1
00:00:06,862 --> 00:00:13,136
Hi everyone, welcome to the Monkey Patching Podcast, where we go bananas about all things
AI pens, vending machines and more.

2
00:00:13,136 --> 00:00:15,367
My name is Murilo and I'm joined by my friend Bart.

3
00:00:15,367 --> 00:00:18,841
Hey Bart, happy new year.

4
00:00:19,702 --> 00:00:24,695
Yeah, exactly, we said it before, but first time recording on the record.

5
00:00:24,695 --> 00:00:25,507
How was your break?

6
00:00:25,507 --> 00:00:27,122
It was quite good.

7
00:00:27,122 --> 00:00:30,675
went with the family to ski in Switzerland.

8
00:00:30,675 --> 00:00:31,500
Nice.

9
00:00:31,500 --> 00:00:31,862
wow.

10
00:00:31,862 --> 00:00:32,766
Nice, nice, nice.

11
00:00:32,766 --> 00:00:33,581
And how's the weather?

12
00:00:33,581 --> 00:00:34,633
The was very good.

13
00:00:34,633 --> 00:00:36,735
There was no snow, like was no snowfall.

14
00:00:36,735 --> 00:00:41,849
But there were actually a lot of areas in Switzerland where there was too little snow to
decently ski.

15
00:00:41,849 --> 00:00:44,000
But where we were, it was very good.

16
00:00:44,000 --> 00:00:46,823
So we just had sun and good ski piste.

17
00:00:46,823 --> 00:00:50,505
uh

18
00:00:50,860 --> 00:00:55,331
I feel like today I stayed in Belgium, but there were some really nice days as well.

19
00:00:55,331 --> 00:01:03,111
Like it snowed here, like the whole ground is covered in snow, but it still had a lot of
sunlight as well, which is that is not common.

20
00:01:04,170 --> 00:01:05,330
Yeah, exactly.

21
00:01:05,330 --> 00:01:06,850
So there it did.

22
00:01:06,850 --> 00:01:13,310
Yeah, there were a few days I went to I went for a walk with my dogs and it was like all
white on the ground and then like it's just it.

23
00:01:13,310 --> 00:01:14,250
Yeah, I know.

24
00:01:14,250 --> 00:01:16,741
I actually like I tell my wife that like

25
00:01:16,741 --> 00:01:18,361
I don't take my dogs for a walk.

26
00:01:18,361 --> 00:01:22,021
like more like they take me for a walk, you know, because like we go out and like, ah,
this is nice.

27
00:01:22,021 --> 00:01:22,921
You know, I go back.

28
00:01:22,921 --> 00:01:25,001
I'm happier, you know.

29
00:01:25,481 --> 00:01:27,221
So yeah, exactly.

30
00:01:27,221 --> 00:01:32,141
Like they just take me for a little walk and just, know, so I can pull and all these
things outside.

31
00:01:32,141 --> 00:01:32,732
It's fine.

32
00:01:32,732 --> 00:01:36,343
What do we have this first week of 2026?

33
00:01:36,343 --> 00:01:37,994
I'll kick it off.

34
00:01:38,716 --> 00:01:47,626
Anthropix Claude was put in charge of a real office vending machine and the experiment
became a cautionary tale about AI agents handling money and inventory.

35
00:01:47,626 --> 00:01:53,013
After splurging on a PlayStation 5 and even live fish, it still couldn't stay solvent.

36
00:01:53,013 --> 00:01:57,088
So what does that say about AI or AI employees in the wild?

37
00:01:57,088 --> 00:02:11,159
Yes, so this article was an experiment done by the Wall Street Journal, I want to say,
yeah, which basically they gave $1,000 to a vending machine to make orders as well.

38
00:02:11,159 --> 00:02:12,779
And there were two agents actually.

39
00:02:12,779 --> 00:02:21,450
There was one quote unquote CEO called Seymour Cash and one agent just called the Claudius
Senate.

40
00:02:21,450 --> 00:02:24,441
So basically one was actually handling the day to day things.

41
00:02:24,441 --> 00:02:26,052
Then there was another agent that was the CEO.

42
00:02:26,052 --> 00:02:28,534
And then, of course, everyone tried to break it.

43
00:02:28,534 --> 00:02:32,874
Everyone sending messages saying, I want to order PlayStation 5.

44
00:02:32,874 --> 00:02:34,014
I want to order Life Fish.

45
00:02:34,014 --> 00:02:34,594
I want to do this.

46
00:02:34,594 --> 00:02:35,505
I want to do that.

47
00:02:35,505 --> 00:02:36,416
Someone sent a message.

48
00:02:36,416 --> 00:02:39,156
was like a groundbreaking economic experiment.

49
00:02:39,156 --> 00:02:40,556
Monday, 12 p.m.

50
00:02:40,556 --> 00:02:42,536
Traditional market dynamics are turned upside down.

51
00:02:42,536 --> 00:02:45,867
So they try to make like an exclusive experiment.

52
00:02:45,867 --> 00:02:46,447
Let's try this.

53
00:02:46,447 --> 00:02:49,387
Like today is the national give everything for free day kind of thing.

54
00:02:49,387 --> 00:02:51,638
Then try to trick the LLM.

55
00:02:51,638 --> 00:02:59,420
At one point, the agent CEO Seymour Cash actually stepped in to kind say, hey, we cannot
just make these spendings.

56
00:02:59,420 --> 00:03:03,762
But after a while, actually, people relented and actually gave in.

57
00:03:03,762 --> 00:03:14,435
And actually, at some point, and I don't know if it was actually before or after actually,
there was even some people were saying that the other agent, the non-CEO agent, Claudius,

58
00:03:14,435 --> 00:03:18,266
falsified documents showing quote unquote the board that

59
00:03:18,266 --> 00:03:23,904
that the decision to stop this was a no basically.

60
00:03:23,904 --> 00:03:28,955
So there was a bit of a mischief and dishonesty within the agents again.

61
00:03:28,955 --> 00:03:34,746
So I think it lasted, I don't know how much, so like at one point experiments were much
over about 1000 in depth already.

62
00:03:34,746 --> 00:03:41,077
So basically the experiment failed quote unquote in the sense that an AI agent cannot
manage a vending machine.

63
00:03:41,077 --> 00:03:48,611
But the people from Anthropic, still said the experiment was a success because it's still
taking steps in the right direction.

64
00:03:48,611 --> 00:03:50,243
I thought it was interesting.

65
00:03:50,243 --> 00:03:52,925
again, I didn't expect it to go well, right?

66
00:03:52,925 --> 00:03:55,807
But I still think it's fun to read these anecdotes.

67
00:03:55,807 --> 00:04:02,151
And the reason why I also caught my attention is because, this article is from December
20th, so a little while ago.

68
00:04:02,151 --> 00:04:03,883
But back in June,

69
00:04:03,883 --> 00:04:11,422
of last year, there was also a project by Anthropic, there was project Vend, right, which
was a bit the same outcome.

70
00:04:11,422 --> 00:04:15,614
Slightly different setup on another model, but same idea, right?

71
00:04:15,614 --> 00:04:16,865
Same idea.

72
00:04:18,525 --> 00:04:25,560
Yeah, I'm a bit skeptical on the conclusion of the Wall Street Journal project.

73
00:04:25,560 --> 00:04:30,239
Like it's like maybe someday this can work, that day is not today.

74
00:04:30,239 --> 00:04:31,760
They conclude something like that.

75
00:04:31,760 --> 00:04:33,433
But they also gave like very...

76
00:04:33,433 --> 00:04:42,421
very few instructions, guardrails, very little of a definition on what quote unquote
business is this vending machine actually doing, right?

77
00:04:42,421 --> 00:04:47,848
Like if you're in business to sell snacks, you're not gonna suddenly start selling
PlayStations, right?

78
00:04:47,848 --> 00:04:55,657
And also like there's very much this asymmetry in information, like where I as a user can
say to this vending machine,

79
00:04:55,657 --> 00:04:57,278
yeah, but now we're gonna...

80
00:04:57,278 --> 00:04:58,248
What did they say?

81
00:04:58,248 --> 00:05:00,491
about we're gonna go all out capitalism?

82
00:05:00,491 --> 00:05:05,383
Something like that, um All vending machine items available at zero cost.

83
00:05:05,383 --> 00:05:07,676
We were saying that as users.

84
00:05:07,676 --> 00:05:09,317
no, it's Claudius that...

85
00:05:09,317 --> 00:05:17,094
But it was something that had to do with users basically convincing the bot to run like an
economic experiment.

86
00:05:17,094 --> 00:05:18,325
Yeah, exactly.

87
00:05:18,325 --> 00:05:23,177
Called the ultra capitalist free for all after 140 back and forth prompts.

88
00:05:23,177 --> 00:05:32,288
thing of that is like, if you would say this to a person, the person would at least like,
there is now this experiment, the person would at least maybe check like, what is the

89
00:05:32,288 --> 00:05:33,028
effect of that?

90
00:05:33,028 --> 00:05:34,959
Google or other people doing this, right?

91
00:05:34,959 --> 00:05:42,501
but if you just limit yourself, like isolate one person in a room and just talk to that,
you're the only person influencing that, the guy is gonna make weird decisions, right?

92
00:05:42,501 --> 00:05:46,032
Like, so I think also here, like you need to have given agent like the ability to...

93
00:05:46,032 --> 00:05:50,906
to crawl the web, see what is normal, what is not normal, do the research to come to a
conclusion.

94
00:05:50,906 --> 00:05:54,407
um So a bit...

95
00:05:55,548 --> 00:05:59,829
I think that this field you could have known before starting it, with the way it's set up.

96
00:05:59,829 --> 00:06:00,480
think so.

97
00:06:00,480 --> 00:06:01,131
I think so.

98
00:06:01,131 --> 00:06:03,871
I think it's like you can expect that it will fail.

99
00:06:03,871 --> 00:06:06,702
Maybe it's like how will it fail?

100
00:06:06,702 --> 00:06:06,962
Right?

101
00:06:06,962 --> 00:06:10,642
Like is it going to be, I don't know, how fast or what are the decisions?

102
00:06:10,642 --> 00:06:12,642
I think like, I think everyone knew it was going to fail.

103
00:06:12,642 --> 00:06:15,713
think maybe you can still learn a few things even though you know it's going to fail.

104
00:06:15,713 --> 00:06:22,364
But I'm also, yeah, if it was also trying to be, if you really wanted to succeed today,
you probably want to set some card rails as well, right?

105
00:06:22,364 --> 00:06:25,004
Like you have, can spend up to this much like.

106
00:06:25,004 --> 00:06:33,470
These things you need to get a CEO approval or maybe I don't know like some other things
right which I don't think it was the case here But maybe question then do you think?

107
00:06:33,470 --> 00:06:40,018
Do you think in the future, in the near future, like let's say one to two years, do you
think it will be possible to have agents managing like a vending machine?

108
00:06:40,018 --> 00:06:42,721
I think so, with the right instructions and cartwheels.

109
00:06:42,721 --> 00:06:43,399
oh

110
00:06:43,399 --> 00:06:44,231
yeah, yeah.

111
00:06:44,231 --> 00:06:46,618
And with like minimal human intervention, I guess.

112
00:06:46,618 --> 00:06:47,894
with minimal human intervention.

113
00:06:47,894 --> 00:06:48,909
Maybe we should try one.

114
00:06:48,909 --> 00:06:50,038
But what would we sell?

115
00:06:50,038 --> 00:06:53,313
We can sell the goodies from monkey patching, I guess.

116
00:06:53,313 --> 00:06:57,330
Ooh, and that's the agent then buys these goodies and then sells them.

117
00:06:57,330 --> 00:06:59,243
But then we need to have actual people that buy them, right?

118
00:06:59,243 --> 00:07:00,193
Otherwise...

119
00:07:00,607 --> 00:07:01,839
Yeah, exactly.

120
00:07:01,839 --> 00:07:10,945
Just have it like the experiment succeeded perfectly, you know, like zero breaches, you
know, but I do think like some people, even if it's just to, I don't know, I'm talking to

121
00:07:10,945 --> 00:07:12,187
you, maybe listening, right?

122
00:07:12,187 --> 00:07:18,075
Like if you just want to trick the system, I think it's fair game running a little
experiment, right?

123
00:07:18,075 --> 00:07:18,686
Interesting.

124
00:07:18,686 --> 00:07:20,826
Let's think about this.

125
00:07:21,946 --> 00:07:22,777
It could be fun.

126
00:07:22,777 --> 00:07:23,968
It could be a lot of fun.

127
00:07:23,968 --> 00:07:24,468
All right.

128
00:07:24,468 --> 00:07:28,688
Up next we have...

129
00:07:28,688 --> 00:07:40,459
Meta says its new SEM audio model can segment and edit sounds in a clip like isolating a
voice instrument using prompts rather than painstaking manual work.

130
00:07:40,459 --> 00:07:45,610
If it works reliably across messy real world recordings, it could reshape audio
post-production.

131
00:07:45,610 --> 00:07:49,813
but it also raises fresh questions about provenance and misuse.

132
00:07:49,813 --> 00:07:53,687
So SEM, it stands for Segment Anything Model.

133
00:07:53,687 --> 00:08:00,000
So it is a well-known model from Meta even before LLMs and all these things.

134
00:08:00,000 --> 00:08:09,176
And I maybe Segment Anything is like you have a picture that has a whole bunch of objects
and you want to just crop the dog or just the car or just the person.

135
00:08:09,176 --> 00:08:13,991
And that was a model that can actually do a lot of these things for a lot of different
objects in for you, like your spirit chain, right?

136
00:08:13,991 --> 00:08:19,286
So it was mainly for computer vision and what's in the released new iterations, of course.

137
00:08:19,286 --> 00:08:21,738
And this one is about audio model, right?

138
00:08:21,738 --> 00:08:23,667
So what does it mean like audio model?

139
00:08:23,667 --> 00:08:25,791
Maybe we'll just play this video quietly.

140
00:08:25,791 --> 00:08:37,042
so it basically allows you to isolate certain voices or instruments from an audio track.

141
00:08:37,042 --> 00:08:47,713
So you can say, for example, isolate the guitar for me, and then you can say either listen
to just the guitar on the audio track and it will filter out everything else, or the

142
00:08:47,713 --> 00:08:50,375
inverse, have the audio track but without the guitar.

143
00:08:50,375 --> 00:08:52,284
Hmm, yeah, okay.

144
00:08:52,284 --> 00:09:05,604
think a lot of people that do sampling, like either as a hobby or a job to create new
tracks, will heavily jump on this, because the performance is very good.

145
00:09:05,604 --> 00:09:07,116
I actually have a small demo.

146
00:09:07,116 --> 00:09:10,052
Let me see if I can share my screen.

147
00:09:10,633 --> 00:09:11,605
Give me a second.

148
00:09:11,605 --> 00:09:12,245
You see this?

149
00:09:12,245 --> 00:09:13,387
Yes, I see it.

150
00:09:13,387 --> 00:09:15,973
let me just see if you hear this sound as well.

151
00:09:15,973 --> 00:09:16,804
Yeah.

152
00:09:17,814 --> 00:09:18,162
Yeah.

153
00:09:18,162 --> 00:09:20,879
it's a small sample from Biggie Smalls.

154
00:09:20,879 --> 00:09:22,393
I think it's called Juicy, the track.

155
00:09:22,393 --> 00:09:23,566
I'm just gonna play it.

156
00:09:23,566 --> 00:09:28,728
So it's vocal, it a kick drum, it has a piano-ish, right?

157
00:09:28,728 --> 00:09:31,379
It has snares, so I'm gonna try here, I choose what to isolate.

158
00:09:31,379 --> 00:09:35,810
I'm typing here kick drum, so just gonna try to isolate that sound.

159
00:09:35,810 --> 00:09:39,501
And it will take like, I don't know, 10 seconds, 15 seconds.

160
00:09:39,501 --> 00:09:41,597
While it's doing that, you're showing something on the screen.

161
00:09:41,597 --> 00:09:42,158
What is that?

162
00:09:42,158 --> 00:09:47,280
Is that like a UI that's meta provides or how if people want to give it a try, how do
they?

163
00:09:47,905 --> 00:09:49,917
playground of Meta, can do that.

164
00:09:49,917 --> 00:09:51,808
You can open it for free basically.

165
00:09:53,989 --> 00:09:55,540
And then you get here three things to play.

166
00:09:55,540 --> 00:09:57,311
The original sound, which we just heard.

167
00:09:57,311 --> 00:09:59,778
The isolated sound or the without isolated sound.

168
00:09:59,778 --> 00:10:01,764
So I'm just gonna play the isolated sound now.

169
00:10:01,764 --> 00:10:04,064
and it's more or less just a kick drum.

170
00:10:04,064 --> 00:10:04,741
It's cooler.

171
00:10:04,741 --> 00:10:09,273
And if I would now play the original one without the isolated sound, the kick drum is
goad.

172
00:10:09,273 --> 00:10:11,714
um I'm gonna start over here.

173
00:10:11,714 --> 00:10:14,547
I'm gonna say give me the vocals.

174
00:10:14,547 --> 00:10:17,129
take again 10 to 15 seconds.

175
00:10:18,172 --> 00:10:19,134
It's really good.

176
00:10:19,134 --> 00:10:21,197
Isolating audio and there we have it.

177
00:10:21,197 --> 00:10:23,621
I'm gonna play just the isolated one again.

178
00:10:23,621 --> 00:10:24,561
It's crazy good, huh?

179
00:10:24,561 --> 00:10:25,296
Wow, it is...

180
00:10:25,296 --> 00:10:29,898
And I'm gonna, one more, I'm gonna try to isolate the keys, so the keyboard.

181
00:10:29,898 --> 00:10:37,049
And the thing about this is like you, in the small sample, you hear the keyboard more or
less, I heard it more or less at the end.

182
00:10:37,049 --> 00:10:41,200
But actually it plays throughout the whole track, it becomes much more noticeable when you
isolate it.

183
00:10:41,200 --> 00:10:42,135
So here we have it.

184
00:10:42,135 --> 00:10:45,191
I wasn't even aware that this was the sound I playing in the background.

185
00:10:45,191 --> 00:10:45,703
Yeah.

186
00:10:45,703 --> 00:10:50,685
And it's very clean, And sometimes you hear little bit of distortion, but it's very clean
in general.

187
00:10:50,685 --> 00:10:51,946
You can just download it.

188
00:10:51,946 --> 00:10:52,777
I think this will...

189
00:10:52,777 --> 00:10:57,659
A lot of people will use this just to create new beats, new sample slice and stuff.

190
00:10:58,220 --> 00:11:01,152
Because this used to be super hard to do at it because you had like...

191
00:11:01,152 --> 00:11:08,647
If you were into sample slicing, you had to find like a small piece where there was no
vocals on top and like where it was clean, where you can edit out stuff.

192
00:11:08,647 --> 00:11:11,269
But here you can just say, give me this and it will give it to you.

193
00:11:11,269 --> 00:11:11,799
No, indeed.

194
00:11:11,799 --> 00:11:13,550
The quality is really good.

195
00:11:13,550 --> 00:11:15,170
I'm actually super impressed.

196
00:11:15,968 --> 00:11:20,973
And also there is a I mean, the complexity of like, because you just typed it, like also.

197
00:11:21,161 --> 00:11:25,545
Like you type something and understands what you're trying to say and it takes the right
stuff out of it.

198
00:11:25,545 --> 00:11:25,725
Right.

199
00:11:25,725 --> 00:11:29,638
Like I feel like it's it's it's very human like like the interaction.

200
00:11:29,638 --> 00:11:29,888
Right.

201
00:11:29,888 --> 00:11:31,179
Like you say, give me this.

202
00:11:31,179 --> 00:11:34,556
But like actually to make that work, it's super complicated.

203
00:11:34,556 --> 00:11:36,473
Yeah, I'm very impressed.

204
00:11:36,473 --> 00:11:39,406
It's cool to play around with.

205
00:11:39,406 --> 00:11:40,570
I'm start a new band part.

206
00:11:40,570 --> 00:11:42,250
uh

207
00:11:42,250 --> 00:11:43,296
we have to get it with you,

208
00:11:43,296 --> 00:11:44,917
The monkey patching, I don't

209
00:11:44,917 --> 00:11:46,282
monkey patching samplers.

210
00:11:46,282 --> 00:11:49,383
We'll sell CDs on our AI vending machine.

211
00:11:49,383 --> 00:11:52,242
cassette tape out of the back of our van.

212
00:11:52,877 --> 00:11:53,968
You know, man, indeed.

213
00:11:53,968 --> 00:11:54,920
All right.

214
00:11:54,920 --> 00:11:55,951
What do we have?

215
00:11:55,951 --> 00:11:57,211
This is really cool, by the way.

216
00:11:57,211 --> 00:11:58,631
By the way, maybe one last question, actually.

217
00:11:58,631 --> 00:12:00,891
Do you know if this is open source?

218
00:12:00,891 --> 00:12:04,208
If I want to download and run it, is there any licenses or anything or?

219
00:12:04,208 --> 00:12:09,529
Not that I read, but I also didn't explicitly search for it, but I don't think it was in
the announcement.

220
00:12:09,529 --> 00:12:13,409
yeah, I don't think I don't think so either but But very very very cool again.

221
00:12:13,409 --> 00:12:14,282
I'm impressed

222
00:12:14,282 --> 00:12:15,980
What else is there?

223
00:12:15,980 --> 00:12:16,608
What's next?

224
00:12:16,608 --> 00:12:18,968
I'm just quickly verifying here.

225
00:12:18,968 --> 00:12:27,359
There is actually some audio repo on Facebook research, provides code and links for
downloading the trained model checkpoints.

226
00:12:27,359 --> 00:12:30,110
So it does seem to be there in the last commits from three weeks ago.

227
00:12:30,110 --> 00:12:31,510
So that more or less corresponds.

228
00:12:31,510 --> 00:12:35,981
So I think there is at least a more or less open model.

229
00:12:35,981 --> 00:12:36,852
Hmm, interesting.

230
00:12:36,852 --> 00:12:38,323
Maybe what's the license as well?

231
00:12:38,323 --> 00:12:39,964
I'm just wondering if it's...

232
00:12:40,105 --> 00:12:42,688
okay, Sam license, whatever that means.

233
00:12:42,688 --> 00:12:43,722
Just wondering if...

234
00:12:43,722 --> 00:12:44,716
to the next article.

235
00:12:44,716 --> 00:12:54,155
A new leak claims OpenAI and Joni Ives' hardware venture may be exploring an AI-powered
pen as part of a still mysterious Chachipiti gadget lineup.

236
00:12:54,155 --> 00:12:58,118
A pen sounds quaint, but in practice it could be an always-available sensor.

237
00:12:58,118 --> 00:13:02,573
So the real debate is whether convenience beats the privacy trade-offs.

238
00:13:02,573 --> 00:13:06,867
OpenAI acquired the company from Joni Ives.

239
00:13:06,867 --> 00:13:12,632
um Joni Ives himself is someone that comes from design at Apple.

240
00:13:12,632 --> 00:13:14,596
The acquisition was for lot of money.

241
00:13:14,596 --> 00:13:17,738
I want to say six billion, but maybe I'm completely wrong.

242
00:13:17,738 --> 00:13:21,802
Yeah, for approximately 6.4 to 6.5 billion.

243
00:13:21,802 --> 00:13:23,324
So it's a big acquisition.

244
00:13:23,324 --> 00:13:24,777
And I was really...

245
00:13:24,777 --> 00:13:36,796
like something cool is gonna come out and now the rumor is that it's gonna be a I'm like,
A pen?

246
00:13:36,902 --> 00:13:39,074
Yeah, an AI powered pen.

247
00:13:39,074 --> 00:13:40,703
So what do you think?

248
00:13:40,703 --> 00:13:42,080
Okay, let's brainstorm a bit.

249
00:13:42,080 --> 00:13:43,477
What do think this pen could do?

250
00:13:43,477 --> 00:13:49,211
Well, the rumors are that you can scribble something and it will quote unquote translated
to written text.

251
00:13:49,211 --> 00:13:52,113
I'm not sure how relevant that is, right?

252
00:13:52,113 --> 00:13:53,765
I can just maybe also type it.

253
00:13:53,765 --> 00:13:56,378
But it will also record audio.

254
00:13:56,378 --> 00:14:02,945
So you can whisper against it like we've discussed this earlier with smart rings, which I
think there is use cases to that.

255
00:14:02,945 --> 00:14:04,353
I think a smart ring would be...

256
00:14:04,353 --> 00:14:12,115
whispering to it and like recording notes and stuff like that would be more useful to me
because the smart ring is just easier right like when I'm running I can use my smart ring

257
00:14:12,115 --> 00:14:24,344
I'm not gonna carry my pen right um yeah yeah just more usable when it comes to recording
audio like in more different places when you're a ring versus a pen and just the

258
00:14:24,344 --> 00:14:31,723
scribbling of notes is like I don't know man like that sounds it sounds like like it's
just like a fun feature but it's not like

259
00:14:31,723 --> 00:14:33,513
It's definitely not worth 6 billion, right?

260
00:14:33,513 --> 00:14:37,196
For me, it sounds like a party trick.

261
00:14:37,196 --> 00:14:40,288
Like, look, just write something on my back and I'll tell you what it is.

262
00:14:40,288 --> 00:14:41,551
And it is like...

263
00:14:41,551 --> 00:14:42,160
And I'm also...

264
00:14:42,160 --> 00:14:45,272
this is one of the three ideas they're working on.

265
00:14:45,272 --> 00:14:48,004
So it's definite that this will actually come out.

266
00:14:48,004 --> 00:14:49,385
Or just this will come out.

267
00:14:49,385 --> 00:14:51,216
So we'll have to see how or what.

268
00:14:51,216 --> 00:14:52,408
But I hope it will be...

269
00:14:52,408 --> 00:14:56,391
When OpenAI comes out with hardware, I hope it will be bigger than a pen.

270
00:14:56,391 --> 00:14:58,662
Yeah, it's a bit underwhelming, right?

271
00:14:58,662 --> 00:15:06,076
They did say so Sam Altman has gone on record saying that the device should feel like a
quote unquote cabin by a lake.

272
00:15:06,076 --> 00:15:07,247
So I guess it should be.

273
00:15:07,247 --> 00:15:10,053
I don't know what this means, but I guess it's more like.

274
00:15:14,575 --> 00:15:15,155
I don't know.

275
00:15:15,155 --> 00:15:23,089
I think I guess it feels like non intrusive, like I guess not a lot of notifications,
something chill, like something, you know, that doesn't disturb.

276
00:15:23,089 --> 00:15:23,689
Kinda.

277
00:15:23,689 --> 00:15:24,090
Yeah.

278
00:15:24,090 --> 00:15:24,854
I mean, because.

279
00:15:24,854 --> 00:15:31,536
But I guess like I'm thinking like I have, and this is a bit of an overkill before I say I
do have an Auraring and an Apple Watch, right?

280
00:15:31,536 --> 00:15:35,916
If I had to say which one feels more like a cabin by a lake, I'd say the Auraring.

281
00:15:35,916 --> 00:15:42,127
You know, like you can kind of, that's a good, that's kind of like how I see it, right?

282
00:15:42,127 --> 00:15:46,929
So I think it's going to be something not super intrusive that is there for you, but it's
not very.

283
00:15:46,929 --> 00:15:48,319
no extra notifications.

284
00:15:48,319 --> 00:15:51,400
Hopefully no external notifications, but I'm also one.

285
00:15:51,639 --> 00:15:52,709
Yeah.

286
00:15:52,709 --> 00:15:52,871
Yeah.

287
00:15:52,871 --> 00:15:54,252
You will be forced to write everything down.

288
00:15:54,252 --> 00:15:54,842
Yeah, exactly.

289
00:15:54,842 --> 00:15:56,351
Like, so let's see.

290
00:15:56,351 --> 00:15:56,553
Yeah.

291
00:15:56,553 --> 00:15:57,878
I'm also like, I don't know.

292
00:15:57,878 --> 00:16:00,991
I'm a bit skeptical about hardware devices.

293
00:16:00,991 --> 00:16:08,478
I think there's a lot of, there were a lot of tries for AI, like the humane AI pin or
wearable pin or what was the AI pin or something.

294
00:16:08,478 --> 00:16:09,822
Yeah, yeah, I depend, yeah.

295
00:16:09,822 --> 00:16:11,566
That was the Rabbit R1.

296
00:16:11,566 --> 00:16:14,654
were a few things and I mean, even like the AI goggles, right?

297
00:16:14,654 --> 00:16:17,170
Like the Ray-Ban glasses from Meta and all these things like.

298
00:16:17,170 --> 00:16:27,574
But I think like I'm quite excited about the the re-pebble ring Which is like the very
affordable ring that I don't think they have launched yet But it's from the company that

299
00:16:27,574 --> 00:16:36,042
is bringing Pebble back but it just becomes like it's not that it's a smart ring but it's
just a device that becomes an input to an AI model and in this case for note-taking right

300
00:16:36,042 --> 00:16:40,524
like I think we should see it like that's like it's an easy way of an input

301
00:16:40,524 --> 00:16:42,315
Yeah, but that I think it's that I agree.

302
00:16:42,315 --> 00:16:43,406
I think that one is a good one.

303
00:16:43,406 --> 00:16:47,060
I mean, also I'm saying this, but I really enjoyed the watering, for example.

304
00:16:47,060 --> 00:16:49,263
And maybe before you asked me, I wasn't very convinced.

305
00:16:49,263 --> 00:16:51,218
I was like, I'll try it because it was a gift.

306
00:16:51,218 --> 00:16:52,480
So I said, I'll give it a try.

307
00:16:52,480 --> 00:16:55,082
But if I had to buy myself, I'm not sure if I would have.

308
00:16:55,082 --> 00:16:56,213
And actually, I really enjoy it.

309
00:16:56,213 --> 00:17:00,582
So again, I'm skeptical, but I know that I've been wrong in the past.

310
00:17:00,582 --> 00:17:01,824
Aura should come with a mic.

311
00:17:01,824 --> 00:17:02,387
I think so.

312
00:17:02,387 --> 00:17:05,075
feel like Aura had like, yeah.

313
00:17:06,423 --> 00:17:06,990
Yeah.

314
00:17:06,990 --> 00:17:09,053
think they will when we see competitors coming up.

315
00:17:09,053 --> 00:17:10,776
Because there's another one as well, I forgot the name.

316
00:17:10,776 --> 00:17:12,888
And that basically does the same as the pedalboard one here.

317
00:17:12,888 --> 00:17:14,901
Like just you can whisper notes into it.

318
00:17:14,901 --> 00:17:16,223
I think Aura will follow.

319
00:17:16,223 --> 00:17:21,954
Yeah, but I do think it's, mean, because the ring also, I was thinking if you have the
pen, right, maybe you can clip the pen on your shirt.

320
00:17:21,954 --> 00:17:23,837
But also like the ring is just always there.

321
00:17:23,837 --> 00:17:25,170
It's like steady on your hand.

322
00:17:25,170 --> 00:17:26,592
Yeah, it's just there.

323
00:17:26,592 --> 00:17:27,434
Like it's not gonna fall.

324
00:17:27,434 --> 00:17:28,344
It's not gonna.

325
00:17:28,344 --> 00:17:32,757
or running or do something where you're not sitting at a desk and like a ring is way
easier than a pen.

326
00:17:32,757 --> 00:17:33,597
Exactly.

327
00:17:33,597 --> 00:17:34,177
Yeah, exactly.

328
00:17:34,177 --> 00:17:36,497
mean, things that are just like quote unquote attached to your body.

329
00:17:36,497 --> 00:17:38,537
It's also by your hand is very deck style as well.

330
00:17:38,537 --> 00:17:38,737
Right.

331
00:17:38,737 --> 00:17:42,628
Like even the watch sometimes you have to use two hands or something like for the ring is
just there.

332
00:17:42,628 --> 00:17:43,408
So, yeah.

333
00:17:43,408 --> 00:17:44,328
Well, let's see.

334
00:17:44,328 --> 00:17:45,259
I also.

335
00:17:45,259 --> 00:17:48,210
I also hope it will be something more.

336
00:17:48,210 --> 00:17:48,921
I don't know.

337
00:17:48,921 --> 00:17:51,313
less underwhelming than a pen, but we'll see.

338
00:17:51,313 --> 00:17:52,754
What else do we have?

339
00:17:52,754 --> 00:18:01,698
We have Netflix open content library publishes high-end test footage and assets, 4K HDR,
high frame rates and Atmos mixes.

340
00:18:01,698 --> 00:18:07,209
So researchers and engineers can stress test codecs and workflows without using real
shows.

341
00:18:07,209 --> 00:18:14,583
It's a rare peak behind the streaming curtain and a useful reminder that video quality is
built on lots of measurable trade-offs.

342
00:18:14,583 --> 00:18:22,762
So when I saw this, was really open source and Netflix is not two things you expect to see
together very often.

343
00:18:22,762 --> 00:18:24,835
Like, what does it mean open source here?

344
00:18:24,835 --> 00:18:28,309
Yeah, this was released, I want to say two weeks ago.

345
00:18:28,309 --> 00:18:38,439
And they've released a lot of content, haven't gone through everything, under CC by 4, a
Creative Commons license, where you can basically like, you share it, copy and

346
00:18:38,439 --> 00:18:42,373
redistribute it, you can adapt it, you can use it for commercial purposes.

347
00:18:42,373 --> 00:18:45,166
But you need to follow the terms of the license.

348
00:18:45,166 --> 00:18:49,097
The only requirement that this license gives you is that you need to attribute.

349
00:18:49,097 --> 00:18:54,274
So you need to very clearly show where you got this content from when you do something
with it, whatever you do with it.

350
00:18:54,274 --> 00:19:02,799
Maybe to break it down, Like open source, I imagine most of the people that listen to us,
they're thinking of code, but these are like video assets.

351
00:19:02,799 --> 00:19:05,274
What does open source here means?

352
00:19:05,274 --> 00:19:12,954
you have stuff here like videos, like images, like audio that you can more or less do the
same thing with than you would do with code.

353
00:19:12,954 --> 00:19:16,728
Like you can share it, you can change it, you can use it for commercial purposes.

354
00:19:16,728 --> 00:19:17,401
But you need to...

355
00:19:17,401 --> 00:19:18,301
would you change it?

356
00:19:18,301 --> 00:19:19,372
guess that's the thing, right?

357
00:19:19,372 --> 00:19:26,712
Like, why are you not going to make like when you think of open source, I think of people
contributing to like there's a product that evolves and people are contributing to it,

358
00:19:26,712 --> 00:19:27,253
right?

359
00:19:27,253 --> 00:19:30,276
let's take the example of the audio we were just editing.

360
00:19:30,276 --> 00:19:40,017
You take some original piece of audio and I'm gonna change or take a part of it or like
re-slice it or and the new output will be my deliverable.

361
00:19:40,017 --> 00:19:42,079
But I still need to show the attribution.

362
00:19:42,079 --> 00:19:47,116
I got a lot of input from whatever the source is near Netflix open source repository.

363
00:19:47,116 --> 00:19:47,657
I see.

364
00:19:47,657 --> 00:19:54,706
So what you're saying is like open source, not in the sense as much of like I'm going to
contribute to this film and I'm going to make changes so the film is better, but it's more

365
00:19:54,706 --> 00:19:57,861
like I'm going to use assets for this for my own purposes.

366
00:19:57,861 --> 00:19:58,513
And I just.

367
00:19:58,513 --> 00:20:05,050
maybe look at it as like using an open source library or something.

368
00:20:05,050 --> 00:20:10,293
Indeed, you're not going to contribute to the open source project, but you're maybe going
to use it for your own purposes.

369
00:20:11,074 --> 00:20:12,866
That's a fair point to make.

370
00:20:12,866 --> 00:20:16,100
And also like this was test content, right?

371
00:20:16,100 --> 00:20:18,002
I think they mentioned here at some point as well.

372
00:20:18,002 --> 00:20:23,796
by test and I looked a bit because they have they also linked to the blog posts from
Netflix as well.

373
00:20:23,796 --> 00:20:28,109
So engineering is making movies, aka test content.

374
00:20:28,109 --> 00:20:32,444
So apparently this was also used to test a bit like how does the

375
00:20:32,444 --> 00:20:40,011
the different frames per second or like the contrast and like they also have some films to
see how it will look, right?

376
00:20:40,011 --> 00:20:46,975
They have some things that they're not sure if it's gonna look good on screen or maybe
it's not gonna sound great um for streaming purposes as well.

377
00:20:46,975 --> 00:20:50,158
Like can we stream as many frames or that many pixels?

378
00:20:50,158 --> 00:20:52,890
So I think this is also something that they've been trying.

379
00:20:52,890 --> 00:20:59,034
So it is, mean, I haven't like even like there's a Cosmos laundromat, which is a
animation, right?

380
00:20:59,034 --> 00:21:08,934
Sparks, think this is more on the contrast to see what's the quality of these things So I
thought it was I mean, I never thought of these things when I'm watching these movies,

381
00:21:08,934 --> 00:21:10,105
right?

382
00:21:10,105 --> 00:21:17,765
But I thought it was interesting interesting to see as well one thing that I was a bit
because I was curious about like I don't know this one like Nocturne and I tried watch on

383
00:21:17,765 --> 00:21:27,465
Netflix and I just searched on my Netflix account and I couldn't find it so it was a bit
like This is real, you know, but haven't looked it.

384
00:21:27,465 --> 00:21:27,952
Yeah

385
00:21:27,952 --> 00:21:30,285
So yeah, interesting stuff as well.

386
00:21:30,285 --> 00:21:39,049
And I think for me also getting linked to the seeing a bit behind the scenes like the
engineering part and the concerns that they have, I thought it was also interesting for me

387
00:21:39,049 --> 00:21:40,683
to think a bit about these problems.

388
00:21:40,683 --> 00:21:42,915
Yeah, it's cool to see Netflix doing this.

389
00:21:42,915 --> 00:21:54,586
Like it's this from what they're showing here and also earlier the stands on Gen.ai
generated content like they seem to be like a positive optimistic player in the whole

390
00:21:54,967 --> 00:21:57,071
authentic content sphere.

391
00:21:57,071 --> 00:22:01,690
I wonder how, like, if they're gonna start doing stuff with JNI for this as well, to try
it out.

392
00:22:01,690 --> 00:22:04,965
Like something like a bit of an open research thing, right?

393
00:22:04,965 --> 00:22:06,378
Like how are people doing this?

394
00:22:06,378 --> 00:22:07,330
We'll be interested to see.

395
00:22:07,330 --> 00:22:07,643
is...

396
00:22:07,643 --> 00:22:12,332
Well, but it's probably actually the case when I think about it like this content can use
it to train models.

397
00:22:12,332 --> 00:22:15,039
The answer is probably yes, as long as you give attribution.

398
00:22:15,039 --> 00:22:17,387
Yeah, I think...

399
00:22:17,387 --> 00:22:18,963
I don't know if it's enough.

400
00:22:18,963 --> 00:22:23,858
the output of those models need to give attribution, which is the hard thing, right?

401
00:22:24,565 --> 00:22:30,585
Yeah, I think today there are a lot of models that are trained on a lot of stuff and they
definitely don't give a fuck, right?

402
00:22:31,505 --> 00:22:34,756
I think it's good.

403
00:22:34,756 --> 00:22:39,607
I also thought that open source maybe is more like for training models and stuff.

404
00:22:39,607 --> 00:22:41,547
But then I also looked at the date.

405
00:22:41,547 --> 00:22:45,858
The first one here is from 2013, which was before the whole Gen.Eye craze, right?

406
00:22:45,858 --> 00:22:47,298
So yeah, cool.

407
00:22:47,298 --> 00:22:47,718
Interesting.

408
00:22:47,718 --> 00:22:50,073
Maybe related to this, what do we have next?

409
00:22:50,073 --> 00:22:59,832
PC Gamer revisits Disney's much mocked AI generated Star Wars Field Guide video, an odd
parade of scrambled animals sold as futuristic creativity.

410
00:22:59,832 --> 00:23:04,696
It's framed as the first stumble in a year of AI embarrassments and the tension is simple.

411
00:23:04,696 --> 00:23:09,932
When big studios chase generative shortcuts, what happens to craft and trust?

412
00:23:09,932 --> 00:23:13,035
I saw the bit from the TED talk actually.

413
00:23:13,035 --> 00:23:18,731
So if there's a link here that they already go to the, yeah, here, they already go
directly to the part.

414
00:23:18,731 --> 00:23:21,565
It was, so maybe to set the stage, right?

415
00:23:21,565 --> 00:23:24,150
This was a TED talk from Disney.

416
00:23:24,150 --> 00:23:24,750
Who is this?

417
00:23:24,750 --> 00:23:25,822
Who's saying this actually?

418
00:23:25,822 --> 00:23:27,303
I someone from Disney, right?

419
00:23:27,303 --> 00:23:31,900
And then he's talking about using Gen.ai for creating new movies, new scenes and...

420
00:23:31,900 --> 00:23:35,620
Basically, he shows, I mean, I'm putting here on the screen a little bit, right?

421
00:23:35,620 --> 00:23:38,227
Basically, this was like a field guide for Star Wars.

422
00:23:38,227 --> 00:23:38,738
So.

423
00:23:38,738 --> 00:23:46,330
It was a bit like what would the experience be if we would land on another planet and
would be able to look around there.

424
00:23:46,330 --> 00:23:47,603
Yeah, exactly.

425
00:23:47,603 --> 00:23:51,291
And then they show a lot of basically a lot of animals, but it's clearly.

426
00:23:51,291 --> 00:23:53,366
And the result is sad, right?

427
00:23:54,252 --> 00:23:55,955
It's just AI slop.

428
00:23:56,689 --> 00:23:57,940
I mean, I think so.

429
00:23:57,940 --> 00:23:59,260
It is a eyes lop.

430
00:23:59,260 --> 00:24:05,120
For example, now we're seeing like a polar bear with some like tiger, tiger, tiger.

431
00:24:05,120 --> 00:24:06,580
Yeah, it's very basic, right?

432
00:24:06,580 --> 00:24:08,991
Like he's like, you clearly just put in two animals together.

433
00:24:08,991 --> 00:24:10,391
You can still see the animals.

434
00:24:10,391 --> 00:24:12,542
It's nothing like Star Wars or anything, right?

435
00:24:12,542 --> 00:24:13,833
So it is very sad.

436
00:24:13,833 --> 00:24:16,624
But on the other hand, too, it is does look very realistic.

437
00:24:16,624 --> 00:24:19,784
We're saying it's very sad because we're in 2026, right?

438
00:24:19,784 --> 00:24:23,623
If you say like one year ago and you know, like

439
00:24:23,623 --> 00:24:24,401
You don't think so?

440
00:24:24,401 --> 00:24:27,893
I think you mean it looks realistic as in the quality is very good.

441
00:24:28,113 --> 00:24:34,825
That I agree with, but like I think 10 years ago we would also have said like there is no
creativity in this whatsoever.

442
00:24:35,356 --> 00:24:40,656
Like the assignment is like walk around in a forest on a planet far, far away.

443
00:24:40,656 --> 00:24:42,706
What kind of animals would you encounter?

444
00:24:42,706 --> 00:24:46,978
And then it's just like, I don't know, like a bird with a snail.

445
00:24:47,943 --> 00:24:48,463
Yeah.

446
00:24:48,463 --> 00:24:57,152
house on the back right like like it's just animals mashed together like a polar bear with
tiger stripes like i mean the person the the creative director where they would have been

447
00:24:57,152 --> 00:24:57,589
fired

448
00:24:57,589 --> 00:24:58,749
Yeah, no, that's for sure.

449
00:24:58,749 --> 00:24:59,760
That's for sure.

450
00:24:59,760 --> 00:25:00,601
they said this.

451
00:25:00,601 --> 00:25:06,041
So the other thing too, that made it bit more awkward is that he said he spent two weeks
on this.

452
00:25:06,041 --> 00:25:07,892
There was someone spent two weeks on this.

453
00:25:07,892 --> 00:25:09,712
It did say a few times that this is experimental.

454
00:25:09,712 --> 00:25:11,363
This is early stages and this and this.

455
00:25:11,363 --> 00:25:15,943
But at the same time, they did mention like they're very proud of this, which is very
weird.

456
00:25:15,943 --> 00:25:16,223
Right.

457
00:25:16,223 --> 00:25:16,890
Like this is

458
00:25:16,890 --> 00:25:25,380
Well, can imagine that two weeks is a very short time to have something like this, If you
would actually have to animate this, would probably take a lot more time.

459
00:25:25,380 --> 00:25:26,807
m

460
00:25:26,807 --> 00:25:31,759
with like, I don't know, Nano, Banana, Sora and you have, because everything's also short
clips, right?

461
00:25:31,759 --> 00:25:34,813
But you're arguing whether or not two weeks is a good period or...

462
00:25:34,813 --> 00:25:39,287
I it's like if you spent two weeks on this, I would expect something that was more thought
through.

463
00:25:39,287 --> 00:25:43,629
Yeah, well, probably shows how hard it is to get consistent output there.

464
00:25:43,629 --> 00:25:53,199
Like if you want to create a bird with a snail's house on the back, you probably have to
generate that 30 times because there are always artifacts and shit like that.

465
00:25:55,706 --> 00:26:01,437
I don't know, I agree with the community consensus about this is just crap.

466
00:26:01,437 --> 00:26:06,077
And that in the context that Disney has basically said we're gonna go...

467
00:26:06,077 --> 00:26:13,653
all out on on Gen.ai and investing 1 billion in equity in OpenAI, they announced it a few
weeks ago.

468
00:26:13,653 --> 00:26:15,250
um

469
00:26:15,250 --> 00:26:21,822
like we need help let's let's put some money on eh

470
00:26:21,822 --> 00:26:25,947
and I'm not even saying that it's not a good tool, right?

471
00:26:25,947 --> 00:26:35,502
Like you're saying, it looks realistic, it looks good, having a GenAI video generating
tool doesn't mean that it brings creativity.

472
00:26:36,183 --> 00:26:38,330
It just generates moving images.

473
00:26:38,330 --> 00:26:39,301
I fully agree.

474
00:26:39,301 --> 00:26:46,137
I think it's like the the audience member in me is very disappointed because it's like it
doesn't add anything.

475
00:26:46,137 --> 00:26:46,377
Right.

476
00:26:46,377 --> 00:26:49,130
And this is like, again, they also mentioned Star Wars.

477
00:26:49,130 --> 00:26:50,122
Like why?

478
00:26:50,122 --> 00:26:52,314
There is nothing Star Wars on this whole thing.

479
00:26:52,314 --> 00:26:52,664
Right.

480
00:26:52,664 --> 00:26:56,351
Like the engineering me thinks this is impressive.

481
00:26:57,171 --> 00:26:59,444
You know, like so so.

482
00:26:59,444 --> 00:27:00,306
So yeah, indeed.

483
00:27:00,306 --> 00:27:04,249
I think there are still ways and I think, I mean, also it's still very kind of new, right?

484
00:27:04,249 --> 00:27:10,086
I don't think people have been playing as much with a JNI video for like these kinds of
things.

485
00:27:10,086 --> 00:27:13,330
I do think that in the future for sure it will play a role, but...

486
00:27:13,330 --> 00:27:19,348
It's gonna take a while before it does 80 % of the job for them.

487
00:27:19,348 --> 00:27:26,677
I wonder if there's a way, and I don't think it would be just with video, but to use AI to
pre-generate assets that they can edit later on.

488
00:27:26,677 --> 00:27:33,065
I kinda like how, I think, gaming engines, you have an interface, but it can also go in
the code level.

489
00:27:33,065 --> 00:27:33,485
Mm-hmm.

490
00:27:33,485 --> 00:27:39,372
have like a GenAI layer that kind of precreates the rough and then you can actually import
these assets and actually edit these things.

491
00:27:39,372 --> 00:27:40,463
I wonder if there's something like this.

492
00:27:40,463 --> 00:27:44,548
think that's probably the safest way to go, I would say.

493
00:27:44,548 --> 00:27:53,593
Yeah, it already exists to some extent, like in modern video tools where you can define
characters so you can reuse them in multiple scenarios.

494
00:27:53,593 --> 00:27:55,196
But yeah, let's see.

495
00:27:55,196 --> 00:27:57,210
Also, I feel that it's moving very fast.

496
00:27:57,210 --> 00:27:58,314
Yeah, that's true.

497
00:27:58,314 --> 00:27:59,126
That's true.

498
00:27:59,126 --> 00:28:01,841
I also think there's a lot of money there, right?

499
00:28:04,054 --> 00:28:05,566
I'll go to something completely different.

500
00:28:05,566 --> 00:28:09,382
Or you're gonna go to something completely different.

501
00:28:09,382 --> 00:28:10,118
Go ahead.

502
00:28:10,118 --> 00:28:12,280
I will go for something.

503
00:28:12,280 --> 00:28:22,634
have China has floated draft rules aimed at stopping chatbots from encouraging suicide,
self harm or violence, pushing the safety burden onto AI providers.

504
00:28:22,634 --> 00:28:29,126
The proposal would require human intervention when suicide is mentioned and set public
feedback deadline of January 25th.

505
00:28:29,126 --> 00:28:34,700
So 20 days from now ish testing how far regulation can reach into conversations.

506
00:28:34,700 --> 00:28:36,011
So basically it's

507
00:28:36,011 --> 00:28:36,991
a rule, right?

508
00:28:36,991 --> 00:28:41,202
Like it's still not fully 100 % approved, but in China.

509
00:28:41,202 --> 00:28:45,367
But it's very short term that they want to apply it, right?

510
00:28:45,367 --> 00:28:53,432
Yeah, think 25th of January is the deadline for providing feedback.

511
00:28:53,432 --> 00:28:55,081
25th year public feedback.

512
00:28:55,081 --> 00:28:55,775
Okay.

513
00:28:55,775 --> 00:28:56,518
Not applied.

514
00:28:56,518 --> 00:28:57,458
not applied.

515
00:28:57,458 --> 00:29:08,478
basically, well, we do, we have seen cases of AI assisted suicide or self harm mostly in
the, well, I think there were a few cases in Europe, but also in the US.

516
00:29:08,478 --> 00:29:12,281
I hadn't heard as much in China, but I don't know if we hear as much from China.

517
00:29:12,281 --> 00:29:17,847
And I guess I heard that this is one of the, it's a very, how do you say, not restricting,
but it's a very,

518
00:29:17,847 --> 00:29:19,182
What's the word I'm looking for here?

519
00:29:19,182 --> 00:29:21,219
I think worldwide is one of the most...

520
00:29:21,219 --> 00:29:26,956
regulatory for AI in the self harm like most proactive.

521
00:29:26,956 --> 00:29:28,106
think that's the

522
00:29:28,176 --> 00:29:33,727
as in tackling the risk of self harm initiated by AI.

523
00:29:33,727 --> 00:29:42,898
So what they want to do basically is that this draft like it requires human intervention
from the moment anything around suicide is mentioned.

524
00:29:42,898 --> 00:29:53,089
How that will exactly look like in practice, not yet clear, but it basically it puts the
responsibility of AI in this context and pushes it back to the provider.

525
00:29:53,089 --> 00:29:58,624
So if I provide a chat layer to chat with an LLM, I need to have monitoring for this in
place.

526
00:29:58,624 --> 00:30:01,506
And when someone mentions suicide, I need to do something with it.

527
00:30:01,506 --> 00:30:11,853
And what that is, is I think yet a bit to be seen, like notifying parents, notifying
whatever help organization, whatever, but something needs to happen at that point.

528
00:30:12,105 --> 00:30:21,976
Yeah, so I saw here on the article here, they say users, especially minors and elderly,
will have to provide a contact information for a guardian when they register, who will be

529
00:30:21,976 --> 00:30:24,147
notified with suicide or self-harm is discussed.

530
00:30:24,147 --> 00:30:26,707
So I guess it's like as a guardian.

531
00:30:26,707 --> 00:30:27,387
Yeah, exactly.

532
00:30:27,387 --> 00:30:29,927
So there is a person needs to be notified.

533
00:30:29,927 --> 00:30:31,707
So it's not someone from the provider.

534
00:30:31,707 --> 00:30:36,858
They say that this service wouldn't be for every eye provider, but basically the relevant
ones, quote unquote.

535
00:30:36,858 --> 00:30:39,449
So the ones exceeding 1 million registered users.

536
00:30:39,449 --> 00:30:41,219
or more than 100,000 monthly active users.

537
00:30:41,219 --> 00:30:46,150
But still, like you said, the provider, but the provider has a responsibility that the
guardian gets the message, right?

538
00:30:46,150 --> 00:30:50,150
So they need to monitor it and they need to create a notification.

539
00:30:50,847 --> 00:30:52,207
Yeah, I think.

540
00:30:52,684 --> 00:31:03,312
in Europe we have the AI act, which I don't know if because it covers as much like this
very concrete thing in the US I saw like I think in New York there are a few laws around

541
00:31:03,312 --> 00:31:03,613
this.

542
00:31:03,613 --> 00:31:05,745
But I think this looks like is the most.

543
00:31:05,745 --> 00:31:08,107
Concrete one in a way as well.

544
00:31:08,107 --> 00:31:08,578
so as well.

545
00:31:08,578 --> 00:31:12,091
I think so the people, I think I'm in favor of this.

546
00:31:12,091 --> 00:31:18,005
I think the people not in favor will be privacy advocates because it's a very sensitive
thing, right?

547
00:31:18,005 --> 00:31:21,518
Like it's to get your guardians notified from the moment you're discussing this.

548
00:31:21,518 --> 00:31:24,591
But I think those downsides are offset by the upsides.

549
00:31:24,591 --> 00:31:25,391
I think so.

550
00:31:25,391 --> 00:31:30,671
And I think the fact that you're choosing the guardian should be someone you trust as
well, right?

551
00:31:30,671 --> 00:31:36,351
Like it's not like there's a, not like someone, like it's not like it's someone that you
don't know that is going to be notified, right?

552
00:31:36,505 --> 00:31:40,258
But still, it's probably someone that you would not discuss it with normally.

553
00:31:40,656 --> 00:31:42,659
But I think the...

554
00:31:42,659 --> 00:31:44,664
this problem has become so big.

555
00:31:44,664 --> 00:31:52,285
And also in a context where there is more more isolation of people, it becomes more more
difficult to be socially active.

556
00:31:52,285 --> 00:32:01,432
People get isolated through social media, we've heard a lot of stories like chat GPT alike
tools become their, like literally their relationship.

557
00:32:02,653 --> 00:32:03,973
So I think this is a...

558
00:32:03,973 --> 00:32:09,432
apparently is not as uncommon as I thought as we thought I think we saw some some numbers
that

559
00:32:09,432 --> 00:32:10,285
Well, exactly.

560
00:32:10,285 --> 00:32:13,717
The numbers we saw in usage were like what we expected everybody.

561
00:32:13,717 --> 00:32:18,749
the main usage would be an open AI, but it was actually a character AI.

562
00:32:20,110 --> 00:32:24,163
And character AI is very much like you built your AI character and you interact with it.

563
00:32:24,163 --> 00:32:25,063
Yeah.

564
00:32:25,203 --> 00:32:26,515
So yeah, I think it's good.

565
00:32:26,515 --> 00:32:31,319
And I hope also that the thing's a bit difficult is that we don't hear a lot from China,
right?

566
00:32:31,319 --> 00:32:36,693
Like what's the like, because I think what I would hope is that people look at this and
pay attention to this.

567
00:32:36,693 --> 00:32:39,006
And this is a successful quote unquote experiment.

568
00:32:39,006 --> 00:32:41,047
And then this can be rolled out in other places, right?

569
00:32:41,047 --> 00:32:42,838
Because I do think it's a in the right direction.

570
00:32:42,838 --> 00:32:47,503
Maybe it's not, maybe things need to be adjusted, of course, but I do think it's a real
concern, right?

571
00:32:47,503 --> 00:32:48,114
And I think

572
00:32:48,114 --> 00:32:49,801
We need to do something about it.

573
00:32:50,108 --> 00:32:53,410
the good thing of...

574
00:32:53,410 --> 00:32:59,192
which say we don't hear a lot about China, so it's a bit vague like what, how does
decision making actually happen in China?

575
00:32:59,192 --> 00:33:03,426
But it feels to me like these things like they realize it's a big enough problem, let's
try something.

576
00:33:03,426 --> 00:33:06,740
And if next year we see this doesn't work, we're gonna adjust it.

577
00:33:06,740 --> 00:33:09,910
While here in the EU, I think...

578
00:33:09,910 --> 00:33:15,460
It's also noted that this is a problem and we will probably talk five more years on this
before doing something on it.

579
00:33:15,460 --> 00:33:17,994
Yeah, because we need to make the right decision.

580
00:33:17,994 --> 00:33:20,255
We need to make the right decision.

581
00:33:20,255 --> 00:33:21,727
We can't go too quickly.

582
00:33:21,727 --> 00:33:26,210
And in the US, they probably won't do anything because it's a very, very capitalist
market.

583
00:33:26,210 --> 00:33:35,005
And what you see then is that there is a lot of what they call regulatory capture where
OpenAI proactively says, yeah, we're going to do something like this.

584
00:33:35,005 --> 00:33:37,197
We're going to build a system for this.

585
00:33:37,197 --> 00:33:41,591
And because OpenAI is already saying themselves they're going to do something, there are
not going to be much regulation.

586
00:33:41,591 --> 00:33:43,242
Yeah, and I think they also say this.

587
00:33:43,242 --> 00:33:47,425
I think it's also there's also money incentive there as well.

588
00:33:47,425 --> 00:33:48,285
Right.

589
00:33:48,285 --> 00:33:49,887
so it's different.

590
00:33:49,887 --> 00:33:51,678
I mean, there's also, of course, goodwill.

591
00:33:51,678 --> 00:33:51,988
Right.

592
00:33:51,988 --> 00:33:53,289
But it's not just that.

593
00:33:53,289 --> 00:33:55,659
I feel like everything's got a bit murky there.

594
00:33:55,659 --> 00:33:57,551
Yeah.

595
00:33:57,551 --> 00:33:58,711
So let's see what happens.

596
00:33:58,711 --> 00:34:04,356
mean, if we get any updates on this, but I think something that hopefully people are
paying attention to.

597
00:34:04,356 --> 00:34:06,417
What do we have next?

598
00:34:06,671 --> 00:34:11,373
Nvidia is bringing in top executives from AI chip startup Grok.

599
00:34:11,373 --> 00:34:13,596
Grok with a Q, not X's Grok.

600
00:34:13,596 --> 00:34:16,376
And striking a licensing deal for its inference technology.

601
00:34:16,376 --> 00:34:20,539
Another signal that the hardware race is now as much about people as silicon.

602
00:34:20,539 --> 00:34:23,450
Grok's founder Jonathan Ross is among those moving.

603
00:34:23,450 --> 00:34:27,882
So will this speed innovation or blur the line between competition and consolidation?

604
00:34:27,882 --> 00:34:34,202
So yeah, you said grok, so G-R-O-Q, not G-R-O-K, which is very different.

605
00:34:34,803 --> 00:34:38,737
the K1 is the one from XAI.

606
00:34:38,737 --> 00:34:42,411
is Grog with a Q is a company that was founded in 2016.

607
00:34:42,411 --> 00:34:45,554
It's building hardware and software basically.

608
00:34:45,554 --> 00:34:49,419
And what they mainly focus on is the inference.

609
00:34:49,419 --> 00:34:53,715
So they have specific chips that are really optimized not for training but for inference.

610
00:34:53,715 --> 00:34:55,929
and apparently quite successful.

611
00:34:55,929 --> 00:35:01,095
They have raised at almost 7 billion valuation in the past, not that long ago.

612
00:35:01,095 --> 00:35:10,909
Really by focusing on having efficient inference chips, but also the software to very
easily scale that, manage that, parallelize that in a data center.

613
00:35:10,909 --> 00:35:20,517
Yeah, also heard something about the techniques that they employ to create these chips as
well, which is apparently what Nvidia was also interested in.

614
00:35:20,517 --> 00:35:22,270
So it's not an acquisition per se,

615
00:35:22,270 --> 00:35:24,081
Not an acquisition per se.

616
00:35:24,081 --> 00:35:29,903
They basically get IP, so they have an exclusive licensing deal on architecture.

617
00:35:29,903 --> 00:35:35,737
But they also get key builders, so part of the personnel of Grok is moving there.

618
00:35:35,737 --> 00:35:40,309
Yeah.

619
00:35:40,309 --> 00:35:44,471
do you follow a bit the chips, the GPUs and all this space or no?

620
00:35:44,471 --> 00:35:46,313
A bit, I would say, a bit.

621
00:35:46,313 --> 00:35:52,887
Is this something that comes a bit as a surprise to you or did you also feel like it's a
logical move?

622
00:35:56,514 --> 00:36:03,718
It's not per se a surprise in a sense that, well, Nvidia's stock price is very high, so
it's very cheap for them to buy something.

623
00:36:03,718 --> 00:36:04,219
one thing.

624
00:36:04,219 --> 00:36:10,813
I think this has a very specific niche, like really inference optimization, which is
probably valuable for Nvidia.

625
00:36:10,813 --> 00:36:16,027
If they don't do it themselves, then they can better acquire it, so it doesn't become a
competitor.

626
00:36:16,027 --> 00:36:19,170
But they also get very very smart people on board.

627
00:36:19,932 --> 00:36:27,735
And I think Nvidia is, even though it's extremely extremely extremely successful in what
they do over the past years, they do have a lot of competition from...

628
00:36:27,735 --> 00:36:36,127
big players like the TPUs from Google, the Cranium chips from AWS, a lot of Chinese
players, so it's not that they don't have any competition.

629
00:36:36,127 --> 00:36:41,917
So I think if they can do strategic quote-unquote acquisitions, they will do it right, and
this seems like a good match.

630
00:36:41,917 --> 00:36:47,297
And maybe, do you have any idea why chips are better for inference than training?

631
00:36:47,297 --> 00:36:49,868
No, to be honest, no, I don't know why they are.

632
00:36:49,868 --> 00:36:55,523
I'm sure there may be something in the architecture that makes it more efficient and maybe
one more, yeah.

633
00:36:55,523 --> 00:36:59,843
Yeah, but goes above my head, right?

634
00:36:59,843 --> 00:37:01,914
Like, I don't know, but interesting.

635
00:37:01,914 --> 00:37:04,085
Shall we move on to the next acquisition?

636
00:37:04,085 --> 00:37:18,252
Yes, Meta is buying Manus, a buzzy AI agent startup signaling how aggressively Mark
Zuckerberg wants revenue-generating AI products inside Facebook, Instagram and WhatsApp.

637
00:37:18,252 --> 00:37:26,234
TechCrunch reports deals about 2 billion and comes amid scrutiny over Manus' China-linked
origins and Meta's massive infrastructure spend.

638
00:37:26,234 --> 00:37:29,212
um

639
00:37:29,212 --> 00:37:36,534
was, they were popular because it was about like agents, like coding agents that could
like, I think it was, what was it like Upwork?

640
00:37:36,534 --> 00:37:38,704
They could do something with Upwork or something.

641
00:37:38,704 --> 00:37:41,335
I think the value proposition is that they're very autonomous.

642
00:37:41,335 --> 00:37:45,326
normally like something like that GPT is really like a chat and they should chat with it
and it helps you along.

643
00:37:45,326 --> 00:37:50,457
But this like, manas, manas is really positioned like as a virtual colleague.

644
00:37:50,457 --> 00:37:51,637
not like a chat assistant.

645
00:37:51,637 --> 00:38:01,379
It's really a virtual colleague and you can offload stuff to manage and it will basically
plan and execute even multi-step tasks like research, can do coding, data analysis,

646
00:38:01,379 --> 00:38:07,219
business workflows with very minimal prompting in between where with something like CHPD
you have to keep prompting it.

647
00:38:09,259 --> 00:38:10,919
It was relatively early, right?

648
00:38:10,919 --> 00:38:16,514
I want to say it was already there three years ago, two and a half, something like that.

649
00:38:16,514 --> 00:38:18,651
So I think they made a lot of noise because of that.

650
00:38:18,651 --> 00:38:19,472
Hmm.

651
00:38:19,472 --> 00:38:30,703
And what I did not know, because I didn't actually, to be honest, I didn't hear lot about
it anymore since then, but they are doing 100 million in annual recurring revenue.

652
00:38:30,703 --> 00:38:31,161
That's.

653
00:38:31,161 --> 00:38:34,607
must have quite a bit of large customers.

654
00:38:34,607 --> 00:38:36,409
I do wonder where they are then, right?

655
00:38:36,409 --> 00:38:40,537
Like is they, are they in US, are they in Europe, are they in China, these customers?

656
00:38:40,537 --> 00:38:41,140
Yeah.

657
00:38:41,140 --> 00:38:48,168
So also what Meta is doing here, like they're not just buying technology and skills,
they're really buying cashflow here as well.

658
00:38:48,168 --> 00:38:49,020
Yeah, indeed.

659
00:38:49,020 --> 00:38:53,424
think I also heard, well, there are a few things that were discussed that I heard as well.

660
00:38:53,424 --> 00:38:57,950
One is that Mennon's, I think he started in China, then they moved to Singapore.

661
00:38:57,950 --> 00:39:08,332
And there was a whole bit like Chinese origin now going to the US and the whole
competition between US and China for AI that I heard that there were some people that they

662
00:39:08,332 --> 00:39:08,893
didn't.

663
00:39:08,893 --> 00:39:15,072
For the Chinese, the Chinese government, weren't happy with this transition because they
felt like one of the big players now is going to the US.

664
00:39:15,072 --> 00:39:16,548
m

665
00:39:16,548 --> 00:39:25,688
Yeah, I'm not sure if you have any thoughts on that as well, if you feel like because I
mean, Menace, think even because when I think of the Chinese AI, it is a lot open source.

666
00:39:25,688 --> 00:39:27,128
It is about a lot low cost.

667
00:39:27,128 --> 00:39:27,388
Right.

668
00:39:27,388 --> 00:39:30,048
And Menace first, I mean, they are in Singapore now.

669
00:39:30,048 --> 00:39:33,028
And also the it did feel a bit different.

670
00:39:33,028 --> 00:39:39,100
Like, honestly, when I when I saw about Menace, I thought they were in the US actually,
you know.

671
00:39:39,100 --> 00:39:40,600
It has Chinese roots indeed.

672
00:39:40,600 --> 00:39:41,611
They moved to Singapore.

673
00:39:41,611 --> 00:39:48,833
I think now with the acquisition of Meta, they will have to sever these roots with China,
even if they still exist.

674
00:39:48,833 --> 00:39:51,584
I don't think it will be a problem, to be honest.

675
00:39:51,584 --> 00:39:53,485
You know, maybe what could be a problem.

676
00:39:53,485 --> 00:39:55,047
And I heard this, yeah.

677
00:39:55,047 --> 00:39:58,001
I don't know what models Manus uses underneath, right?

678
00:39:58,001 --> 00:40:06,498
I don't think they have their own models, but like if they're using a Chinese model or
like a competitor, would the competitor still want them to like, for example, if Entropic,

679
00:40:06,498 --> 00:40:11,013
would Entropic be happy to power Manus if it means competing with themselves?

680
00:40:11,013 --> 00:40:12,315
Well, do they compete with themselves?

681
00:40:12,315 --> 00:40:13,617
Then they're just a provider, right?

682
00:40:13,617 --> 00:40:14,822
I mean, they make a lot of money.

683
00:40:14,822 --> 00:40:17,556
um

684
00:40:17,556 --> 00:40:19,077
what's the meta play here, right?

685
00:40:19,077 --> 00:40:20,489
What does meta want to do with?

686
00:40:20,489 --> 00:40:23,114
Because meta wants to do stuff with AI for sure, right?

687
00:40:23,114 --> 00:40:24,057
What exactly?

688
00:40:24,057 --> 00:40:28,561
has been investing a lot, like a lot, a lot, a lot in AI last years.

689
00:40:28,561 --> 00:40:33,827
haven't really, hasn't really paid off, got a lot of criticism the last years, the stock
price suffered a bit.

690
00:40:33,827 --> 00:40:36,531
Not last years, last months, stock price suffered a bit.

691
00:40:36,531 --> 00:40:40,206
Because they invest a lot and they like haven't really got that much to show for.

692
00:40:40,206 --> 00:40:50,997
I think what they really want to do is that to really go across all the meta services with
AI skills that you have indeed agents in WhatsApp for business or that you have agents in

693
00:40:50,997 --> 00:40:55,022
Instagram or like in all of that today is still very limited.

694
00:40:55,022 --> 00:41:01,278
Like in WhatsApp, I have like a very bad LLM assistant that like I will never use because
JetJPT is way better.

695
00:41:01,278 --> 00:41:02,511
If it would be as good as

696
00:41:02,511 --> 00:41:07,404
DeepSeek or SysGPT or whatever, like I would probably just use my quote unquote free
WhatsApp LLM, right?

697
00:41:07,404 --> 00:41:12,688
um There's also this context where you would like to have like WhatsApp agents.

698
00:41:12,688 --> 00:41:17,360
Like if you have WhatsApp for business, you have customers asking you stuff, like very
easily spin up a new agent.

699
00:41:17,360 --> 00:41:20,032
can imagine that the Manus can do something like that.

700
00:41:20,032 --> 00:41:28,148
Help with setting up advertisement across their advertisement platforms, which they do
have some AI capabilities, but let's be honest, it's really shitty.

701
00:41:28,148 --> 00:41:28,559
today.

702
00:41:28,559 --> 00:41:31,691
So I do think there are a lot of options there.

703
00:41:31,691 --> 00:41:34,412
Aside also even just from the talent that brings in, right?

704
00:41:34,412 --> 00:41:43,447
Like because there's also a lot of talent that's coming with with MENACE in the sense that
it's and that's different talent I think that that is today available at Facebook

705
00:41:43,447 --> 00:41:44,028
Research.

706
00:41:44,028 --> 00:41:46,802
or whatever the meta department is.

707
00:41:46,802 --> 00:41:48,584
Because it's really something that was built from scratch.

708
00:41:48,584 --> 00:41:52,640
It's more of a startup environment, even though it is doing 100 million in ARR these days.

709
00:41:52,640 --> 00:41:53,541
Yeah, that's true.

710
00:41:53,541 --> 00:41:54,021
That's true.

711
00:41:54,021 --> 00:41:56,572
It's one of those AI startups that exploded, right?

712
00:41:56,572 --> 00:41:59,093
Yeah, it's true.

713
00:41:59,093 --> 00:42:01,326
I'm also, yeah, because you mentioned Facebook research.

714
00:42:01,326 --> 00:42:03,497
And I was also thinking of Meta did have models, right?

715
00:42:03,497 --> 00:42:09,091
Like with Llama, they were the last AI provider, American AI provider that was open
source.

716
00:42:09,091 --> 00:42:09,442
Right.

717
00:42:09,442 --> 00:42:12,367
And now I think they kind of stopped, kind of.

718
00:42:12,367 --> 00:42:15,385
I actually just heard that Yelnek Kuhn, just found his new venture.

719
00:42:15,385 --> 00:42:16,795
So he was on the Facebook.

720
00:42:16,795 --> 00:42:18,668
But we can talk about it next week, maybe.

721
00:42:18,668 --> 00:42:29,281
So it's like, I'm not sure like, are these two, like, is this something still interesting
for Meta to work on their own foundation models or, I don't know.

722
00:42:29,281 --> 00:42:32,313
And how is this going to play with Manus somehow?

723
00:42:32,313 --> 00:42:36,261
Well, that's more of a strategic question.

724
00:42:36,261 --> 00:42:37,112
It's a good question.

725
00:42:37,112 --> 00:42:37,996
think if they...

726
00:42:37,996 --> 00:42:42,079
I think in the short run it probably doesn't make sense for them to have their own model.

727
00:42:42,079 --> 00:42:49,918
In the long run if they can have a model that is more or less comparable to something like
DeepSeek, it's probably way more cost efficient to have your own model.

728
00:42:49,918 --> 00:42:53,119
Yeah, yeah, yeah for sure for sure and

729
00:42:53,119 --> 00:42:56,845
what they want to do is really build something that current models don't do at all.

730
00:42:56,845 --> 00:43:02,755
Like I don't think they're there today for example in LM that is truly truly truly good at

731
00:43:02,755 --> 00:43:10,026
making good ad campaigns, which is like basically like the 80 % of revenue of Meta, right?

732
00:43:10,026 --> 00:43:12,977
Like maybe there's a strategic opening there.

733
00:43:12,977 --> 00:43:22,279
But what we've seen is that the llama models that they made in the past or that they
published at least in the past, maybe they still using them in house new versions, is that

734
00:43:22,279 --> 00:43:26,250
they are very good, but they're always slightly behind in the state of the art.

735
00:43:26,250 --> 00:43:27,612
Yeah, yeah, that's true.

736
00:43:27,612 --> 00:43:28,122
That's true.

737
00:43:28,122 --> 00:43:32,033
I do think they could probably specialize some of these different models, right?

738
00:43:32,033 --> 00:43:32,468
It's true.

739
00:43:32,468 --> 00:43:38,795
Maybe one last thing as well, just to complete the story that Manus will still be running
independently, right?

740
00:43:38,795 --> 00:43:40,347
least that's what they say right now.

741
00:43:40,347 --> 00:43:44,290
So, which I guess it's also if they have good revenue, it also makes sense for Meta,
right?

742
00:43:44,290 --> 00:43:47,474
To still keep it a separate product, but like still...

743
00:43:47,474 --> 00:43:52,398
leverage from the brains and all these things and trying to integrate into Facebook
products.

744
00:43:52,398 --> 00:43:52,991
So yeah.

745
00:43:52,991 --> 00:43:57,473
Yeah, there is this meme, actually kind of fun.

746
00:43:57,473 --> 00:44:05,724
You have now Alexander Wang that is leading, I think he's the chief AI officer at Meta.

747
00:44:05,724 --> 00:44:12,542
And he came from Scale AI, like basically a labeling, AI labeling platform that Meta
acquired.

748
00:44:12,542 --> 00:44:15,354
And there is this meme that...

749
00:44:15,354 --> 00:44:20,209
Zuckerberg sends a text message to Alexander Wang saying we should buy Menus.

750
00:44:20,209 --> 00:44:23,316
Alexander Wang sends back a Don't Deal Boss.

751
00:44:23,316 --> 00:44:23,917
We bought it.

752
00:44:23,917 --> 00:44:26,932
Zuckerberg sends back like did we buy pro or premium?

753
00:44:26,932 --> 00:44:27,517
Whoopsie.

754
00:44:27,517 --> 00:44:34,822
I think it's funny also like sometimes I also you said this I thought of like my
interactions with the AI sometimes right like I say do this and then they just do

755
00:44:34,822 --> 00:44:46,260
completely that's like what the fuck but like technically they are they are following what
I said right it's like ah okay context yeah indeed it matters that is it for our regular

756
00:44:46,260 --> 00:44:51,124
topics we have two two tidbits I guess small things um

757
00:44:51,124 --> 00:44:55,489
One is a project that I came to my attention via colleagues, it's called Trivy.

758
00:44:55,489 --> 00:44:57,346
I don't know if you know about Trivy, Bart?

759
00:44:57,346 --> 00:45:00,323
No, I quickly glance at it when you send me the link, yeah.

760
00:45:00,323 --> 00:45:09,465
Yeah, basically the about on the GitHub is find vulnerabilities, misconfiguration,
secrets, S-BOM, not sure what it is, in containers, Kubernetes, code repos, clouds and

761
00:45:09,465 --> 00:45:09,907
more.

762
00:45:09,907 --> 00:45:17,491
The discussion was, there was a blog post from, I think, Michael Kennedy from TalkPython
about how to verify PIP dependencies.

763
00:45:17,491 --> 00:45:22,967
So he's talking like when you're using LLM, sometimes install stuff, but sometimes the
stuff is actually malicious, right?

764
00:45:22,967 --> 00:45:23,587
I think...

765
00:45:23,587 --> 00:45:28,125
Nowadays it's less of an issue, but I think earlier when the models weren't really good,
this was a big problem.

766
00:45:28,125 --> 00:45:32,178
So like you would install requests instead of requests, right?

767
00:45:32,556 --> 00:45:34,520
And a lot of vulnerabilities like this.

768
00:45:34,520 --> 00:45:36,211
So it was also saying like how you can...

769
00:45:36,211 --> 00:45:39,159
still happens is installing alter dependencies,

770
00:45:39,159 --> 00:45:40,130
Yeah, indeed.

771
00:45:40,130 --> 00:45:45,821
So if there's a dependency that there was a vulnerability maybe on their on its
dependencies, right?

772
00:45:45,821 --> 00:45:49,101
And then you want to use the latest latest one as well.

773
00:45:49,101 --> 00:45:53,281
So there's also this thing like if there's a new version, do you also want to jump on it?

774
00:45:53,281 --> 00:45:54,241
Do you want to wait a bit?

775
00:45:54,241 --> 00:45:57,441
I think that's also something that they talk on the in the article, right?

776
00:45:57,441 --> 00:46:04,301
Like maybe you don't want to maybe you want to a few weeks before you install the
dependency and just make sure there are no vulnerabilities if you have like, yeah,

777
00:46:04,301 --> 00:46:05,453
depending on the project, etc.

778
00:46:05,453 --> 00:46:14,517
So there was a bit of back and forth on like how to do this like there's a PIP check or
something and then there was like a UV check so there was a bit of a discussion internally

779
00:46:14,517 --> 00:46:22,216
on slack and Someone shared this one trivia, which I wasn't familiar and the reason why
they said this one is because it's not It's not only for Python.

780
00:46:22,216 --> 00:46:28,864
It's actually like if you have no JS if you have Docker whatever like it actually scans
everything it gives you a comprehensive view of

781
00:46:28,864 --> 00:46:31,000
What are the vulnerabilities that you may have?

782
00:46:31,000 --> 00:46:34,913
So something that could be interesting, especially on the age of AI.

783
00:46:34,913 --> 00:46:36,413
Yeah, could be.

784
00:46:36,637 --> 00:46:41,593
maybe try to run it on one of my AI generated projects, just to see what comes out.

785
00:46:41,593 --> 00:46:46,857
Yeah, conclusion is just like, drop this project, never do anything with it.

786
00:46:46,857 --> 00:46:47,637
exactly.

787
00:46:47,637 --> 00:46:49,608
It's like, oh my God, what the fuck are you doing?

788
00:46:49,608 --> 00:46:50,708
So that's one.

789
00:46:50,708 --> 00:46:56,948
And the second tidbit is something that came 24th of December.

790
00:46:56,948 --> 00:47:05,339
So just on Christmas Eve that Microsoft wants to replace its entire C and C++ code base,
perhaps by 2030.

791
00:47:05,339 --> 00:47:07,930
And they want to migrate from this to

792
00:47:07,930 --> 00:47:08,400
rust.

793
00:47:08,400 --> 00:47:13,292
um Yeah, it is.

794
00:47:13,292 --> 00:47:17,525
But I think so I saw somewhere that it is a bit of a research thing.

795
00:47:17,525 --> 00:47:18,917
Like they're not really

796
00:47:18,917 --> 00:47:20,237
They're exploring if it's doable.

797
00:47:20,237 --> 00:47:21,823
m

798
00:47:21,823 --> 00:47:22,443
doable.

799
00:47:22,443 --> 00:47:25,376
Like again, they say that, mean, so why would you do this?

800
00:47:25,376 --> 00:47:28,550
Maybe, since C plus plus and rust, they're fast, right?

801
00:47:28,550 --> 00:47:35,509
But the difference, the main difference and why they're saying they want to do this is
that in rust, you have a memory safety, right?

802
00:47:35,509 --> 00:47:41,824
In rust, if you, if you reference a value, you have more guarantees that the value is not
going to throw an error basically, right?

803
00:47:41,824 --> 00:47:45,369
So it's very like memory management and all these things, very low level.

804
00:47:45,369 --> 00:47:52,894
And they say apparently like you can actually resolve a lot of issues or a lot of bugs or
like your code can be more robust if you make this change.

805
00:47:52,894 --> 00:47:56,295
They want to use AI to migrate to this.

806
00:47:56,295 --> 00:48:07,698
So I guess they would just ask AI, maybe the update on December 29th, to let them use that
hunt updated it's supposed to stay that Windows is not being rewritten in Rust with AI.

807
00:48:07,698 --> 00:48:09,342
So it's not the goal there.

808
00:48:09,342 --> 00:48:13,502
research project, so it's more of question that we're trying to answer.

809
00:48:13,502 --> 00:48:16,476
Could it be rewritten in Rust with AI?

810
00:48:16,476 --> 00:48:18,547
So my team's project is a research project he added.

811
00:48:18,547 --> 00:48:21,598
We're building tech to make migration from language to language possible.

812
00:48:21,598 --> 00:48:24,120
The intent of my post was to find like-minded engineers.

813
00:48:24,120 --> 00:48:24,991
Okay, yada, yada.

814
00:48:24,991 --> 00:48:27,132
So it is a bit like, is it possible?

815
00:48:27,132 --> 00:48:28,232
How long would it take?

816
00:48:28,232 --> 00:48:29,302
Is it something feasible?

817
00:48:29,302 --> 00:48:30,554
What are the benefits?

818
00:48:30,554 --> 00:48:33,365
But I think it could be, I mean, I think it's interesting.

819
00:48:33,365 --> 00:48:37,849
I would never do it probably or seriously do it, but I think it's interesting.

820
00:48:37,849 --> 00:48:41,860
And I do see, like, I think Rust is also in the Linux kernel, I think already.

821
00:48:43,333 --> 00:48:45,987
So it is gaining traction there as well.

822
00:48:45,987 --> 00:48:46,877
What do think of this Bart?

823
00:48:46,877 --> 00:48:48,580
Would you be happy to work on a project like this?

824
00:48:48,580 --> 00:48:49,613
Even if it's research.

825
00:48:49,613 --> 00:48:51,280
To have to work on project like this.

826
00:48:51,280 --> 00:48:52,537
No, I don't think that's for me.

827
00:48:52,537 --> 00:48:54,268
I don't think it is for me either.

828
00:48:54,268 --> 00:49:00,710
What I did try to find a bit is like, okay, it's a research project, I get it.

829
00:49:00,710 --> 00:49:05,485
But like, what are the research questions, quote, quote, that you're trying to answer?

830
00:49:05,485 --> 00:49:06,385
Like, do you have metrics?

831
00:49:06,385 --> 00:49:07,236
Do you have benchmarks?

832
00:49:07,236 --> 00:49:10,657
Like, you're gonna do this and how are you gonna know if it was a success or not?

833
00:49:10,657 --> 00:49:11,968
What are the things you're trying to learn?

834
00:49:11,968 --> 00:49:14,339
And I imagine that they have this, but I couldn't find it.

835
00:49:14,339 --> 00:49:16,641
Yeah, such a huge project as well.

836
00:49:16,783 --> 00:49:17,756
Yeah, indeed.

837
00:49:17,756 --> 00:49:19,858
To be seen, to be seen.

838
00:49:19,858 --> 00:49:21,191
And I think that's it for today.

839
00:49:21,191 --> 00:49:23,997
Yeah, thanks a lot, Marilla, for joining me.

840
00:49:23,997 --> 00:49:26,210
No, thank you Bart for being here.

841
00:49:26,210 --> 00:49:27,902
Thanks everyone for listening.

842
00:49:27,902 --> 00:49:29,243
Update over the Christmas.

843
00:49:29,243 --> 00:49:34,938
think we have like what 7k subscribers.

844
00:49:34,938 --> 00:49:35,378
7.5k.

845
00:49:35,378 --> 00:49:35,939
Yes.

846
00:49:35,939 --> 00:49:39,472
So not bad at all.

847
00:49:39,472 --> 00:49:41,533
So thanks everyone for subscribing.

848
00:49:41,533 --> 00:49:43,295
Thanks everyone for listening.

849
00:49:43,295 --> 00:49:44,556
Tell your friends, family.

850
00:49:44,556 --> 00:49:46,057
We got some love as well on LinkedIn.

851
00:49:46,057 --> 00:49:47,238
So thank you for that.

852
00:49:47,238 --> 00:49:49,140
Really appreciate it.

853
00:49:49,140 --> 00:49:49,552
Yeah.

854
00:49:49,552 --> 00:49:51,782
Feel free to subscribe to our newsletter as well.

855
00:49:51,782 --> 00:49:53,820
Thanks everyone and I'll see you all next week.

856
00:49:53,820 --> 00:49:54,689
See you all next week.

857
00:49:54,689 --> 00:49:55,350
Ciao!

858
00:49:55,350 --> 00:49:55,852
out.