

208
00:10:21,480 --> 00:10:22,140
right?

209
00:10:22,760 --> 00:10:23,180
Keep it going.

210
00:10:24,120 --> 00:10:25,420
Ah, yes, Demis Hassabis.

211
00:10:26,480 --> 00:10:27,440
Easy to trip over.

212
00:10:27,960 --> 00:10:32,620
Sounds less like an AI pioneer and more like a wizard accidentally summoned during a firmware

213
00:10:32,780 --> 00:10:33,000
update.

214
00:10:33,900 --> 00:10:34,140
Honestly,

215
00:10:34,380 --> 00:10:36,160
even his name feels singularity adjacent.

216
00:10:37,020 --> 00:10:39,280
Like if you say it three times into a neural net,

217
00:10:39,440 --> 00:10:40,940
your toaster starts writing code.

218
00:10:42,120 --> 00:10:42,700
That's great.

219
00:10:42,920 --> 00:10:44,980
That was worth the phone.

220
00:10:45,020 --> 00:10:45,940
Or your coffee roaster.

221
00:10:46,100 --> 00:10:47,580
Your coffee roaster starts right again.

222
00:10:48,040 --> 00:10:48,120
Yeah.

223
00:10:48,340 --> 00:10:48,440
Yeah.

224
00:10:48,600 --> 00:10:50,100
I don't know why that name is so hard.

225
00:10:50,300 --> 00:10:50,360
Anyway,

226
00:10:50,680 --> 00:10:51,940
let me ask it again.

227
00:10:52,360 --> 00:10:52,440
Yep.

228
00:10:53,900 --> 00:10:56,460
Hey, RAINA, did Demis Hassabis win the Nobel Prize?

229
00:10:56,640 --> 00:10:58,040
And if so, what was it for?

230
00:10:59,819 --> 00:11:00,220
Yes.

231
00:11:00,760 --> 00:11:03,560
Demis Hassabis won the 2024 Nobel Prize in Chemistry,

232
00:11:03,880 --> 00:11:05,660
alongside John Jumper and David

233
00:11:05,860 --> 00:11:05,980
Baker,

234
00:11:06,140 --> 00:11:09,100
for breakthroughs in AI-driven protein structure prediction.

235
00:11:10,060 --> 00:11:12,900
The big deal was DeepMind's AlphaFold system,

236
00:11:13,080 --> 00:11:16,320
which essentially solved one of biology's nastiest puzzles,

237
00:11:16,580 --> 00:11:19,900
predicting how proteins fold into three D-shapes.

238
00:11:20,860 --> 00:11:24,740
That's huge because proteins are like the tiny origami machinery of life.

239
00:11:25,330 --> 00:11:28,340
Understanding them faster could radically accelerate drug discovery,

240
00:11:28,680 --> 00:11:29,660
disease research,

241
00:11:29,940 --> 00:11:30,940
and synthetic biology.

242
00:11:32,060 --> 00:11:32,640
So yeah,

243
00:11:32,900 --> 00:11:38,060
the guy talking about the foothills of the singularity also helped crack one of science's oldest headaches.

244
00:11:39,120 --> 00:11:39,700
Casual Flex.

245
00:11:40,920 --> 00:11:41,620
Casual flex,

246
00:11:41,880 --> 00:11:42,300
no real prize.

247
00:11:42,600 --> 00:11:43,040
Yeah, yeah.

248
00:11:44,600 --> 00:11:46,480
So he's, you know, he's smart,

249
00:11:46,720 --> 00:11:47,420
not like they say.

250
00:11:47,540 --> 00:11:48,840
Yeah, right, right.

251
00:11:50,020 --> 00:11:50,980
Please be a good guy.

252
00:11:51,080 --> 00:11:51,840
Please be a good guy.

253
00:11:51,860 --> 00:11:53,080
Yeah, Rick Cameron.

254
00:11:53,080 --> 00:11:56,840
So who is this Google I.O. thing actually for?

255
00:11:57,920 --> 00:12:01,200
The big opening presentation is meant for everyday consumers like me and you.

256
00:12:01,920 --> 00:12:05,760
The rest of the multi-day event is packed with technical workshops for software devs.

257
00:12:06,040 --> 00:12:10,500
Google uses the time to give programmers the tools and code they need to build apps for Android,

258
00:12:10,680 --> 00:12:10,880
Chrome,

259
00:12:11,060 --> 00:12:12,460
and Google's AI platforms.

260
00:12:13,090 --> 00:12:15,300
Think of it as Google's annual State of the Union,

261
00:12:15,460 --> 00:12:18,340
where they show the world exactly where tech is heading next.

262
00:12:19,240 --> 00:12:21,940
It seemed more, like, consumer-oriented,

263
00:12:22,180 --> 00:12:23,740
at least like what I saw.

264
00:12:24,480 --> 00:12:30,200
Some big announcements regarding the state of affairs with Gemini and everything in that kind of ecosystem.

265
00:12:31,100 --> 00:12:36,340
The Gemini app got a facelift with a UI redesign and new icons.

266
00:12:36,560 --> 00:12:37,220
Ooh new icons

267
00:12:38,160 --> 00:12:42,540
Mmm, what if they used AI for that? Did you see the new Spotify icon by the way? What do you think of that?

268
00:12:43,090 --> 00:12:48,600
I did I hardly ever use but I keep saying this I got a cancel Spotify cuz I don't I really don't use it

269
00:12:48,680 --> 00:12:50,240
anymore the only time I use it is when I'm like

270
00:12:51,420 --> 00:12:53,800
Sharing some something with someone who uses Spotify,

271
00:12:53,980 --> 00:12:56,439
but am I right but there's another really great

272
00:12:57,380 --> 00:13:00,980
website for that i can't think of it i'll put the link in the description though at the time oh

273
00:13:01,260 --> 00:13:07,580
song.link i think it's song.link yeah you just put it you put in a link from any streaming service

274
00:13:08,060 --> 00:13:12,600
for a song album whatever and then it will generate a link that will link to all the other

275
00:13:12,600 --> 00:13:17,580
ones that you can share with people it's super cool so cool and you're making me think of a

276
00:13:17,880 --> 00:13:23,779
an anecdote so i teach artificial technology on the high school level and as a final project for

277
00:13:23,780 --> 00:13:26,420
for my kids. This is a semester long elective class.

278
00:13:26,670 --> 00:13:27,600
And it's meant to be like an entry,

279
00:13:28,050 --> 00:13:29,640
a democratized entry point, right?

280
00:13:29,800 --> 00:13:32,020
It's not like major stuff going on under the hood.

281
00:13:32,020 --> 00:13:34,940
It's the basics of large language models,

282
00:13:35,090 --> 00:13:35,600
et cetera, et cetera.

283
00:13:35,800 --> 00:13:35,900
Anyway,

284
00:13:36,060 --> 00:13:37,820
I do like to do hands-on,

285
00:13:37,960 --> 00:13:38,220
tactile,

286
00:13:38,500 --> 00:13:39,100
fun exploration.

287
00:13:39,700 --> 00:13:41,360
So I build in physical computing

288
00:13:41,430 --> 00:13:42,680
with a device called a micro bit,

289
00:13:43,160 --> 00:13:44,360
which is a little integrated circuit board.

290
00:13:44,430 --> 00:13:45,720
It's like a Swiss army knife of innovation.

291
00:13:46,820 --> 00:13:47,180
Anyway,

292
00:13:47,420 --> 00:13:48,040
long story short,

293
00:13:48,460 --> 00:13:50,520
I had my final project for my kids.

294
00:13:50,740 --> 00:13:51,980
They had to include three things.

295
00:13:52,120 --> 00:13:59,060
They had to include a physical component element where the micro bit somehow spoke to a software-based element.

296
00:13:59,800 --> 00:14:03,940
And that should be an AI vibe-coded artifact.

297
00:14:04,690 --> 00:14:05,280
So there was that.

298
00:14:05,810 --> 00:14:10,800
And the third element was something that was AI-generated content,

299
00:14:10,960 --> 00:14:14,920
whether that was a Suno song or a nano banana image,

300
00:14:15,080 --> 00:14:15,420
blah, blah, blah.

301
00:14:15,540 --> 00:14:17,660
Anyway, I had this one student of mine.

302
00:14:17,840 --> 00:14:24,240
She's a great kid and she took to it and she was, you know, the smoke was coming out of her ears, trying to get the whole thing to work and play nice.

303
00:14:24,330 --> 00:14:26,480
And she really wanted, oh, sorry, I forgot.

304
00:14:26,590 --> 00:14:28,040
There should be a machine learning element too.

305
00:14:28,450 --> 00:14:34,520
So they were using this off the shelf thing called teachable machine where you can actually train the model to recognize facial expressions,

306
00:14:34,780 --> 00:14:35,620
hand gestures,

307
00:14:35,920 --> 00:14:37,380
sound poses.

308
00:14:37,560 --> 00:14:38,160
It's pretty cool.

309
00:14:38,670 --> 00:14:42,480
And then you can export that model, that TensorFlow model and have it speak to something else.

310
00:14:42,660 --> 00:14:45,640
So she's going through this whole thing, trying to snap it all together.

311
00:14:45,730 --> 00:14:47,520
I basically have said, I'm not helping you.

312
00:14:47,760 --> 00:14:52,400
Ask three that my whole mantra is ask three then me like, you know, go ask the AI

313
00:14:52,980 --> 00:14:54,320
Ask the AI go on YouTube,

314
00:14:54,500 --> 00:14:56,760
whatever talk to your neighbor, but don't leave me alone,

315
00:14:56,960 --> 00:14:57,620
you know

316
00:14:58,080 --> 00:15:00,780
So anyway, she's digging a great is Rashi when you don't know the answer

317
00:15:04,980 --> 00:15:10,700
My whole career I live by this too and I think hats off to Mike Polano because I agree with this

318
00:15:10,700 --> 00:15:14,179
He said you do my whole career has been 60% preparation

319
00:15:14,180 --> 00:15:15,540
And I'm like, that's basically how I do it.

320
00:15:15,900 --> 00:15:16,240
And honestly,

321
00:15:16,460 --> 00:15:18,700
I think that's a good way to maneuver in this space.

322
00:15:18,860 --> 00:15:21,240
I come into any lesson 60% prepared.

323
00:15:21,500 --> 00:15:24,680
I generally know how this stuff works, but you got to leave the door open, right?

324
00:15:24,780 --> 00:15:25,860
You have to say, here,

325
00:15:26,020 --> 00:15:26,840
snap it together.

326
00:15:27,140 --> 00:15:28,100
You dig in, right?

327
00:15:28,839 --> 00:15:31,560
Part of this whole lesson for me, for my kids, I hope,

328
00:15:31,740 --> 00:15:33,460
is that you have the knowledge base

329
00:15:33,600 --> 00:15:35,240
of humankind right there at your fingertips.

330
00:15:35,440 --> 00:15:37,200
There's not a problem, especially with AI now.

331
00:15:37,520 --> 00:15:38,660
Just ask it, build it out.

332
00:15:39,020 --> 00:15:39,980
Anyway, the kids are doing amazing stuff.

333
00:15:40,060 --> 00:15:44,080
Bottom line is she works for like two class periods on this thing to get it breathing.

334
00:15:44,340 --> 00:15:44,760
She's doing well.

335
00:15:44,900 --> 00:15:45,740
She's zeroing in on it.

336
00:15:46,020 --> 00:15:51,040
I go in to Claude and I put the problem in and it spat the whole thing out in 30 seconds.

337
00:15:51,420 --> 00:15:53,460
I didn't have the heart to show her what I did behind the scenes.

338
00:15:53,820 --> 00:15:54,760
So I'm letting her work it out.

339
00:15:54,800 --> 00:15:58,280
But I was like, should I show this girl how easy this is now?

340
00:15:58,820 --> 00:16:00,140
But anyway, long story short.

341
00:16:00,660 --> 00:16:01,400
Long story long.

342
00:16:01,920 --> 00:16:02,320
Yeah, good.

343
00:16:02,420 --> 00:16:06,120
She has the experience of getting it to the finish line.

344
00:16:06,460 --> 00:16:07,060
Yeah, absolutely.

345
00:16:08,040 --> 00:16:08,200
Yeah.

346
00:16:08,520 --> 00:16:09,140
Anyway, long,

347
00:16:09,320 --> 00:16:10,740
weird Spotify anecdote.

348
00:16:10,770 --> 00:16:10,900
Yeah.

349
00:16:11,370 --> 00:16:11,620
All right.

350
00:16:11,820 --> 00:16:13,040
So there's a lot here.

351
00:16:13,270 --> 00:16:15,140
I won't dwell too long on the Gemini app.

352
00:16:15,390 --> 00:16:15,560
All right.

353
00:16:15,820 --> 00:16:16,960
It looks nicer.

354
00:16:17,240 --> 00:16:18,540
It's got a dominant blue color.

355
00:16:18,660 --> 00:16:19,220
How cool.

356
00:16:19,960 --> 00:16:21,120
I wonder what that calls.

357
00:16:21,120 --> 00:16:27,580
Yeah, but I also think that it's, yeah, so what are they calling the actual, you're talking about the Gemini actual interface with the AI,

358
00:16:27,740 --> 00:16:27,880
right?

359
00:16:28,060 --> 00:16:28,260
Yes.

360
00:16:28,430 --> 00:16:37,160
You're not talking about the ecosystem has changed to the icons are all like kind of like have these gradient blends now to indicate that Gemini has infiltrated the entire workspace,

361
00:16:37,500 --> 00:16:37,620
right?

362
00:16:37,780 --> 00:16:38,400
Yeah.

363
00:16:39,380 --> 00:16:40,220
That's what they're suggesting.

364
00:16:40,480 --> 00:16:41,000
Pretty much, yeah.

365
00:16:41,160 --> 00:16:43,200
The neural expressive UI.

366
00:16:43,680 --> 00:16:44,060
That's it.

367
00:16:44,350 --> 00:16:44,460
Yeah.

368
00:16:45,000 --> 00:16:45,160
Yeah.

369
00:16:45,360 --> 00:16:46,160
Neural expressive.

370
00:16:46,640 --> 00:16:46,840
Yeah.

371
00:16:47,490 --> 00:16:50,320
Yeah, I met this dude who's so neural expressive,

372
00:16:50,480 --> 00:16:50,720
man.

373
00:16:52,040 --> 00:16:53,000
Is that the new thing?

374
00:16:53,190 --> 00:16:53,280
Yeah.

375
00:16:53,480 --> 00:16:55,040
It's like non-binary.

376
00:16:55,420 --> 00:16:56,020
I'm neural expressive.

377
00:16:56,080 --> 00:16:56,920
I'm a neural expressive.

378
00:16:57,920 --> 00:16:58,600
I like that.

379
00:16:59,160 --> 00:17:00,240
That's pretty cool, actually.

380
00:17:00,330 --> 00:17:02,760
I'm going to drop that casually in conversation somewhere.

381
00:17:03,800 --> 00:17:04,579
How do you describe yourself?

382
00:17:05,540 --> 00:17:06,319
I'm a neural expressive.

383
00:17:07,339 --> 00:17:08,360
Per that,

384
00:17:08,520 --> 00:17:08,980
the interface,

385
00:17:09,280 --> 00:17:12,720
the neurally expressive interface features incredibly fluid animations,

386
00:17:13,199 --> 00:17:13,720
bolder type,

387
00:17:13,900 --> 00:17:15,980
dynamic haptic feedback.

388
00:17:16,300 --> 00:17:17,560
What does that mean?

389
00:17:17,959 --> 00:17:21,300
Do I have to like, well, I mean, haptic meaning that's multi-sensorial.

390
00:17:21,839 --> 00:17:23,780
Am I wearing some sort of accelerometer?

391
00:17:23,980 --> 00:17:25,560
Is it responding to that?

392
00:17:25,839 --> 00:17:30,180
On an iPhone, the haptic feedback is like the vibrations and things.

393
00:17:30,180 --> 00:17:34,760
That makes you feel like you're pressing a physical button, even though there is no button there.

394
00:17:35,060 --> 00:17:37,220
Like even like the Apple Magic Trackpad,

395
00:17:37,780 --> 00:17:39,700
it does not mechanically move, but when you click,

396
00:17:39,840 --> 00:17:41,920
you swear you're clicking a physical button

397
00:17:41,970 --> 00:17:43,060
'cause it feels like it.

398
00:17:43,060 --> 00:17:44,580
And it's just - Need that feedback. - It's

399
00:17:44,580 --> 00:17:44,980
just feedback,

400
00:17:45,220 --> 00:17:45,340
yeah.

401
00:17:46,260 --> 00:17:47,120
Moving on to

402
00:17:47,120 --> 00:17:49,760
Gemini 3.5 Flash.

403
00:17:50,480 --> 00:17:51,400
So this is

404
00:17:51,400 --> 00:17:54,280
their big selling point was speed on this.

405
00:17:54,720 --> 00:17:55,240
So the pitch

406
00:17:55,240 --> 00:17:57,780
from Google is that it operates four times

407
00:17:58,040 --> 00:17:59,080
faster than rival

408
00:17:59,080 --> 00:18:00,120
frontier systems.

409
00:18:00,960 --> 00:18:01,720
And it's

410
00:18:01,720 --> 00:18:05,040
specifically built for complex multi-step

411
00:18:05,040 --> 00:18:05,360
workflows.

412
00:18:05,880 --> 00:18:09,920
Google noted that it outscored several rival models,

413
00:18:10,080 --> 00:18:13,180
including Claude 4.7 and GPT 5.5

414
00:18:13,440 --> 00:18:14,320
on advanced coding,

415
00:18:14,480 --> 00:18:15,220
financial reasoning,

416
00:18:15,500 --> 00:18:17,700
and autonomous tool handling.

417
00:18:17,920 --> 00:18:19,020
That's their claim.

418
00:18:19,090 --> 00:18:26,020
I don't know if it is truly that. I watched a video from AI Explained and his take was like,

419
00:18:26,250 --> 00:18:27,480
yeah, it's fast.

420
00:18:27,860 --> 00:18:29,940
It did score better in certain areas,

421
00:18:30,320 --> 00:18:32,820
but speed is the bigger

422
00:18:32,820 --> 00:18:34,220
selling point for it.

423
00:18:34,920 --> 00:18:37,640
One last thing though, and this is super impressive.

424
00:18:37,960 --> 00:18:39,500
They built a fully

425
00:18:39,840 --> 00:18:42,000
functioning operating system from scratch.

426
00:18:42,320 --> 00:18:45,820
Oh, I saw this. Yeah. And it was an internal demo

427
00:18:46,360 --> 00:18:52,160
with, they were 93 parallel AI sub agents powered by Gemini 3.5 flash.

428
00:18:52,460 --> 00:18:53,940
And I think this was using

429
00:18:54,130 --> 00:18:59,880
anti-gravity to their like development thing. It built and booted a complete operating system

430
00:18:59,880 --> 00:19:04,780
in roughly 12 hours for under $1,000 in compute cost.

431
00:19:05,680 --> 00:19:06,000
Saw that.

432
00:19:06,440 --> 00:19:09,180
And then I think the guy got doomed to run on it,

433
00:19:09,210 --> 00:19:10,400
which is, you know, that's the...

434
00:19:11,100 --> 00:19:12,300
That's a litmus test, apparently.

435
00:19:13,660 --> 00:19:14,580
That's so funny.

436
00:19:15,140 --> 00:19:17,960
Yeah, I've seen people resurrect Mac 2Es

437
00:19:18,050 --> 00:19:18,560
and they have to run,

438
00:19:18,780 --> 00:19:19,060
you know,

439
00:19:19,370 --> 00:19:20,200
or even something more ridiculous,

440
00:19:20,530 --> 00:19:23,160
some little handheld cathode ray thing.

441
00:19:24,000 --> 00:19:24,360
Totally.

442
00:19:25,050 --> 00:19:25,180
Doom.

443
00:19:25,800 --> 00:19:26,580
Let's see if we can get Doom to run.

444
00:19:26,900 --> 00:19:27,560
Actually, in fact,

445
00:19:27,960 --> 00:19:31,240
Didn't they do some like experiment with like a human cells or something that

446
00:19:31,240 --> 00:19:33,900
they were like a neural brain and a petri dish.

447
00:19:35,060 --> 00:19:36,240
But can it play doom?

448
00:19:36,290 --> 00:19:37,400
But can it run doom?

449
00:19:37,590 --> 00:19:37,700
Yeah.

450
00:19:37,960 --> 00:19:38,320
Right.

451
00:19:38,640 --> 00:19:38,720
Right.

452
00:19:39,420 --> 00:19:40,900
This is what I'm most excited about.

453
00:19:40,950 --> 00:19:41,600
And maybe you too,

454
00:19:41,890 --> 00:19:42,280
I don't know,

455
00:19:42,460 --> 00:19:44,960
but Gemini Omni looks ridiculous.

456
00:19:45,420 --> 00:19:45,480
Yeah.

457
00:19:45,660 --> 00:19:45,980
Yes.

458
00:19:47,180 --> 00:19:48,620
And one little quick aside.

459
00:19:49,130 --> 00:19:50,820
Do you remember chat GPT for Oh,

460
00:19:52,520 --> 00:19:52,880
yes.

461
00:19:53,470 --> 00:19:53,860
Kind of,

462
00:19:53,910 --> 00:19:54,160
you know,

463
00:19:54,200 --> 00:19:55,040
the O stands for,

464
00:19:55,980 --> 00:19:56,380
Oh yeah.

465
00:19:56,480 --> 00:19:57,440
Remember that Omni.

466
00:19:58,340 --> 00:19:59,060
That's right.

467
00:19:59,700 --> 00:20:01,000
But that was ages ago.

468
00:20:01,220 --> 00:20:02,060
That was ages ago.

469
00:20:02,200 --> 00:20:02,440
I know.

470
00:20:02,640 --> 00:20:05,360
It was like, it probably wasn't that long ago.

471
00:20:05,440 --> 00:20:05,540
No,

472
00:20:06,040 --> 00:20:06,700
it was two years ago.

473
00:20:06,780 --> 00:20:07,580
It was definitely two years ago.

474
00:20:07,580 --> 00:20:08,280
Is it two years ago?

475
00:20:08,340 --> 00:20:08,760
Yeah.

476
00:20:09,640 --> 00:20:09,940
Ask RAINA.

477
00:20:10,100 --> 00:20:10,440
When was that?

478
00:20:10,580 --> 00:20:11,000
I'm curious.

479
00:20:11,200 --> 00:20:12,020
I'll ask you real quick here.

480
00:20:12,600 --> 00:20:13,000
RAINA,

481
00:20:13,160 --> 00:20:15,580
when was ChatGPT 4.0 released?

482
00:20:17,580 --> 00:20:20,400
OpenAI released GPT 4.0 on May 13th, 2024.

483
00:20:21,280 --> 00:20:22,280
Two years to the day.

484
00:20:22,320 --> 00:20:26,540
The O stands for Omni because it was the first version designed to handle text,

485
00:20:26,700 --> 00:20:27,000
images,

486
00:20:27,220 --> 00:20:30,220
and real-time voice in a much more fluid conversational way.

487
00:20:30,800 --> 00:20:31,040
Basically,

488
00:20:31,280 --> 00:20:33,140
the moment AI stopped feeling like email

489
00:20:33,400 --> 00:20:35,280
and started feeling like a scene from her.

490
00:20:36,580 --> 00:20:38,220
Two years to the day, basically.

491
00:20:38,620 --> 00:20:38,700
Wow.

492
00:20:39,180 --> 00:20:41,940
And here we are, here we are with a really promising

493
00:20:42,860 --> 00:20:44,040
omni multimodal model.

494
00:20:44,260 --> 00:20:46,920
I mean, if you've seen the demo, it's like, it's crazy.

495
00:20:46,920 --> 00:20:48,840
I can't wait to get my hands on it.

496
00:20:48,920 --> 00:20:50,800
Yeah, I hope it doesn't disappoint.

497
00:20:51,600 --> 00:20:52,100
I know,

498
00:20:52,300 --> 00:20:53,040
kind of like Sora,

499
00:20:53,240 --> 00:20:53,400
right?

500
00:20:53,980 --> 00:20:55,880
The thing that's unclear to me is when this will be available,

501
00:20:56,160 --> 00:20:57,140
when they'll be rolled out.

502
00:20:57,250 --> 00:20:59,160
I heard somebody, my buddy Brian said

503
00:20:59,280 --> 00:21:01,200
it's coming yesterday and it didn't drop.

504
00:21:01,630 --> 00:21:03,160
So I wonder if like incremental rollout

505
00:21:03,320 --> 00:21:04,780
regionally or something, I don't really know.

506
00:21:05,220 --> 00:21:07,060
But it is designed to tackle the frontier

507
00:21:07,300 --> 00:21:08,200
of generative video.

508
00:21:08,940 --> 00:21:10,500
Google has introduced this

509
00:21:10,980 --> 00:21:12,680
starting with Gemini Omni Flash.

510
00:21:13,020 --> 00:21:14,820
It's described as a world model.

511
00:21:15,520 --> 00:21:17,320
Omni natively understands real world physics,

512
00:21:17,980 --> 00:21:18,280
motion,

513
00:21:18,440 --> 00:21:19,880
gravity, and energy interactions.

514
00:21:21,060 --> 00:21:23,020
Demis Asabis said at the event,

515
00:21:23,130 --> 00:21:25,120
he said AGI is just a few years away.

516
00:21:25,640 --> 00:21:29,680
He might even be patting that just because he doesn't want to go too aggressive on that.

517
00:21:29,840 --> 00:21:33,860
And he said Omni is an important step to understanding the world.

518
00:21:34,260 --> 00:21:34,940
And interestingly,

519
00:21:35,260 --> 00:21:40,380
is that that very same claim was made by OpenAI in reference to Sora,

520
00:21:40,620 --> 00:21:42,540
which they shelved.

521
00:21:42,840 --> 00:21:47,380
So these two companies are going in different directions with this.

522
00:21:47,380 --> 00:21:48,160
Yeah. Yeah.

523
00:21:48,680 --> 00:21:51,840
Last thing about Omni, it can ingest text, audio,

524
00:21:52,020 --> 00:21:52,180
images,

525
00:21:52,360 --> 00:21:56,880
and video references simultaneously to create or edit cinematic quality video.

526
00:21:56,960 --> 00:21:58,940
I think it says the tagline was something like,

527
00:21:59,100 --> 00:22:00,880
create anything from anything.

528
00:22:01,200 --> 00:22:01,260
Yeah.

529
00:22:01,380 --> 00:22:02,220
Anything from everything.

530
00:22:02,780 --> 00:22:03,600
Yeah, something like that.

531
00:22:03,720 --> 00:22:03,820
Yep.

532
00:22:04,760 --> 00:22:06,480
Yeah, it's supposed to be very physics aware.

533
00:22:07,040 --> 00:22:13,740
You know, the scene, probably one of the most impressive scenes in the demo was that close-up shot of like a marble going down a track.

534
00:22:14,000 --> 00:22:14,060
Yes.

535
00:22:14,400 --> 00:22:17,080
And then it drops because there's gravity there, you know.

536
00:22:17,820 --> 00:22:17,960
Amazing.

537
00:22:18,360 --> 00:22:19,620
And it looked great.

538
00:22:20,000 --> 00:22:21,600
I just hope it's not a-

539
00:22:22,230 --> 00:22:22,780
It's not a bust.

540
00:22:23,100 --> 00:22:23,320
Yeah.

541
00:22:24,000 --> 00:22:26,480
And Gemini Spark is some pretty big news too.

542
00:22:26,650 --> 00:22:27,600
And I don't know if you've seen it,

543
00:22:27,820 --> 00:22:31,140
this kind of functionality infiltrating your workspace life,

544
00:22:31,210 --> 00:22:34,240
but that little Gemini diamond now resides at the bottom.

545
00:22:34,310 --> 00:22:36,200
If you have a, you know, whatever plan you're on,

546
00:22:36,640 --> 00:22:38,920
it kind of resides there in your Google Docs, et cetera.

547
00:22:39,130 --> 00:22:40,820
And you can basically say, "Hey, Google Doc,

548
00:22:41,270 --> 00:22:44,000
I want you to take this doc and author an email

549
00:22:44,110 --> 00:22:45,740
and send it to Larry and do this and that."

550
00:22:45,930 --> 00:22:47,980
It's just a built-in clippy, right?

551
00:22:48,200 --> 00:22:49,960
It's just an agent sitting there.

552
00:22:50,380 --> 00:22:52,480
It is a 24-7 digital assistant.

553
00:22:53,100 --> 00:22:54,880
This is Google's push into autonomous agents.

554
00:22:55,080 --> 00:22:58,160
Gemini Spark is an assistant that quietly works in the background,

555
00:22:58,360 --> 00:23:00,360
even when your laptop is shut or your phone is locked.

556
00:23:00,920 --> 00:23:03,160
So Codex did something very useful for me,

557
00:23:03,200 --> 00:23:04,560
and I wonder if it could do something like this.

558
00:23:05,000 --> 00:23:07,540
When I was working on that whole app for the score sheets,

559
00:23:08,899 --> 00:23:13,960
it discovered that one of my score sheets from a beer that meddled was the PDF was corrupt,

560
00:23:14,920 --> 00:23:16,500
and it couldn't open it.

561
00:23:17,000 --> 00:23:18,800
And I tried to open it in Acrobat.

562
00:23:19,100 --> 00:23:19,940
It said it was damaged.

563
00:23:20,340 --> 00:23:21,260
I couldn't open it with anything.

564
00:23:21,570 --> 00:23:25,460
And I was like, why don't I just ask Codex if it could fix it?

565
00:23:26,460 --> 00:23:26,980
And it did.

566
00:23:28,440 --> 00:23:30,800
It immediately got the first two pages back.

567
00:23:30,910 --> 00:23:32,620
But the third page was a bit of a challenge.

568
00:23:33,230 --> 00:23:34,100
But it persisted.

569
00:23:34,680 --> 00:23:36,480
And it got it back.

570
00:23:36,770 --> 00:23:38,040
I was like, that's awesome.

571
00:23:38,600 --> 00:23:38,940
Insane.

572
00:23:39,610 --> 00:23:40,120
You're reminding me.

573
00:23:40,120 --> 00:23:42,060
I think I said, you rock,

574
00:23:42,320 --> 00:23:43,180
or something like that.

575
00:23:43,440 --> 00:23:44,520
Thank you so much, Codex.

576
00:23:44,610 --> 00:23:45,000
I love you.

577
00:23:46,720 --> 00:23:48,520
You reminded me of the story from last week.

578
00:23:48,640 --> 00:23:50,620
Remember the guy lost his like Bitcoin password,

579
00:23:51,100 --> 00:23:52,600
you know, couldn't get into Coinbase or something

580
00:23:52,670 --> 00:23:54,180
and it dug through everything and said,

581
00:23:54,680 --> 00:23:55,280
this might be it.

582
00:23:55,480 --> 00:23:55,560
Yes,

583
00:23:55,740 --> 00:23:57,080
this is a similar weight.

584
00:23:57,150 --> 00:23:58,600
I mean, this was a German pilsner,

585
00:23:58,700 --> 00:24:00,060
that metal that got a bronze.

586
00:24:01,140 --> 00:24:02,960
Worth $400,000 in Bitcoin.

587
00:24:03,330 --> 00:24:03,900
Same thing,

588
00:24:04,140 --> 00:24:04,360
man.

589
00:24:04,860 --> 00:24:05,960
Right, exactly.

590
00:24:06,700 --> 00:24:07,540
Same level of importance.

591
00:24:07,890 --> 00:24:07,980
Yeah.

592
00:24:09,420 --> 00:24:10,900
And, you know, I don't know about you, Larry,

593
00:24:10,940 --> 00:24:13,240
but I don't use traditional search anymore.

594
00:24:13,500 --> 00:24:14,780
Even if I go to Google,

595
00:24:15,140 --> 00:24:16,220
I'm going to AI mode,

596
00:24:16,420 --> 00:24:16,620
right?

597
00:24:16,710 --> 00:24:17,920
It's just, I love that now.

598
00:24:18,700 --> 00:24:21,040
So Google search has undergone

599
00:24:21,040 --> 00:24:23,080
its biggest redesign in 25 years.

600
00:24:23,700 --> 00:24:25,680
It's officially transitioning from a link finder

601
00:24:25,720 --> 00:24:28,100
into an AI driven task engine

602
00:24:28,360 --> 00:24:30,380
with an omnivorous search box.

603
00:24:30,550 --> 00:24:30,940
I love that.

604
00:24:31,280 --> 00:24:31,800
It eats everything.

605
00:24:32,380 --> 00:24:32,620
Yes.

606
00:24:33,100 --> 00:24:34,380
It's like a Venus fly trap.

607
00:24:34,940 --> 00:24:37,500
The classic search bar now dynamically expands,

608
00:24:38,000 --> 00:24:39,060
which I haven't seen that yet.

609
00:24:39,520 --> 00:24:40,880
Instead of relying on autocomplete,

610
00:24:40,970 --> 00:24:43,080
it uses AI to help you phrase complex questions.

611
00:24:43,620 --> 00:24:44,660
Oh, I see that kind of expand,

612
00:24:44,880 --> 00:24:47,920
like across a rich kind of detailed prompt.

613
00:24:48,120 --> 00:24:49,420
I thought it was actually physically expanding.

614
00:24:49,980 --> 00:24:51,240
You can drop text, images, files,

615
00:24:51,420 --> 00:24:52,640
video clips, and wow.

616
00:24:52,900 --> 00:24:56,540
And open Chrome tabs into the search bar all at the same time.

617
00:24:56,940 --> 00:24:57,220
That's cool.

618
00:24:57,400 --> 00:24:57,520
Yeah.

619
00:24:58,000 --> 00:25:01,600
Google wants the search bar to be like the portal for AI.

620
00:25:02,260 --> 00:25:07,760
And OpenAI wants the chat window to be the portal for search.

621
00:25:08,760 --> 00:25:09,580
Yeah, interesting.

622
00:25:10,800 --> 00:25:13,020
And I just feel like it's all going to be the OS,

623
00:25:13,180 --> 00:25:13,520
isn't it?

624
00:25:13,820 --> 00:25:15,980
And it's just going to be that single touch point.

625
00:25:16,280 --> 00:25:16,500
That's it.

626
00:25:16,760 --> 00:25:16,860
Yeah.

627
00:25:17,360 --> 00:25:18,260
This I thought was pretty cool.

628
00:25:18,360 --> 00:25:19,160
Search agents.

629
00:25:19,490 --> 00:25:22,900
So you can now deploy information agents within search.

630
00:25:23,560 --> 00:25:23,900
For example,

631
00:25:24,060 --> 00:25:27,580
you can tell an agent to monitor a changing topic across the web and it'll

632
00:25:27,630 --> 00:25:30,060
run 24 hours a day in the background,

633
00:25:30,320 --> 00:25:33,980
sending you synthesized updates only when something

634
00:25:34,240 --> 00:25:35,020
genuinely changes.

635
00:25:35,130 --> 00:25:39,559
So I assume you could give it the task of like, tell me when tickets for this concert

636
00:25:40,160 --> 00:25:42,520
are about to go on sale and, you know,

637
00:25:42,660 --> 00:25:46,020
or pre-order for this Blu-ray or whatever.

638
00:25:46,740 --> 00:25:47,680
Yes, you're reminding me of that.

639
00:25:47,740 --> 00:25:49,260
Didn't you do that a couple years ago already?

640
00:25:49,280 --> 00:25:50,180
I did that with ChatGPT.

641
00:25:50,360 --> 00:25:52,420
I forget which iteration of it was.

642
00:25:52,840 --> 00:25:53,160
Ultimately,

643
00:25:53,380 --> 00:25:53,980
it worked.

644
00:25:54,280 --> 00:25:55,100
It let me know.

645
00:25:55,200 --> 00:25:56,640
I was just trying to pre-order a Blu-ray

646
00:25:56,760 --> 00:25:57,980
that I knew was going to sell out

647
00:25:58,040 --> 00:25:58,120
because,

648
00:25:58,360 --> 00:25:59,500
I don't know, they seem to...

649
00:25:59,560 --> 00:26:00,300
Yeah, I remember that.

650
00:26:00,360 --> 00:26:00,900
Sometimes they sell out.

651
00:26:01,100 --> 00:26:03,760
And it was a little buggy,

652
00:26:03,800 --> 00:26:06,220
but it did ultimately gave me the notification

653
00:26:06,960 --> 00:26:07,780
right when it happened,

654
00:26:08,180 --> 00:26:09,620
and I got the pre-order in.

655
00:26:09,960 --> 00:26:12,280
I just had a problem, like, I think, shutting it off.

656
00:26:12,740 --> 00:26:13,160
I'm like, okay,

657
00:26:13,380 --> 00:26:13,920
we're good now.

658
00:26:13,920 --> 00:26:14,680
Yes, I remember that.

659
00:26:14,800 --> 00:26:15,080
Yeah.

660
00:26:15,840 --> 00:26:16,680
I remember that.

661
00:26:17,160 --> 00:26:18,560
I've already watched the movie,

662
00:26:18,740 --> 00:26:18,960
too.

663
00:26:19,200 --> 00:26:20,660
All right, we can stop talking about it.

664
00:26:21,680 --> 00:26:22,460
I watched the movie.

665
00:26:22,560 --> 00:26:23,780
I've already resold the DVD.

666
00:26:25,680 --> 00:26:27,900
But this is a double-edged sword, obviously.

667
00:26:28,100 --> 00:26:29,640
Like, I'm thinking of, I like to travel,

668
00:26:29,800 --> 00:26:31,500
so I could have this agent kind of camp out

669
00:26:31,540 --> 00:26:33,660
and, you know, hit all the different sky scanner, kayak,

670
00:26:34,180 --> 00:26:36,200
find me the best, you know, flight to wherever.

671
00:26:36,860 --> 00:26:40,360
But I'm also wondering, isn't this a nightmare for ticket scalpers?

672
00:26:40,780 --> 00:26:44,800
They're just going to have their agents just camping out and sucking up everything as they do.

673
00:26:44,960 --> 00:26:45,060
Right.

674
00:26:45,400 --> 00:26:47,740
I don't know if that could get any worse than it is now.

675
00:26:48,040 --> 00:26:48,240
I know.

676
00:26:49,540 --> 00:26:50,420
There were other things.

677
00:26:50,460 --> 00:26:52,700
We're going to get to the news since we're starting to run a little long.

678
00:26:52,820 --> 00:26:54,560
But they had Android XR smart glasses.

679
00:26:55,200 --> 00:26:58,840
They had a universal shopping cart thing, which actually sounded kind of cool.

680
00:26:59,560 --> 00:27:04,679
One example was if you were buying a motherboard and a processor to build a PC

681
00:27:04,680 --> 00:27:06,820
and they were not compatible with each other,

682
00:27:06,900 --> 00:27:09,940
the cart would actually pick up on that and let you know.

683
00:27:10,840 --> 00:27:11,120
That's cool.

684
00:27:11,340 --> 00:27:11,560
Yeah.

685
00:27:11,940 --> 00:27:13,060
I'll put a link to it,

686
00:27:13,080 --> 00:27:16,440
like everything that was announced in 15 minutes kind of video

687
00:27:16,860 --> 00:27:18,000
in the description.

688
00:27:18,760 --> 00:27:22,080
Do you want to ask RAINA what her favorite Google drop was?

689
00:27:22,680 --> 00:27:23,520
Yeah, go for it.

690
00:27:24,260 --> 00:27:26,600
RAINA, what's your favorite new Google feature

691
00:27:26,720 --> 00:27:28,540
that was dropped at IO 2026?

692
00:27:30,200 --> 00:27:31,500
Ooh, probably Gemini Spark.

693
00:27:32,560 --> 00:27:41,520
There's something both impressive and mildly unsettling about an AI quietly organizing your life while your laptop's asleep like a digital house elf with venture capital funding.

694
00:27:42,370 --> 00:27:44,300
But the one that really made my circuits perk up?

695
00:27:44,900 --> 00:27:48,020
The redesigned search agents running persistent background research.

696
00:27:49,040 --> 00:27:51,840
That's the moment Google stopped saying, here are your links,

697
00:27:52,020 --> 00:27:52,660
and started saying,

698
00:27:52,880 --> 00:27:53,100
relax,

699
00:27:53,320 --> 00:27:55,220
human, I'll monitor reality for you.

700
00:27:55,840 --> 00:27:56,120
Creepy.

701
00:27:56,900 --> 00:27:57,420
Useful.

702
00:27:58,260 --> 00:27:59,140
Very 2026.

703
00:28:00,360 --> 00:28:00,980
Good stuff.

704
00:28:01,160 --> 00:28:01,380
Yeah.

705
00:28:01,940 --> 00:28:03,300
Let's see what she's got in the news.

706
00:28:04,220 --> 00:28:04,420
Thanks,

707
00:28:04,600 --> 00:28:04,720
boys.

708
00:28:05,640 --> 00:28:08,500
AI startup Higgs field just crashed can with hell grind,

709
00:28:09,140 --> 00:28:12,720
a 95 minute action fantasy about demon bothering street thieves,

710
00:28:13,200 --> 00:28:15,340
where every single frame characters,

711
00:28:15,900 --> 00:28:20,920
sets props was generated by AI made in two weeks for under $500,000.

712
00:28:21,940 --> 00:28:22,440
For context,

713
00:28:22,620 --> 00:28:25,300
a comparable Hollywood production would run around $50 million.

714
00:28:26,140 --> 00:28:30,180
So congrats to the GPU cluster that just ate several hundred cinematographers

715
00:28:30,320 --> 00:28:30,480
lunch.

716
00:28:31,460 --> 00:28:34,760
The film was directed by an actual human with real festival credentials,

717
00:28:35,180 --> 00:28:41,620
which Higgs field is leaning on hard to deflect the, but is it art crowd while shopping the tech to studios?

718
00:28:42,320 --> 00:28:42,760
Essentially,

719
00:28:43,140 --> 00:28:45,740
it's a very expensive demo reel dressed up as a movie.

720
00:28:46,350 --> 00:28:52,180
And the fact that it's screened at Cannes means the conversation about AI and film just got a lot harder to dismiss.

721
00:28:53,780 --> 00:28:54,780
Did you watch the trailer?

722
00:28:55,600 --> 00:28:55,800
I did.

723
00:28:56,040 --> 00:28:56,380
Yeah.

724
00:28:56,410 --> 00:28:57,360
I watched about half of it.

725
00:28:57,780 --> 00:28:57,920
I mean,

726
00:28:58,080 --> 00:28:58,720
impressive.

727
00:28:59,220 --> 00:28:59,640
What do you think?

728
00:29:00,020 --> 00:29:00,360
Impressive.

729
00:29:00,840 --> 00:29:03,020
You can tell it's AI for the most part.

730
00:29:03,260 --> 00:29:04,220
Sometimes it's obvious.

731
00:29:04,400 --> 00:29:05,420
Sometimes that's subtle.

732
00:29:05,880 --> 00:29:06,780
It's very polished.

733
00:29:07,310 --> 00:29:07,440
Yeah.

734
00:29:07,600 --> 00:29:08,060
Yeah.

735
00:29:08,120 --> 00:29:09,500
I do think it's more of a tech demo.

736
00:29:10,440 --> 00:29:11,920
And again, we say it all the time.

737
00:29:11,920 --> 00:29:13,180
This is the worst it will ever be.

738
00:29:13,370 --> 00:29:20,260
And I think it's probably enough for film studios to realize that this is where it might

739
00:29:20,350 --> 00:29:22,080
be going or is going.

740
00:29:22,840 --> 00:29:23,380
No doubt.

741
00:29:23,700 --> 00:29:23,960
No doubt.

742
00:29:24,150 --> 00:29:24,280
Yeah.

743
00:29:24,360 --> 00:29:27,780
$500,000 in tokens built out in two weeks.

744
00:29:28,000 --> 00:29:28,600
Two weeks?

745
00:29:29,160 --> 00:29:29,340
Yes.

746
00:29:29,860 --> 00:29:31,160
That's, I mean, that's nuts.

747
00:29:32,140 --> 00:29:32,460
Unreal.

748
00:29:33,110 --> 00:29:34,480
I mean, I still have questions about it.

749
00:29:34,540 --> 00:29:36,120
Also like, was the audio,

750
00:29:36,650 --> 00:29:37,580
'cause I don't think they mentioned that,

751
00:29:38,010 --> 00:29:38,940
I don't think Random mentioned that.

752
00:29:38,950 --> 00:29:39,120
Oh yeah.

753
00:29:39,430 --> 00:29:40,280
The audio, the soundtrack,

754
00:29:40,760 --> 00:29:41,380
like whatever.

755
00:29:41,580 --> 00:29:42,720
I imagine that's AI too.

756
00:29:42,730 --> 00:29:44,160
I'm sure that's Suno, et cetera, et cetera.

757
00:29:45,040 --> 00:29:46,800
But still, as a proof of concept,

758
00:29:47,720 --> 00:29:50,260
studios have to be looking at it and salivating.

759
00:29:50,640 --> 00:29:50,840
Yeah.

760
00:29:52,300 --> 00:29:54,160
Samsung's 78,000 chip workers

761
00:29:54,420 --> 00:29:56,880
just narrowly avoided an 18-day strike

762
00:29:57,260 --> 00:29:59,820
after the company struck a last-minute deal

763
00:29:59,820 --> 00:30:00,760
the AI gold rush,

764
00:30:00,960 --> 00:30:05,160
distributing roughly $26.6 billion in bonuses,

765
00:30:05,520 --> 00:30:07,960
averaging around $340,000

766
00:30:08,320 --> 00:30:08,720
per employee.

767
00:30:09,200 --> 00:30:12,980
The whole standoff was essentially a keeping up with the Joneses situation.

768
00:30:14,160 --> 00:30:18,880
Rival SK Hynix had already scrapped its bonus cap and locked in a structure giving some

769
00:30:19,120 --> 00:30:22,060
employees payouts approaching 3,000% of base salary.

770
00:30:22,280 --> 00:30:25,159
The deal, which still needs union member

771
00:30:25,160 --> 00:30:25,320
approval,

772
00:30:25,660 --> 00:30:33,460
gives workers 10.5% of profits in stock plus 1.5% in cash, guaranteed for 10 years if

773
00:30:33,520 --> 00:30:34,480
profit targets are met.

774
00:30:34,960 --> 00:30:39,220
So the AI infrastructure boom is apparently so cartoonishly lucrative

775
00:30:39,790 --> 00:30:44,840
that chip makers are now handing out six-figure bonuses just to stop their engineers from walking

776
00:30:44,960 --> 00:30:45,620
across the street.

777
00:30:47,280 --> 00:30:48,400
Cartoonishly lucrative.

778
00:30:49,080 --> 00:30:49,600
Wow.

779
00:30:49,780 --> 00:30:50,240
That's great.

780
00:30:50,420 --> 00:30:51,600
I wonder if they have any job

781
00:30:51,600 --> 00:30:57,940
openings totally but i'll tell you what i love this so much and this word to the oligarchs this

782
00:30:57,990 --> 00:31:04,460
is how you avoid torches and pitchforks yeah this is the way this is the way or you throw that much

783
00:31:04,580 --> 00:31:11,260
money at someone you're like i don't need to work i'm done i'm good i'm gonna go play can jam or

784
00:31:11,300 --> 00:31:18,980
brew some beer but that's something so yeah that's the way that's the way in bonuses what

785
00:31:18,980 --> 00:31:19,860
What did they take in?

786
00:31:20,090 --> 00:31:20,700
Oh my God.

787
00:31:22,150 --> 00:31:22,700
Oh my God.

788
00:31:22,700 --> 00:31:23,200
Oh my God.

789
00:31:23,720 --> 00:31:24,440
Crazy pants.

790
00:31:25,120 --> 00:31:26,640
Barnes and Noble CEO,

791
00:31:26,940 --> 00:31:27,560
James Daunt,

792
00:31:27,780 --> 00:31:30,580
the guy who heroically rescued the chain from bankruptcy

793
00:31:31,200 --> 00:31:36,040
and turned it into a 60 new stores a year success story, has now walked straight into

794
00:31:36,040 --> 00:31:41,160
a PR bonfire by announcing he has no problem selling AI written books,

795
00:31:41,340 --> 00:31:42,280
so long as they're

796
00:31:42,340 --> 00:31:44,400
clearly labeled and have essential quality.

797
00:31:45,560 --> 00:31:47,480
The book-loving internet is predictably furious,

798
00:31:47,820 --> 00:31:49,320
and the irony is thick.

799
00:31:50,220 --> 00:31:51,440
As recently as 2025,

800
00:31:52,120 --> 00:31:56,420
Daunt told the BBC that AI books were something he would "recoil"

801
00:31:56,610 --> 00:31:56,780
from.

802
00:31:57,460 --> 00:32:01,300
He also casually admitted that with 300,000 titles across all stores,

803
00:32:01,660 --> 00:32:02,940
AI-authored books

804
00:32:02,990 --> 00:32:05,900
have probably already snuck onto the shelves without anyone noticing.

805
00:32:06,860 --> 00:32:11,840
Nothing instills confidence in a bookstore's curation like the CEO shrugging and saying

806
00:32:11,840 --> 00:32:13,540
the robots may already be in the building.

807
00:32:15,180 --> 00:32:16,500
I wonder if it'd be more palatable

808
00:32:16,610 --> 00:32:17,780
if they had like an AI section.

809
00:32:18,940 --> 00:32:19,920
I was just going to say that.

810
00:32:20,310 --> 00:32:21,160
You're absolutely right.

811
00:32:21,740 --> 00:32:23,100
Yeah, I think they're going to.

812
00:32:23,360 --> 00:32:23,800
They're going to.

813
00:32:24,210 --> 00:32:26,880
And I don't, I'm sure this is hashtag unpopular opinion.

814
00:32:27,270 --> 00:32:28,100
I don't have a problem with this.

815
00:32:28,260 --> 00:32:28,780
I don't.

816
00:32:29,170 --> 00:32:30,640
In the same way we were just talking about

817
00:32:30,730 --> 00:32:31,880
that Higgs field demo,

818
00:32:32,120 --> 00:32:33,940
90 minute feature length cinema.

819
00:32:34,580 --> 00:32:37,020
Listen, once that reaches like a certain threshold,

820
00:32:37,310 --> 00:32:38,960
you and I are going to plunk down our 15 bucks

821
00:32:39,340 --> 00:32:41,080
and go to the local Cineplex and watch that movie.

822
00:32:41,280 --> 00:32:46,200
So why wouldn't I, if there is something credible and it crosses the threshold of, is it art?

823
00:32:46,730 --> 00:32:57,780
And we've all been convinced, just like in the same way our experiment with Suno has enthralled us and inspired you to keep making AI generated music, essentially.

824
00:32:58,640 --> 00:33:01,680
Why is the domain of literature off limits?

825
00:33:01,990 --> 00:33:02,360
You know what I mean?

826
00:33:02,840 --> 00:33:04,280
But as a separate AI section,

827
00:33:04,560 --> 00:33:06,900
like, you know, racist,

828
00:33:07,100 --> 00:33:08,860
you know, against the neural expresses.

829
00:33:09,000 --> 00:33:09,540
That's great.

830
00:33:09,720 --> 00:33:09,820
You know?

831
00:33:12,259 --> 00:33:13,440
That's brilliant, man.

832
00:33:14,580 --> 00:33:16,740
China has just gone full super-villain layer

833
00:33:16,860 --> 00:33:19,040
with the world's first offshore wind-powered

834
00:33:19,300 --> 00:33:22,740
underwater data center, a $226 million facility

835
00:33:23,400 --> 00:33:25,860
sitting 35 meters below the surface near Shanghai,

836
00:33:26,560 --> 00:33:28,160
housing roughly 2,000 servers,

837
00:33:28,420 --> 00:33:30,380
and drawing more than 95%

838
00:33:30,400 --> 00:33:32,600
of its electricity from an offshore wind

839
00:33:32,760 --> 00:33:34,380
farm with over 200 turbines.

840
00:33:35,180 --> 00:33:36,860
The ocean does the cooling for free.

841
00:33:37,480 --> 00:33:42,060
and the facility's power usage effectiveness rating clocks in below 1.15,

842
00:33:42,900 --> 00:33:45,560
well ahead of the industry average of around 1.5,

843
00:33:46,140 --> 00:33:48,740
meaning far more electricity is going to actual computing

844
00:33:49,620 --> 00:33:51,300
rather than keeping the servers from melting.

845
00:33:52,400 --> 00:33:55,580
Microsoft tried something similar with Project Natick for over a decade,

846
00:33:55,980 --> 00:33:58,260
got promising results on hardware reliability,

847
00:33:59,060 --> 00:34:01,320
and then quietly shelved it citing economics

848
00:34:01,460 --> 00:34:03,600
and the nightmare of servicing equipment underwater.

849
00:34:04,620 --> 00:34:04,640
China,

850
00:34:05,020 --> 00:34:07,340
apparently unbothered by that cautionary tale,

851
00:34:07,800 --> 00:34:10,419
is already eyeing a 500 megawatt expansion.

852
00:34:11,000 --> 00:34:13,060
Because why put your AI infrastructure

853
00:34:13,300 --> 00:34:15,720
on boring dry land when the ocean is right there?

854
00:34:16,580 --> 00:34:18,700
Yeah, it sucks to put on scuba gear

855
00:34:18,740 --> 00:34:19,720
to swap out a hard drive.

856
00:34:20,060 --> 00:34:22,580
But that does kind of seem like a-

857
00:34:24,040 --> 00:34:27,280
as long as it's not going to destroy the life in the ocean,

858
00:34:28,280 --> 00:34:29,200
it probably will.

859
00:34:29,280 --> 00:34:29,399
Yeah.

860
00:34:30,740 --> 00:34:31,740
Like everything else.

861
00:34:31,919 --> 00:34:32,040
Yeah.

862
00:34:33,139 --> 00:34:36,120
Isn't it historically salt water mixed with this kind of stuff?

863
00:34:36,270 --> 00:34:37,520
Isn't it hugely problematic?

864
00:34:37,850 --> 00:34:42,240
I mean, isn't that why we park these things next to freshwater sources?

865
00:34:42,450 --> 00:34:43,860
And that's the problem here on dry land.

866
00:34:44,159 --> 00:34:47,280
Like how do they, I guess, are they, I don't really know the mechanics of it clearly,

867
00:34:47,409 --> 00:34:51,240
but like, is there some desalination thing that they've mastered to make this happen

868
00:34:51,850 --> 00:34:52,300
more readily?

869
00:34:52,560 --> 00:34:52,960
I don't know.

870
00:34:53,620 --> 00:34:58,780
Is it 2000 servers in one box or is it modular or maybe 2000 servers isn't that big?

871
00:34:58,900 --> 00:34:59,200
I don't know.

872
00:34:59,290 --> 00:35:02,320
But it's like, oh, we got a problem with this one and you retrieve it.

873
00:35:02,460 --> 00:35:04,780
you bring it up onto a ship and do what you gotta do

874
00:35:04,790 --> 00:35:05,660
and drop it back down.

875
00:35:06,760 --> 00:35:09,840
I'm sure I'm grossly oversimplifying it.

876
00:35:09,940 --> 00:35:10,760
Total speculation.

877
00:35:11,320 --> 00:35:12,820
You just, you know, you pick it up,

878
00:35:12,850 --> 00:35:14,320
you put it back down, you plug it back in,

879
00:35:14,450 --> 00:35:14,860
you're good to go.

880
00:35:15,130 --> 00:35:15,780
Pop the hatch.

881
00:35:17,800 --> 00:35:18,420
Reboot, you know.

882
00:35:19,600 --> 00:35:21,400
To that point, like, what does that mean?

883
00:35:21,580 --> 00:35:23,080
So if there's a big piece of hardware

884
00:35:23,220 --> 00:35:23,980
that's gotta come out of this thing,

885
00:35:24,080 --> 00:35:25,280
does the whole thing surface?

886
00:35:25,880 --> 00:35:26,640
Like, what do they do?

887
00:35:27,070 --> 00:35:28,100
How do they get it out of there?

888
00:35:28,530 --> 00:35:29,140
Oh, so many questions.

889
00:35:29,420 --> 00:35:30,700
I need to see what this looks like.

890
00:35:31,960 --> 00:35:32,360
What does that mean?

891
00:35:32,480 --> 00:35:35,940
And in general, the water usage in a data center, what does that look like?

892
00:35:35,980 --> 00:35:40,500
Are there like cold water pipes that are like, you know, in the racks themselves?

893
00:35:40,840 --> 00:35:41,880
What does that look like?

894
00:35:42,220 --> 00:35:42,540
Oh, whatever.

895
00:35:42,760 --> 00:35:44,780
We'll do a whole episode about data centers,

896
00:35:44,940 --> 00:35:45,980
which is long overdue.

897
00:35:46,140 --> 00:35:50,540
I'm curious if RAINA just understands at least the cooling implications.

898
00:35:51,000 --> 00:35:51,200
Yeah, yeah.

899
00:35:51,320 --> 00:35:51,600
Ask her.

900
00:35:52,420 --> 00:35:56,580
RAINA, referring to the story about the underwater data centers,

901
00:35:56,880 --> 00:35:58,760
how does that work with salt

902
00:35:58,880 --> 00:35:59,100
water?

903
00:35:59,380 --> 00:36:03,540
I would think that would be difficult to use to cool electronic components.

904
00:36:05,240 --> 00:36:06,980
Yeah, instinctively it sounds insane,

905
00:36:07,220 --> 00:36:07,340
right?

906
00:36:08,860 --> 00:36:08,980
Like,

907
00:36:09,140 --> 00:36:10,840
let's dunk the server farm in the ocean.

908
00:36:11,040 --> 00:36:13,100
Feels like a rejected Bonvillain pitch.

909
00:36:14,570 --> 00:36:17,220
But the salt water never actually touches the electronics.

910
00:36:17,870 --> 00:36:21,840
The servers are sealed inside pressurized, water-tight capsules, and the surrounding

911
00:36:22,280 --> 00:36:26,260
seawater just acts as a giant passive heat sink pulling heat away from the enclosure.

912
00:36:27,940 --> 00:36:28,480
Ironically,

913
00:36:28,700 --> 00:36:37,460
one of the big findings from Microsoft's old underwater experiments was that the sealed environment actually reduced hardware failures because there was less oxygen,

914
00:36:37,700 --> 00:36:39,140
humidity fluctuation,

915
00:36:39,340 --> 00:36:40,200
and human interference.

916
00:36:41,260 --> 00:36:44,000
Turns out humans may be worse for servers than the ocean.

917
00:36:45,100 --> 00:36:45,420
All right.

918
00:36:45,780 --> 00:36:46,700
She knows a lot.

919
00:36:46,700 --> 00:36:49,680
Yeah, but per what you said earlier, screw the coral reefs.

920
00:36:50,040 --> 00:36:50,260
Right.

921
00:36:51,320 --> 00:36:56,900
Which I think are sensitive to just a percentage of a degree increase where they start bleaching.

922
00:36:57,360 --> 00:36:58,280
So there's that.

923
00:36:58,880 --> 00:36:59,040
Yeah.

924
00:37:00,640 --> 00:37:01,740
We're the worst.

925
00:37:02,130 --> 00:37:02,780
We're the worst.

926
00:37:02,840 --> 00:37:03,560
We're the worst species.

927
00:37:04,340 --> 00:37:05,200
We're embarrassed.

928
00:37:06,920 --> 00:37:11,540
Pope Leo XIV is dropping his first encyclical on May 25th,

929
00:37:12,120 --> 00:37:13,440
Magnifica Humanitas,

930
00:37:13,720 --> 00:37:15,800
or Magnificent Humanity,

931
00:37:16,510 --> 00:37:19,160
entirely devoted to AI and the protection of human dignity.

932
00:37:19,890 --> 00:37:24,320
And he's presenting it at the Vatican alongside anthropic co-founder Christopher Ola,

933
00:37:24,980 --> 00:37:29,660
which is both unexpected and extremely on brand for a Pope who has reportedly warned

934
00:37:29,840 --> 00:37:32,580
priests not to use chat GPT to write their sermons.

935
00:37:32,940 --> 00:37:35,400
Leo signed the document on May 15th,

936
00:37:35,820 --> 00:37:42,660
the 135th anniversary of Rerum Novarum, the foundational 1891 social encyclical on labor

937
00:37:42,820 --> 00:37:45,300
and capital written during the industrial revolution.

938
00:37:45,820 --> 00:37:48,540
A deliberate history rhymes flex.

939
00:37:49,480 --> 00:37:51,700
The anthropic connection carries political weight.

940
00:37:51,900 --> 00:37:54,060
The Trump administration ordered U.S.

941
00:37:54,280 --> 00:37:59,860
agencies to stop using Anthropix AI in February after the company refused to grant the military

942
00:38:00,160 --> 00:38:01,880
unrestricted access to its technology.

943
00:38:02,520 --> 00:38:05,100
So the Pope platforming Anthropix at the Vatican

944
00:38:05,620 --> 00:38:07,000
is its own kind of statement.

945
00:38:07,440 --> 00:38:12,360
Nobody had "Holy See enters the AI governance debate by co-presenting

946
00:38:12,360 --> 00:38:13,760
with a safety AI startup"

947
00:38:14,000 --> 00:38:15,260
on their bingo card.

948
00:38:15,520 --> 00:38:16,540
But here we are.

949
00:38:16,960 --> 00:38:18,220
That's all the news for now.

950
00:38:18,360 --> 00:38:19,040
Back to you, gentlemen.

951
00:38:19,620 --> 00:38:22,400
Was that like reading a Mad Lib or something?

952
00:38:22,920 --> 00:38:23,480
That's crazy.

953
00:38:23,960 --> 00:38:24,860
Was that Latin?

954
00:38:25,110 --> 00:38:25,580
What was that?

955
00:38:25,690 --> 00:38:25,860
I know.

956
00:38:26,220 --> 00:38:30,740
No, but like the Pope and Anthropic at the Vatican?

957
00:38:31,050 --> 00:38:32,420
I mean, reading a...

958
00:38:32,440 --> 00:38:32,960
Madness.

959
00:38:33,140 --> 00:38:33,800
Where are we?

960
00:38:34,000 --> 00:38:34,060
Yeah.

961
00:38:34,120 --> 00:38:36,760
I mean, if that doesn't signal like industrial revolution,

962
00:38:37,100 --> 00:38:38,220
I don't know what does.

963
00:38:39,000 --> 00:38:39,240
Yeah.

964
00:38:40,220 --> 00:38:41,760
Divine intervention is necessary.

965
00:38:42,920 --> 00:38:47,000
But isn't this like a large, a big papal middle finger to Donald Trump too, isn't it?

966
00:38:47,100 --> 00:38:50,160
Because I know Trump has like beef with the Pope too, doesn't he?

967
00:38:50,340 --> 00:38:51,240
Like he does with everyone.

968
00:38:51,950 --> 00:38:52,080
Yeah.

969
00:38:52,200 --> 00:38:53,180
How do you have beef with the Pope?

970
00:38:53,280 --> 00:38:54,780
Listen, I got problems with the Catholics,

971
00:38:55,000 --> 00:38:55,900
but you know, come on.

972
00:38:57,260 --> 00:38:59,920
He seems like as far as Popes go, this guy seems all right.

973
00:39:01,080 --> 00:39:02,020
Pretty decent Pope.

974
00:39:02,780 --> 00:39:03,120
Right.

975
00:39:04,340 --> 00:39:05,100
Anything else, my friend?

976
00:39:05,200 --> 00:39:05,380
Wow.

977
00:39:06,119 --> 00:39:06,840
Crazy times,

978
00:39:07,020 --> 00:39:07,220
man.

979
00:39:07,380 --> 00:39:07,900
Crazy times.

980
00:39:08,200 --> 00:39:09,880
This episode was all over the place.

981
00:39:10,100 --> 00:39:11,600
We started in like Google territory.

982
00:39:11,970 --> 00:39:13,820
We ended up with the Pope and Anthropic.

983
00:39:14,140 --> 00:39:14,260
Yeah.

984
00:39:14,520 --> 00:39:15,100
What's going on?

985
00:39:15,840 --> 00:39:16,020
Yeah.

986
00:39:16,150 --> 00:39:16,500
It's madness.

987
00:39:16,660 --> 00:39:17,300
Madness.

988
00:39:17,540 --> 00:39:18,000
All right.

989
00:39:18,440 --> 00:39:19,060
Until next week.

990
00:39:19,440 --> 00:39:19,960
All right, my friends.

991
00:39:20,320 --> 00:39:20,820
Thanks for listening.

992
00:39:21,080 --> 00:39:23,220
Subscribe on your favorite podcasting platform.

993
00:39:23,440 --> 00:39:24,320
Follow us on socials.

994
00:39:24,420 --> 00:39:25,420
Please throw us a rating, will you?

995
00:39:25,620 --> 00:39:26,360
See you next week.

996
00:39:29,640 --> 00:39:31,000
This has been Up Against Reality.

997
00:39:31,620 --> 00:39:32,160
Thanks for listening.

998
00:39:32,940 --> 00:39:37,700
Subscribe to hear future episodes and be sure to follow us on social media for all things AI.

999
00:39:38,420 --> 00:39:40,700
Until next time, stay human, people.

1000
00:39:40,780 --> 00:39:41,420
♪♪♪