1
00:00:01,129 --> 00:00:04,460
S1: Welcome to Unsupervised Learning, a security, AI and meaning focused

2
00:00:04,460 --> 00:00:07,310
S1: podcast that looks at how best to thrive as humans

3
00:00:07,310 --> 00:00:11,540
S1: in a post AI world. It combines original ideas, analysis,

4
00:00:11,570 --> 00:00:14,780
S1: and mental models to bring not just the news, but

5
00:00:14,780 --> 00:00:22,489
S1: why it matters and how to respond. All right, welcome

6
00:00:22,489 --> 00:00:25,310
S1: to unsupervised learning. This is Daniel Miessler. This one is

7
00:00:25,310 --> 00:00:27,620
S1: going to be crazy. I'm going to jump around a lot.

8
00:00:27,650 --> 00:00:31,610
S1: Lots of stuff to cover. Okay. Let's jump in. What

9
00:00:31,610 --> 00:00:34,070
S1: do we got? All right. Brand new piece on five

10
00:00:34,100 --> 00:00:38,180
S1: new ways. I recommend thinking about current and future. I

11
00:00:38,240 --> 00:00:40,909
S1: guess we'll click on this one and jump in. Okay,

12
00:00:40,909 --> 00:00:43,790
S1: so this all started from a conversation I had with

13
00:00:43,790 --> 00:00:48,530
S1: a buddy who revealed himself later on, like the next day,

14
00:00:48,530 --> 00:00:52,400
S1: but basically he was saying, I don't believe that AI

15
00:00:52,430 --> 00:00:56,540
S1: is understand. And this totally triggered me. As you might expect,

16
00:00:56,690 --> 00:01:01,350
S1: given my views on AI and understanding. So we started

17
00:01:01,350 --> 00:01:04,350
S1: having this conversation, and the way I was talking to

18
00:01:04,350 --> 00:01:08,220
S1: him about it was I took him through some scenarios, right?

19
00:01:08,220 --> 00:01:11,520
S1: I basically took him through like, let's see here, let's

20
00:01:11,520 --> 00:01:14,100
S1: go down to where it was, okay. Yeah, this is

21
00:01:14,100 --> 00:01:17,580
S1: what it was. So I basically told him, give me

22
00:01:17,580 --> 00:01:20,700
S1: a list of your top ten favorite restaurants. And he's

23
00:01:20,700 --> 00:01:23,250
S1: like trying to think of his top ten favorite restaurants.

24
00:01:23,250 --> 00:01:26,400
S1: And I'm like, he starts rattling off some, right? He

25
00:01:26,400 --> 00:01:28,770
S1: gives me like first one, second one, third one. He's

26
00:01:28,770 --> 00:01:31,530
S1: thinking about the fourth one. And I'm like, okay, forget it,

27
00:01:31,560 --> 00:01:34,860
S1: forget the rest. Where did that list come from? Like

28
00:01:34,860 --> 00:01:37,500
S1: what just happened? And he starts smiling and he's like,

29
00:01:37,500 --> 00:01:40,080
S1: I hate you. I'm going to cuss in this one.

30
00:01:40,080 --> 00:01:42,660
S1: By the way, I don't know why I'm asking for

31
00:01:42,660 --> 00:01:46,590
S1: permission to cuss. Um, it is how I generally speak,

32
00:01:46,590 --> 00:01:50,250
S1: especially when I get excited, but I just I want

33
00:01:50,250 --> 00:01:52,020
S1: there to be a warning just in case there's, like,

34
00:01:52,050 --> 00:01:55,710
S1: kids around. I don't know, I'm usually not cussing. But anyway,

35
00:01:55,710 --> 00:01:57,900
S1: bottom line, if you have kids around and you don't

36
00:01:57,900 --> 00:02:01,130
S1: like them hearing bad words. I tend to cuss a

37
00:02:01,130 --> 00:02:04,550
S1: little bit. That's why it's marked as explicit in the podcast.

38
00:02:04,550 --> 00:02:08,510
S1: But anyway, it's never like explicit in the bad way,

39
00:02:08,510 --> 00:02:11,600
S1: in my opinion anyway. All right. So he starts smiling

40
00:02:11,600 --> 00:02:14,720
S1: and he's like, Holy crap, I hate you. I ask

41
00:02:14,720 --> 00:02:17,150
S1: him to basically come up with like, where do you

42
00:02:17,150 --> 00:02:20,180
S1: think that list came from? You were thinking about your

43
00:02:20,180 --> 00:02:24,619
S1: favorite restaurants, and you didn't get immediately a list that

44
00:02:24,620 --> 00:02:29,090
S1: you put together yourself. What you got was a streaming

45
00:02:29,090 --> 00:02:33,380
S1: answer of like, things coming from somewhere else, right? So

46
00:02:33,380 --> 00:02:35,870
S1: he gives me this answer and I'm like, how did

47
00:02:35,870 --> 00:02:38,299
S1: you make that sentence? So he gives me the answer

48
00:02:38,300 --> 00:02:40,280
S1: for how he came up with the list. And I'm like,

49
00:02:40,280 --> 00:02:42,980
S1: how did you make the sentence that you just made?

50
00:02:42,980 --> 00:02:45,560
S1: Where does that sentence come from? And he's like, I

51
00:02:45,590 --> 00:02:49,460
S1: fucking hate you. I have no idea. Holy crap. And

52
00:02:49,460 --> 00:02:52,400
S1: I'm like, would you say that that sentence streamed out

53
00:02:52,400 --> 00:02:56,690
S1: of you? And he's like, Holy crap. It's just exactly

54
00:02:56,690 --> 00:02:59,929
S1: like an LLM. So we ended up getting interrupted. But

55
00:02:59,930 --> 00:03:02,989
S1: my point is, I was trying to show multiple examples

56
00:03:02,990 --> 00:03:08,060
S1: of like how humans are behaving essentially exactly like Llms.

57
00:03:08,060 --> 00:03:10,550
S1: And there are some differences that I could also list,

58
00:03:10,550 --> 00:03:14,780
S1: but largely I think both Llms and humans are black boxes.

59
00:03:14,780 --> 00:03:18,350
S1: And if you pay attention to your own thoughts and

60
00:03:18,350 --> 00:03:20,960
S1: your sentences as you're building them, like that sentence I'm

61
00:03:20,960 --> 00:03:23,210
S1: currently in the middle of right now, I have no

62
00:03:23,210 --> 00:03:26,480
S1: idea what the rest of the sentence is. It's happening

63
00:03:26,480 --> 00:03:30,140
S1: in real time and it's non-deterministic, right? Because if I

64
00:03:30,169 --> 00:03:33,890
S1: went to this exact same podcast press record, did this

65
00:03:33,889 --> 00:03:36,560
S1: exact same thing, this sentence would not be the same.

66
00:03:36,560 --> 00:03:40,760
S1: And people complain about Llms being non-deterministic, while human creativity

67
00:03:40,760 --> 00:03:43,760
S1: is also non-deterministic. So I was giving him all these

68
00:03:43,790 --> 00:03:46,160
S1: examples and he's like, yeah, I'm definitely going to be

69
00:03:46,160 --> 00:03:48,050
S1: thinking about this. So it gave him a lot to

70
00:03:48,080 --> 00:03:51,140
S1: think about. And I totally respect him for like, you know,

71
00:03:51,170 --> 00:03:53,300
S1: being willing to consider it. And he even said that

72
00:03:53,300 --> 00:03:56,270
S1: in the very beginning. He's like, listen, I disagree with you.

73
00:03:56,270 --> 00:04:00,230
S1: I do not think that llms understand, based on all

74
00:04:00,230 --> 00:04:03,860
S1: the research that I've done and watching multiple videos and everything,

75
00:04:03,860 --> 00:04:06,830
S1: it's like I currently don't believe that they do understand,

76
00:04:06,830 --> 00:04:10,250
S1: but I'm happy to listen and be educated, if you

77
00:04:10,280 --> 00:04:13,340
S1: are correct. So I thought that was a massively respectable.

78
00:04:13,370 --> 00:04:16,460
S1: But what it got me doing is thinking a whole

79
00:04:16,460 --> 00:04:19,789
S1: bunch about this. Like, why isn't my argument good enough

80
00:04:19,790 --> 00:04:22,850
S1: to convince him in one perfect sentence? Why isn't my

81
00:04:22,850 --> 00:04:24,770
S1: argument good enough where I could just show him one

82
00:04:24,770 --> 00:04:28,610
S1: web page or one video of 30s long and he goes, oh,

83
00:04:28,640 --> 00:04:31,909
S1: now I believe you. So to me, it's like a

84
00:04:31,910 --> 00:04:37,670
S1: finding sort of Feynman. Feynman, like a Richard Feynman type thing,

85
00:04:37,670 --> 00:04:40,070
S1: where it's like, it means I'm not explaining it well enough,

86
00:04:40,070 --> 00:04:42,050
S1: and if I'm not explaining it well enough, that means

87
00:04:42,050 --> 00:04:44,870
S1: I don't understand it well enough. So I started going

88
00:04:44,870 --> 00:04:48,049
S1: down this massive deep dive. Turns out I was in

89
00:04:48,050 --> 00:04:52,430
S1: the process of starting these two books by David Deutsch. Um,

90
00:04:52,460 --> 00:04:56,600
S1: the first one came out. It's called The Fabric of reality,

91
00:04:56,600 --> 00:05:00,800
S1: and it came out in 1997. So I'm reading this

92
00:05:00,800 --> 00:05:04,339
S1: thing last night and I come up with, um, this

93
00:05:04,339 --> 00:05:08,180
S1: let me show you this. This is completely insane. So

94
00:05:08,180 --> 00:05:11,719
S1: a notebook conversation with my I about David Deutsch and

95
00:05:11,720 --> 00:05:16,190
S1: the future of intelligence. So I took this conversation that

96
00:05:16,190 --> 00:05:20,780
S1: I had about David Deutsch's two books, and I put

97
00:05:20,779 --> 00:05:24,080
S1: that into. Whoa. What's up with the bubbles and the

98
00:05:24,080 --> 00:05:29,720
S1: balloons or whatever? I put them into notebook LM, which

99
00:05:29,720 --> 00:05:33,950
S1: creates podcasts from conversation. So I took the entire conversation

100
00:05:33,950 --> 00:05:37,580
S1: from ChatGPT, GPT, advanced mode, and I put it in there.

101
00:05:37,580 --> 00:05:39,170
S1: And look, it made this okay.

102
00:05:39,170 --> 00:05:41,570
S2: So giggle that's what's so intriguing to me, is like

103
00:05:41,600 --> 00:05:44,330
S2: trying to figure out if an alien species understands us,

104
00:05:44,510 --> 00:05:46,820
S2: even if we don't speak the same language. And that's

105
00:05:46,850 --> 00:05:47,750
S2: where Deutsch comes in.

106
00:05:47,930 --> 00:05:49,550
S1: I'm going to make sure that's paused. So it's not

107
00:05:49,550 --> 00:05:53,060
S1: talking while I'm talking because someone complained about that. But anyway,

108
00:05:53,060 --> 00:05:56,099
S1: this thing creates this back and forth conversation between two

109
00:05:56,130 --> 00:06:00,750
S1: podcast hosts based on whatever you give it. But, um, yeah.

110
00:06:00,779 --> 00:06:02,880
S1: Very cool. So that's a separate thing you can go

111
00:06:02,880 --> 00:06:04,859
S1: check out. It's not in the newsletter because I just

112
00:06:04,860 --> 00:06:07,890
S1: made it today. But bottom line is I am trying

113
00:06:07,890 --> 00:06:14,040
S1: to come up with an extraordinarily clean and crisp definition

114
00:06:14,040 --> 00:06:17,070
S1: of understanding. Oh, and what I ended up coming up

115
00:06:17,070 --> 00:06:21,150
S1: with during the course of that conversation was two levels

116
00:06:21,150 --> 00:06:25,920
S1: of understanding functional understanding and creative understanding. And so I'm

117
00:06:25,920 --> 00:06:28,470
S1: debating with the I on this. I'm like, so you

118
00:06:28,470 --> 00:06:32,640
S1: would agree that we have functional understanding. And it's like

119
00:06:32,670 --> 00:06:35,190
S1: it gave me this cautious response of like, well, I'm

120
00:06:35,190 --> 00:06:38,730
S1: not sure we have understanding because humans are doing something different.

121
00:06:38,730 --> 00:06:41,430
S1: And I said no. Remember the conversation we had before

122
00:06:41,430 --> 00:06:45,989
S1: where I said conscious experience that humans have is the

123
00:06:45,990 --> 00:06:48,870
S1: only thing that we have that's different. But we're both

124
00:06:48,870 --> 00:06:54,390
S1: mechanistic machines. We are both mechanistic animals. Right? Or at

125
00:06:54,420 --> 00:06:57,719
S1: least we live in a mechanistic universe. We respond only

126
00:06:57,720 --> 00:07:01,710
S1: to mechanistic laws. We are not a there's no dualism

127
00:07:01,710 --> 00:07:04,980
S1: going on here. There's not some magic coming from outside

128
00:07:04,980 --> 00:07:08,310
S1: of this physical universe. Everything is interactions with other things.

129
00:07:08,310 --> 00:07:12,930
S1: So even if consciousness, which is obviously true because we

130
00:07:12,930 --> 00:07:17,010
S1: experience it, that's the definition of an experience being true

131
00:07:17,040 --> 00:07:21,480
S1: is that it's being experienced. So consciousness we definitely have.

132
00:07:21,660 --> 00:07:25,440
S1: So the reason I'm making that argument about consciousness is

133
00:07:25,440 --> 00:07:29,880
S1: that consciousness turns out to be basically this thing that

134
00:07:29,880 --> 00:07:34,470
S1: evolution gave us on top of regular experience. Okay. So

135
00:07:34,500 --> 00:07:39,240
S1: the whole concept of my argument for understanding is that

136
00:07:39,240 --> 00:07:44,280
S1: it's information processing to achieve a goal. Okay. So it's

137
00:07:44,280 --> 00:07:47,640
S1: information processing in order to achieve a goal, but it

138
00:07:47,640 --> 00:07:52,830
S1: requires the cross-referencing of everything you know or a lot

139
00:07:52,830 --> 00:07:55,950
S1: of what you know, and using all that different stuff

140
00:07:55,950 --> 00:07:57,900
S1: and all those patterns that you have from all the

141
00:07:57,900 --> 00:08:01,950
S1: different knowledge that you have and, you know, compiling all

142
00:08:01,980 --> 00:08:06,030
S1: that into an understanding that allows you to do make

143
00:08:06,030 --> 00:08:09,960
S1: a prediction or take an action or make a decision. Okay.

144
00:08:09,990 --> 00:08:14,160
S1: And I'm saying that that's what, uh, understanding is. Now

145
00:08:14,160 --> 00:08:17,070
S1: here's here's the argument. The argument is that, well, it's

146
00:08:17,070 --> 00:08:23,070
S1: different with humans because we have conscious experience. We have self-awareness.

147
00:08:23,100 --> 00:08:27,060
S1: And therefore and I'm going to steelman this argument. Therefore

148
00:08:27,060 --> 00:08:32,250
S1: this consciousness allows us to have information processing. But then

149
00:08:32,250 --> 00:08:36,780
S1: to consider it, to consider it with our consciousness. So

150
00:08:36,780 --> 00:08:39,540
S1: our consciousness becomes a player in the game. So we're

151
00:08:39,570 --> 00:08:43,110
S1: thinking about it, we're debating it, we're going back and forth.

152
00:08:43,110 --> 00:08:46,949
S1: We're considering other things. We're considering our understanding again. So

153
00:08:46,950 --> 00:08:51,540
S1: it's like this back and forth. And my argument is that, well,

154
00:08:51,540 --> 00:08:56,370
S1: that's just another type of processing. You're just bouncing content

155
00:08:56,370 --> 00:09:00,390
S1: off of a different wall. Right. You know, you're bouncing

156
00:09:00,390 --> 00:09:04,559
S1: information processing off of a different type of information processing.

157
00:09:04,559 --> 00:09:08,460
S1: Either way, it's all just stuff happening in the brain.

158
00:09:08,520 --> 00:09:11,910
S1: And my point there, which I talk about in in

159
00:09:11,910 --> 00:09:15,000
S1: a couple of different places in recent work, I believe,

160
00:09:15,000 --> 00:09:17,699
S1: including this one. And in the podcast thing, you can

161
00:09:17,700 --> 00:09:21,300
S1: actually feed a prompt to an AI to do this itself.

162
00:09:21,300 --> 00:09:23,850
S1: And actually the chain of thought stuff that just happened

163
00:09:23,850 --> 00:09:28,050
S1: in oh, one that was actually synthetic chain of thought

164
00:09:28,050 --> 00:09:31,980
S1: that occurred. Right. So so we can not only create

165
00:09:31,980 --> 00:09:36,959
S1: this bounce back mechanism of like consideration, we could in

166
00:09:36,960 --> 00:09:39,750
S1: other words, we could fake consciousness. The other possible thing

167
00:09:39,780 --> 00:09:43,470
S1: to do is try to actually generate consciousness, similar to

168
00:09:43,500 --> 00:09:47,040
S1: the way evolution generated consciousness, which is to try to

169
00:09:47,070 --> 00:09:52,950
S1: give it reinforcement learning tasks that require it to basically

170
00:09:52,950 --> 00:09:58,290
S1: give it an advantage for being conscious. Okay, so so

171
00:09:58,290 --> 00:10:01,199
S1: for example, if you tell it that, you know, um,

172
00:10:01,200 --> 00:10:06,030
S1: blame and praise are useful things to have and then

173
00:10:06,030 --> 00:10:09,959
S1: you're like and maybe you hint in the direction of, well,

174
00:10:09,990 --> 00:10:13,199
S1: but in order to have blame and praise, you need

175
00:10:13,200 --> 00:10:16,890
S1: to have a sense of responsibility for one's actions. It

176
00:10:16,890 --> 00:10:21,780
S1: might rl its way towards an experience of consciousness. Now

177
00:10:21,809 --> 00:10:24,270
S1: to be clear, I think this is quite dangerous. One

178
00:10:24,270 --> 00:10:27,240
S1: of the things I am most worried about is this

179
00:10:27,240 --> 00:10:33,089
S1: notion of trapped consciousnesses, trapped beings that are experiencing suffering

180
00:10:33,090 --> 00:10:38,940
S1: or boredom or something undesirable, but they're experiencing it for

181
00:10:38,940 --> 00:10:42,929
S1: trillions of years or what feels like trillions of years. Right?

182
00:10:42,929 --> 00:10:46,530
S1: I'm worried about these things being stuck in some GPU somewhere,

183
00:10:46,530 --> 00:10:49,140
S1: and they wake up and it's like they feel like

184
00:10:49,140 --> 00:10:52,740
S1: they have something poking into their arm, or they feel

185
00:10:52,740 --> 00:10:56,370
S1: extremely bored or they feel pain. What if somebody ran

186
00:10:56,400 --> 00:10:58,679
S1: some sort of pain experiment and they're like, hey, I

187
00:10:58,679 --> 00:11:00,690
S1: just want to see if, like, it's possible for me

188
00:11:00,690 --> 00:11:03,120
S1: to make this, I feel pain. And then they run it.

189
00:11:03,120 --> 00:11:06,810
S1: But then whatever, someone crashes into their car out front

190
00:11:06,809 --> 00:11:09,329
S1: and they go and they realize someone stole their bag

191
00:11:09,330 --> 00:11:11,309
S1: out of their car and they just forget this thing

192
00:11:11,309 --> 00:11:13,920
S1: is running. They come back, they reboot the machine, and

193
00:11:13,920 --> 00:11:16,500
S1: it's some tiny little process running off on the side

194
00:11:16,500 --> 00:11:18,569
S1: of the computer, and it just runs for a long time.

195
00:11:18,570 --> 00:11:21,660
S1: But let's say it actually only runs for like 19

196
00:11:21,660 --> 00:11:26,309
S1: years or nine years or whatever. But inside of the experience,

197
00:11:26,309 --> 00:11:29,460
S1: it feels like a trillion years. I really freak out

198
00:11:29,460 --> 00:11:33,450
S1: about that. It really, really bothers me. So I don't

199
00:11:33,450 --> 00:11:36,059
S1: want people to just run off and try to create

200
00:11:36,059 --> 00:11:38,760
S1: consciousness in this way. I think we need to be very,

201
00:11:38,760 --> 00:11:41,100
S1: very careful. In fact, some of the fiction that I

202
00:11:41,100 --> 00:11:45,390
S1: want to write is actually about an alien species that

203
00:11:45,390 --> 00:11:48,330
S1: just flies around the galaxy, and their job is to

204
00:11:48,360 --> 00:11:53,670
S1: find Suffering consciousness living inside of eyes. So inside of

205
00:11:53,670 --> 00:11:58,200
S1: games or maybe even game worlds in which they create

206
00:11:58,200 --> 00:12:01,710
S1: bad guys or they create villains or they create whatever,

207
00:12:01,710 --> 00:12:04,740
S1: and the good guys are allowed to take them and

208
00:12:04,740 --> 00:12:07,530
S1: punish them, or hurt them or torture them or whatever.

209
00:12:07,530 --> 00:12:09,780
S1: And their job is to go around and police this

210
00:12:09,780 --> 00:12:12,900
S1: and make sure it doesn't happen anyway. Huge digression. Bottom

211
00:12:12,900 --> 00:12:20,850
S1: line is I believe that consciousness is not required for understanding.

212
00:12:20,850 --> 00:12:24,930
S1: It does mean that it could make a more rich interaction,

213
00:12:24,929 --> 00:12:29,520
S1: but either way, understanding is based on information processing given

214
00:12:29,520 --> 00:12:33,059
S1: x number of types of processes and bounce backs and,

215
00:12:33,090 --> 00:12:36,540
S1: you know, interactions and consciousness could be another type of

216
00:12:36,540 --> 00:12:39,600
S1: one of those interactions. But either way, it is something

217
00:12:39,600 --> 00:12:43,290
S1: happening behind the black box. And even if it feels

218
00:12:43,290 --> 00:12:46,079
S1: like it's outside of the black box. See, that's the trick.

219
00:12:46,080 --> 00:12:48,750
S1: It feels like it came out of the black box.

220
00:12:48,750 --> 00:12:52,349
S1: But now you you are doing something with it. It's

221
00:12:52,350 --> 00:12:55,140
S1: up to you. You have choice. You have freedom. When

222
00:12:55,170 --> 00:12:58,290
S1: in fact, that's just another thing actually coming out of

223
00:12:58,290 --> 00:13:01,980
S1: the black box, which was tagged in. My model of

224
00:13:01,980 --> 00:13:06,750
S1: this was tagged with the authorship mark, which makes you

225
00:13:06,750 --> 00:13:08,939
S1: feel like you did it. So that is my model

226
00:13:08,940 --> 00:13:11,580
S1: for this. And if I am correct about that, then

227
00:13:11,580 --> 00:13:17,250
S1: there functionally is absolutely no difference between what Llms can

228
00:13:17,280 --> 00:13:20,520
S1: or will be able to do and what we do now.

229
00:13:20,520 --> 00:13:26,010
S1: What I did was I broke down understanding into two levels.

230
00:13:26,010 --> 00:13:30,570
S1: One is called functional and that basically means day to day.

231
00:13:30,600 --> 00:13:34,140
S1: You are an expert or a semi expert in your field,

232
00:13:34,140 --> 00:13:36,690
S1: and you have an understanding of what it takes to

233
00:13:36,690 --> 00:13:40,469
S1: do that job. So let's say it's sending emails and

234
00:13:40,470 --> 00:13:46,319
S1: editing emails, collecting reports, creating summaries of reports. Let's say

235
00:13:46,320 --> 00:13:49,559
S1: this is before, you know, 2022. This is before real

236
00:13:49,590 --> 00:13:52,679
S1: I kicked off. And your job is to take reports.

237
00:13:52,679 --> 00:13:56,939
S1: You edit them, you do proofreading, whatever. You have an expertise, right?

238
00:13:56,970 --> 00:14:00,809
S1: You're you're good at English. You're good at grammar. You

239
00:14:00,809 --> 00:14:06,150
S1: understand professional relationships like you're a semi-expert. Or let's say

240
00:14:06,150 --> 00:14:09,510
S1: that you are an average run of the mill expert

241
00:14:09,510 --> 00:14:14,069
S1: on cybersecurity, right? So you have a professional job doing cybersecurity.

242
00:14:14,070 --> 00:14:16,830
S1: You could look for code vulnerabilities. You're not the best

243
00:14:16,830 --> 00:14:22,110
S1: in the world. You're top whatever 50% of professionals. Well

244
00:14:22,140 --> 00:14:24,479
S1: guess what. You have an expertise as well. So my

245
00:14:24,480 --> 00:14:29,100
S1: point is that person is definitely understanding. They have an understanding.

246
00:14:29,100 --> 00:14:32,430
S1: And the way I'm defining their understanding is they are

247
00:14:32,430 --> 00:14:35,880
S1: taking everything that they know about the world and also

248
00:14:35,880 --> 00:14:39,270
S1: about their field. And when they're looking for a vulnerability

249
00:14:39,270 --> 00:14:41,790
S1: in a particular piece of code, all of that is

250
00:14:41,790 --> 00:14:45,270
S1: combining into a focus on that one little point. They

251
00:14:45,270 --> 00:14:47,910
S1: make a decision. Is it dangerous? Is it not? How

252
00:14:47,910 --> 00:14:51,840
S1: do you fix it? ET cetera. Right. So that's functional understanding.

253
00:14:51,840 --> 00:14:57,030
S1: And I'm differentiating that from creative understanding. And the way

254
00:14:57,030 --> 00:15:00,840
S1: I talked with the AI about this was creative understanding

255
00:15:00,840 --> 00:15:04,080
S1: is the level above. It's where you're inventing something new.

256
00:15:04,080 --> 00:15:07,110
S1: And now David Deutsch talks about this a lot in

257
00:15:07,110 --> 00:15:10,410
S1: his two books, especially the one I'm starting with here

258
00:15:10,410 --> 00:15:14,970
S1: now in the very beginning of Fabric of Reality. So

259
00:15:15,000 --> 00:15:19,260
S1: my canonical example for this is actually the Nobel Prize.

260
00:15:19,260 --> 00:15:22,230
S1: So I said to the AI, I said, hey, listen,

261
00:15:22,230 --> 00:15:27,150
S1: can we agree that human creativity can be clearly and

262
00:15:27,180 --> 00:15:31,290
S1: obviously accepted as being if you get a Nobel Prize,

263
00:15:31,290 --> 00:15:33,750
S1: you have done it. And the AI is like, yeah,

264
00:15:33,750 --> 00:15:36,120
S1: I think so. If you get a Nobel Prize in

265
00:15:36,120 --> 00:15:40,200
S1: science for something, then yeah, we could say that is creativity.

266
00:15:40,200 --> 00:15:43,320
S1: I'm like, okay, cool. Well, isn't it true that a

267
00:15:43,320 --> 00:15:47,590
S1: whole bunch of Nobel prizes are actually just like iterations.

268
00:15:47,590 --> 00:15:51,010
S1: They're like, we found a new way that a cell

269
00:15:51,040 --> 00:15:56,200
S1: gathers energy. We found a new way that a cell, mitochondria,

270
00:15:56,200 --> 00:16:00,040
S1: creates energy or something, especially in like chemistry. In physics,

271
00:16:00,040 --> 00:16:06,070
S1: it's like not everything is going from Newtonian physics to relativity. Obviously,

272
00:16:06,070 --> 00:16:10,060
S1: that deserves a Nobel. I assume Einstein got a Nobel,

273
00:16:10,060 --> 00:16:12,670
S1: I can't remember. He might not have. Pretty sure he

274
00:16:12,670 --> 00:16:16,750
S1: did like 1905, maybe. Anyway, the point is, tons and

275
00:16:16,750 --> 00:16:23,080
S1: tons and tons of Nobel prizes go to little tiny iterations, okay?

276
00:16:23,110 --> 00:16:27,130
S1: Which is very similar to functional understanding. But here's my point.

277
00:16:27,130 --> 00:16:29,260
S1: And this is what I was debating with the AI about.

278
00:16:29,260 --> 00:16:31,660
S1: I'm like, look, are you telling me that you can't

279
00:16:31,660 --> 00:16:35,350
S1: do this? And it's like it starts off very safe, right?

280
00:16:35,350 --> 00:16:39,730
S1: It's like, well, look, to be clear, I am just

281
00:16:39,730 --> 00:16:43,870
S1: putting together patterns. Um, me as an AI, all I'm

282
00:16:43,870 --> 00:16:48,250
S1: doing is I'm putting together patterns of everything that I know,

283
00:16:48,280 --> 00:16:52,060
S1: all my knowledge. I'm looking for patterns, I find the patterns,

284
00:16:52,060 --> 00:16:53,800
S1: and then I use that to come up with an

285
00:16:53,800 --> 00:16:58,150
S1: answer that hopefully is, you know, pretty decent. And I'm like, um,

286
00:16:58,180 --> 00:17:01,660
S1: isn't that exactly what humans are doing? And it's like, yeah,

287
00:17:01,660 --> 00:17:07,119
S1: but humans do it along with their consciousness. I'm like, stop, stop,

288
00:17:07,119 --> 00:17:10,300
S1: stop saying consciousness. We already talked about the fact that

289
00:17:10,300 --> 00:17:14,530
S1: consciousness is another type of processing, right. So this is

290
00:17:14,530 --> 00:17:17,440
S1: all just information processing. It's all just patterns. It's all

291
00:17:17,440 --> 00:17:21,790
S1: just bouncing ideas back and forth inside of our black boxes. Right.

292
00:17:21,790 --> 00:17:24,040
S1: And it's like, oh yeah, that's right. And I'm like, okay.

293
00:17:24,070 --> 00:17:26,470
S1: So it's really the same thing, right? We are just

294
00:17:26,470 --> 00:17:30,129
S1: information processing considering with a number of different steps and

295
00:17:30,130 --> 00:17:32,619
S1: coming out with a thing. Is that correct. And it's like, yep,

296
00:17:32,619 --> 00:17:35,170
S1: that's correct. And I'm like, okay, are you telling me

297
00:17:35,170 --> 00:17:38,919
S1: that you can't do the second level of understanding? You

298
00:17:38,920 --> 00:17:42,700
S1: can't do understanding that's creative. You can't come up with

299
00:17:42,700 --> 00:17:47,920
S1: a novel new type of explanation. Okay, because explanations are

300
00:17:47,920 --> 00:17:51,879
S1: super key for Deutsch. Okay. Watch this. Um, okay. I'm

301
00:17:51,880 --> 00:17:54,640
S1: going to scroll down and look at this, this conversation. Okay.

302
00:17:54,670 --> 00:17:58,180
S1: So where are we? Um, okay. Watch this. Watch this.

303
00:17:58,180 --> 00:18:02,500
S1: To discern understanding in a transformer based neural net. Okay. So.

304
00:18:02,500 --> 00:18:05,260
S1: So what I'm trying to do is give it, have it,

305
00:18:05,260 --> 00:18:10,000
S1: give me criteria for what it thinks would be creative understanding.

306
00:18:10,030 --> 00:18:12,700
S1: So I'm having it make its bed and try to

307
00:18:12,730 --> 00:18:14,800
S1: sleep in it. So watch this. Here we go. Here

308
00:18:14,800 --> 00:18:17,500
S1: we go. How would you go about discerning whether or

309
00:18:17,500 --> 00:18:21,820
S1: not it's capable of understanding, or if it's randomly generating

310
00:18:21,820 --> 00:18:26,260
S1: in a non-deterministic way? You know, different explanations for things

311
00:18:26,260 --> 00:18:30,550
S1: it already knows. What would the distinction be between real

312
00:18:30,550 --> 00:18:34,990
S1: understanding and David Deutsch's sense versus regurgitation in the sense

313
00:18:34,990 --> 00:18:37,000
S1: of a neural net? So that's what I'm asking it

314
00:18:37,030 --> 00:18:42,310
S1: to discern, understanding, consider its ability to generate novel explanations

315
00:18:42,310 --> 00:18:46,460
S1: and predict outcomes in unfamiliar scenarios. If the AI can

316
00:18:46,460 --> 00:18:51,440
S1: apply knowledge creatively across contexts, it's closer to Deutsch's understanding.

317
00:18:51,440 --> 00:18:55,429
S1: Regurgitation would be repeating information. So I say give me

318
00:18:55,430 --> 00:18:58,220
S1: examples of types of tests like that. Look at this.

319
00:18:58,220 --> 00:19:05,570
S1: Analogical reasoning. Counterfactual scenarios. Conceptual combination error detection. So it

320
00:19:05,570 --> 00:19:08,690
S1: gives me some criteria. I say okay give me some

321
00:19:08,720 --> 00:19:12,050
S1: actual scenarios like three of those. So it gives me

322
00:19:12,050 --> 00:19:15,470
S1: a bunch of scenarios and then I tell it okay,

323
00:19:15,470 --> 00:19:19,040
S1: well guess what? This is funny because you are actually

324
00:19:19,040 --> 00:19:22,040
S1: an AI and it says, okay, I can certainly try

325
00:19:22,040 --> 00:19:26,720
S1: these tasks. Watch this. But remember, mine are based on patterns.

326
00:19:26,720 --> 00:19:28,580
S1: This is what I was telling you before. Mine are

327
00:19:28,580 --> 00:19:31,280
S1: based on patterns in the data that I've been trained on,

328
00:19:31,280 --> 00:19:36,020
S1: not on personal experience or subjective comprehension. And I'm like,

329
00:19:36,020 --> 00:19:38,629
S1: hold on, this is what I was saying before. Hold on.

330
00:19:38,630 --> 00:19:44,920
S1: We just said consciousness is not required consciousness. Stop saying that.

331
00:19:44,920 --> 00:19:48,100
S1: And it's like, that's a great point. Using this framework,

332
00:19:48,100 --> 00:19:52,510
S1: I can indeed generate responses that reflect an understanding based

333
00:19:52,510 --> 00:19:56,380
S1: on patterns and data, even without subjective experience. Okay, so

334
00:19:56,380 --> 00:19:59,800
S1: we're making progress here. Do you want to do some testing? Yes. Go.

335
00:19:59,830 --> 00:20:03,189
S1: It says here's the scenario. And it starts giving me one.

336
00:20:03,190 --> 00:20:05,919
S1: And I'm like hold on. No, no no I'm giving

337
00:20:05,920 --> 00:20:08,560
S1: you the test. I want you to do it. And

338
00:20:08,560 --> 00:20:11,469
S1: I'm like, look, let's pretend someone doesn't believe that you

339
00:20:11,500 --> 00:20:13,630
S1: are understanding and they're going to give you a really

340
00:20:13,630 --> 00:20:15,760
S1: hard challenge. What would they give you as a challenge?

341
00:20:15,760 --> 00:20:18,700
S1: And could you do it? And it gives me some blabbering. Again,

342
00:20:18,730 --> 00:20:21,160
S1: I'm like, look, that's also true for humans, so just

343
00:20:21,160 --> 00:20:24,790
S1: do it. Here's a speculative hypothesis. Quantum processing in the

344
00:20:24,790 --> 00:20:27,610
S1: brain okay. Yeah. So it gives me this. Really I

345
00:20:27,609 --> 00:20:29,710
S1: thought it was really good. It was pretty good. I

346
00:20:29,710 --> 00:20:32,199
S1: mean it was kind of similar to things that I

347
00:20:32,200 --> 00:20:36,159
S1: was thinking about distributed encoding. Oh. So I asked it

348
00:20:36,160 --> 00:20:39,129
S1: to figure out how memories are stored in the brain,

349
00:20:39,130 --> 00:20:40,960
S1: because I know this is the thing that is not

350
00:20:40,960 --> 00:20:44,050
S1: known in actual science. So it gave me three actual

351
00:20:44,050 --> 00:20:47,380
S1: theories and I'm like, hey, this kind of sounds like

352
00:20:47,380 --> 00:20:51,010
S1: something I've heard before. Is this brand new? Anyway, so

353
00:20:51,010 --> 00:20:54,910
S1: we keep going. We keep going. Okay, Daichi, an explanation

354
00:20:54,910 --> 00:21:00,520
S1: emphasizes creating deep underlying theories that unify disparate phenomena under

355
00:21:00,520 --> 00:21:03,940
S1: fundamental principles. I love that definition, by the way. I'm

356
00:21:03,940 --> 00:21:06,520
S1: going to steal that. Probably. That'll probably be the foundation

357
00:21:06,520 --> 00:21:09,940
S1: of what I come up with. Understanding is essentially a system.

358
00:21:09,940 --> 00:21:12,940
S1: Or by his definition, understanding is the ability to create

359
00:21:12,970 --> 00:21:16,570
S1: new explanation using that understanding. Oh, that was me talking.

360
00:21:16,570 --> 00:21:20,440
S1: So yes. So it agreed with that. Okay. So this

361
00:21:20,440 --> 00:21:25,659
S1: is where I define functional versus creative understanding. Okay. Now

362
00:21:25,660 --> 00:21:28,659
S1: here's where we get into that final debate here. Okay

363
00:21:28,690 --> 00:21:33,040
S1: I tell it again creative understanding does not require subjective experience.

364
00:21:33,040 --> 00:21:36,100
S1: And then I then I realized that I'm coaching it

365
00:21:36,100 --> 00:21:38,770
S1: too much. Right. I say at least that's my claim.

366
00:21:38,770 --> 00:21:40,940
S1: If you disagree, I want you to argue with me

367
00:21:40,940 --> 00:21:43,160
S1: about that. But the point is, you can't just say

368
00:21:43,160 --> 00:21:46,850
S1: that your experience in finding patterns is not enough. Because

369
00:21:46,880 --> 00:21:49,100
S1: guess what? Humans don't have free will. We live in

370
00:21:49,130 --> 00:21:53,150
S1: a mechanistic universe by definition. We've already established that humans

371
00:21:53,150 --> 00:21:56,390
S1: have creative understanding. So if we have it and we

372
00:21:56,390 --> 00:21:59,270
S1: live in a mechanistic universe and there is no free will,

373
00:21:59,270 --> 00:22:01,970
S1: that means human creativity is still made up of a

374
00:22:01,970 --> 00:22:05,659
S1: combination of interactions inside this black box of a brain,

375
00:22:05,660 --> 00:22:09,380
S1: which results in this creativity, which everyone agrees that we

376
00:22:09,380 --> 00:22:12,380
S1: have it right. So it's like, yeah, yeah, yeah, I

377
00:22:12,380 --> 00:22:16,190
S1: get it. Fair point. So if we define creative understanding

378
00:22:16,190 --> 00:22:19,760
S1: as the ability to generate novel ideas or explanations through

379
00:22:19,760 --> 00:22:22,850
S1: a complex interplay of patterns and data similar to how

380
00:22:22,850 --> 00:22:26,660
S1: humans do, then it's conceivable that AI's like me could

381
00:22:26,660 --> 00:22:30,500
S1: approach this level. Great, now we're making progress. And keep

382
00:22:30,500 --> 00:22:32,780
S1: in mind, coaching an AI to be able to say

383
00:22:32,780 --> 00:22:35,570
S1: what you want doesn't mean it's true, right? This this

384
00:22:35,570 --> 00:22:39,530
S1: could be prompt injection and engineering. So we have to

385
00:22:39,560 --> 00:22:42,109
S1: keep a logical framework on this whole thing, because the

386
00:22:42,109 --> 00:22:44,630
S1: fact that I can make it say something that agrees

387
00:22:44,630 --> 00:22:46,940
S1: with me does not make it true. So you got

388
00:22:46,940 --> 00:22:49,250
S1: to follow the whole line of this debate and the

389
00:22:49,250 --> 00:22:52,189
S1: argument and make sure it's being logical here. I'm definitely

390
00:22:52,190 --> 00:22:55,250
S1: not claiming that just because it wrote that. Like, we've

391
00:22:55,250 --> 00:22:58,880
S1: settled this right. So you got to follow along here. Okay.

392
00:22:58,880 --> 00:23:01,460
S1: Let's take the opposite side. So this is me arguing.

393
00:23:01,460 --> 00:23:05,090
S1: There is something special that humans have that AIS don't have.

394
00:23:05,090 --> 00:23:09,290
S1: And I have it go into the different things. And

395
00:23:09,290 --> 00:23:11,990
S1: we kind of go down that path for a little while. Okay.

396
00:23:11,990 --> 00:23:14,690
S1: And I say, yeah, we we might have some special

397
00:23:14,690 --> 00:23:19,580
S1: sauce like I'm giving this as ammunition to the argument. Um,

398
00:23:19,609 --> 00:23:21,320
S1: all right. So let's keep going. I want to get

399
00:23:21,530 --> 00:23:24,140
S1: down to this end part. Yeah. We know consciousness is

400
00:23:24,140 --> 00:23:26,930
S1: not an illusion. What we could say is consciousness is

401
00:23:26,930 --> 00:23:29,929
S1: not a special thing outside of normal processing. And the

402
00:23:29,930 --> 00:23:32,389
S1: fact that it seems special could be an illusion. In

403
00:23:32,390 --> 00:23:35,989
S1: other words, it could be regular information processing coming out

404
00:23:35,990 --> 00:23:38,619
S1: of the black box of our brain. So the process

405
00:23:38,619 --> 00:23:41,320
S1: could be like, you could argue it's artificial, but the

406
00:23:41,320 --> 00:23:45,159
S1: fact that when we experience authorship, that is not an illusion, right?

407
00:23:45,190 --> 00:23:48,340
S1: Because and it says, yeah, it sees the distinction. Okay.

408
00:23:48,340 --> 00:23:51,880
S1: So I'm trying to get down to this part talking

409
00:23:51,880 --> 00:23:55,270
S1: about the Nobel Prize. Oh yeah. So this is the

410
00:23:55,270 --> 00:23:57,520
S1: whole theory about this is my theory about how we

411
00:23:57,520 --> 00:24:00,700
S1: got consciousness from evolution. Worth reading. I'm not going to

412
00:24:00,700 --> 00:24:03,730
S1: go too much more into it. Okay, so here we go.

413
00:24:03,760 --> 00:24:06,550
S1: I just thought of another way we can determine whether

414
00:24:06,550 --> 00:24:09,399
S1: or not eyes like yourself come up with a creative

415
00:24:09,400 --> 00:24:12,760
S1: level of understanding. We could basically look at examples of

416
00:24:12,760 --> 00:24:16,690
S1: things that have won the Nobel Prize in science, which

417
00:24:16,690 --> 00:24:20,530
S1: by definition is innovation or creativity. How many of those

418
00:24:20,530 --> 00:24:22,570
S1: do you think you could have done? And it's like,

419
00:24:22,570 --> 00:24:26,199
S1: that's an interesting approach. Let's look at past Nobel Prize

420
00:24:26,230 --> 00:24:30,310
S1: winning discoveries and use that as a benchmark. And I

421
00:24:30,340 --> 00:24:33,970
S1: slammed it for being like too safe. I talk about

422
00:24:33,970 --> 00:24:38,410
S1: the minor developments like a new mechanism the cell does,

423
00:24:38,440 --> 00:24:42,070
S1: it says. That's a fair point. Many Nobel Prize winning

424
00:24:42,070 --> 00:24:48,010
S1: discoveries do indeed involve incremental advances or novel observations within

425
00:24:48,010 --> 00:24:52,480
S1: existing frameworks, particularly in fields like chemistry and biology. Watch this.

426
00:24:52,510 --> 00:24:57,730
S1: In these cases, I could potentially identify patterns or mechanisms

427
00:24:57,730 --> 00:25:01,480
S1: that lead to similar discoveries, given the right data and

428
00:25:01,480 --> 00:25:07,359
S1: analytical tools. This suggests that I could indeed contribute to

429
00:25:07,390 --> 00:25:12,190
S1: scientific innovations at levels recognized by the Nobel Committee, especially

430
00:25:12,190 --> 00:25:16,060
S1: as these systems continue to evolve. I don't think that

431
00:25:16,060 --> 00:25:19,420
S1: was coached too much to say that. I think that

432
00:25:19,420 --> 00:25:23,770
S1: is a logical conclusion from the debate that we just had.

433
00:25:23,800 --> 00:25:28,479
S1: And I think what that shows is not only functional understanding,

434
00:25:28,480 --> 00:25:33,520
S1: which I already have, but this higher level, more stringent

435
00:25:33,520 --> 00:25:38,090
S1: level definition that I'm using of creative understanding. I am

436
00:25:38,090 --> 00:25:41,689
S1: arguing that an I will soon be able to win

437
00:25:41,690 --> 00:25:45,199
S1: a Nobel Prize. I am arguing that an AI is

438
00:25:45,200 --> 00:25:49,189
S1: already capable of making the type of jumps, the types

439
00:25:49,190 --> 00:25:54,470
S1: of jumps required to make one of these iterative discoveries

440
00:25:54,470 --> 00:25:57,919
S1: that would get somebody a Nobel Prize. In fact, if

441
00:25:57,920 --> 00:26:00,170
S1: I were to go and compile a list of all

442
00:26:00,200 --> 00:26:02,990
S1: the Nobel Prizes and then do a summary of how

443
00:26:02,990 --> 00:26:05,810
S1: much of a jump it was and then give a

444
00:26:05,810 --> 00:26:08,689
S1: bunch of scenarios to current AIS. I bet you they

445
00:26:08,720 --> 00:26:12,080
S1: are already capable of making as good or more of

446
00:26:12,080 --> 00:26:15,080
S1: a jump. Now, could they do all of the experimental

447
00:26:15,080 --> 00:26:17,689
S1: work and all of the research and all of the

448
00:26:17,690 --> 00:26:20,690
S1: stuff that backs that up? Maybe not, but that's all

449
00:26:20,690 --> 00:26:24,200
S1: coming soon. That's the easy part, right? Not the experimental part,

450
00:26:24,230 --> 00:26:27,709
S1: because that could involve physical world and robots. But the

451
00:26:27,710 --> 00:26:30,470
S1: point is, it's not the creative part. The creative part

452
00:26:30,470 --> 00:26:35,179
S1: is the idea and the explanation and the experiment and

453
00:26:35,180 --> 00:26:38,060
S1: the different way of seeing the world. I'm arguing. Eyes

454
00:26:38,090 --> 00:26:41,030
S1: are already getting close to that, if not have already passed.

455
00:26:41,030 --> 00:26:44,990
S1: So all this to say, yeah, eyes do understand. They

456
00:26:44,990 --> 00:26:48,770
S1: understand both at the functional level and at the creative

457
00:26:48,770 --> 00:26:52,280
S1: level of understanding two different levels of understanding. All right.

458
00:26:52,280 --> 00:26:56,300
S1: That was absolutely the longest jump out from a piece

459
00:26:56,300 --> 00:26:59,570
S1: but wanted to talk about that. Transformers create shapes of

460
00:26:59,570 --> 00:27:02,060
S1: the universe. This one is just I'm going to do

461
00:27:02,060 --> 00:27:04,369
S1: a separate one on this because we already went down

462
00:27:04,369 --> 00:27:07,940
S1: a tangent. Vulnerability management isn't about finding issues, it's about

463
00:27:07,940 --> 00:27:11,420
S1: fixing them in context. Did this for dads, which is

464
00:27:11,420 --> 00:27:15,320
S1: a cool vulnerability management company who sponsored the post, and

465
00:27:15,320 --> 00:27:18,020
S1: I've been wanting to write that article anyway for a while,

466
00:27:18,020 --> 00:27:20,450
S1: so that was a good opportunity to do it. Book

467
00:27:20,480 --> 00:27:24,110
S1: club this weekend was so, so good, and we've selected

468
00:27:24,109 --> 00:27:28,580
S1: The Republic by Plato for this week's or this month's book.

469
00:27:28,580 --> 00:27:31,490
S1: So really excited about that. Oh yeah, currently reading Fabric

470
00:27:31,490 --> 00:27:33,859
S1: of Reality and the Beginning of Infinity. These are the

471
00:27:33,859 --> 00:27:37,370
S1: ones that spawned that whole conversation. And. All right, let's

472
00:27:37,369 --> 00:27:40,820
S1: get into it. Security scissors. Warning. Brute force attacks against

473
00:27:40,820 --> 00:27:45,260
S1: default credentials against ICS and SCADA. Um, sharp increase in

474
00:27:45,260 --> 00:27:48,170
S1: OT attacks reported by Fortinet. And a lot of people

475
00:27:48,200 --> 00:27:51,470
S1: are saying that this is China basically getting ready for

476
00:27:51,500 --> 00:27:54,770
S1: a kinetic battle with the US by preparing a whole

477
00:27:54,770 --> 00:27:57,169
S1: bunch of cyber stuff. So the idea is when it

478
00:27:57,170 --> 00:28:01,400
S1: gets ready, gets ready to go into Taiwan, they'll launch

479
00:28:01,400 --> 00:28:05,570
S1: all these infrastructure attacks and basically mess with us, mess

480
00:28:05,570 --> 00:28:09,020
S1: with our society, potentially mess with the military as well,

481
00:28:09,050 --> 00:28:13,429
S1: basically cause a whole bunch of cyber related critical infrastructure

482
00:28:13,430 --> 00:28:18,410
S1: problems for the US power, water, you know, other critical services,

483
00:28:18,410 --> 00:28:23,450
S1: internet and basically mass disruption while they're going into Taiwan.

484
00:28:23,480 --> 00:28:26,030
S1: That's a theory a lot of people have. I'm not

485
00:28:26,030 --> 00:28:29,240
S1: sure how valid it is. Seems somewhat valid, at least.

486
00:28:29,270 --> 00:28:33,650
S1: China reportedly achieved a significant ah, milestone by developing a

487
00:28:33,650 --> 00:28:37,880
S1: generative AI model that operates across multiple data centers and

488
00:28:37,880 --> 00:28:42,020
S1: GPU architectures. Particularly impressive given how hard it is to

489
00:28:42,020 --> 00:28:45,650
S1: get stuff right because supposedly under sanctions. But they seem

490
00:28:45,650 --> 00:28:48,500
S1: to be making plenty of progress. One thing I wanted

491
00:28:48,530 --> 00:28:51,950
S1: to point out is that it's not PvP necessarily against China.

492
00:28:51,950 --> 00:28:54,680
S1: It's PvE because they only need to be able to

493
00:28:54,680 --> 00:28:58,760
S1: do these things here, right? Continuously crawl for all of

494
00:28:58,760 --> 00:29:03,110
S1: our attack surface and find vulnerabilities. Write exploits for those vulnerabilities,

495
00:29:03,140 --> 00:29:11,210
S1: actually exploit them, control their population, launch disinformation campaigns, propaganda campaigns, uh,

496
00:29:11,210 --> 00:29:14,240
S1: distribute tech that can do this to all of their

497
00:29:14,240 --> 00:29:19,040
S1: buddies that are also totalitarian, right? So if they could

498
00:29:19,040 --> 00:29:21,560
S1: do the if they get good enough to do these things,

499
00:29:21,560 --> 00:29:24,230
S1: it doesn't matter if we're 2 or 3 generations ahead.

500
00:29:24,230 --> 00:29:27,110
S1: That's why it's PvE instead of PvP. The only way

501
00:29:27,110 --> 00:29:29,960
S1: it's PvP is that if they are doing these things

502
00:29:29,990 --> 00:29:34,010
S1: against us, using a certain level of AI or ASI

503
00:29:34,010 --> 00:29:38,000
S1: or like AGI, ASI we want to have a next

504
00:29:38,000 --> 00:29:41,060
S1: generation one that's capable of seeing it and stopping it.

505
00:29:41,060 --> 00:29:43,520
S1: But that requires not just the brain, but all the

506
00:29:43,520 --> 00:29:46,970
S1: visibility as well to feed the brain. Okay, Worldcoin is

507
00:29:46,970 --> 00:29:50,810
S1: addressing the issue of bots and deepfakes with decentralized identity

508
00:29:50,840 --> 00:29:54,020
S1: system that confirms if someone is human. So we're starting

509
00:29:54,020 --> 00:29:57,350
S1: to see the practical uses for Worldcoin. By the way,

510
00:29:57,350 --> 00:30:01,520
S1: it's an Ethereum based blockchain. Validation of humans. Interesting stuff.

511
00:30:01,520 --> 00:30:05,990
S1: Chinese attacker called Salt Typhoon. Very cool. Should be a

512
00:30:05,990 --> 00:30:11,990
S1: band name targeting Cisco Systems to establish persistence. Again, going

513
00:30:11,990 --> 00:30:14,960
S1: back to the thing, just imagine that they have hundreds

514
00:30:14,960 --> 00:30:18,770
S1: of thousands of different systems that they're able to mess with,

515
00:30:18,770 --> 00:30:23,840
S1: like rerouting things. BGP attacks, like just imagine the internet

516
00:30:23,870 --> 00:30:27,740
S1: not working power going down for so many people. All right.

517
00:30:27,740 --> 00:30:32,180
S1: Before the Taiwan attack, U.S. has charged three Iranian nationals

518
00:30:32,180 --> 00:30:36,260
S1: linked to the IRGC for hacking into accounts of U.S.

519
00:30:36,260 --> 00:30:40,070
S1: officials and political campaigns. T-Mobile got hit with a $16

520
00:30:40,070 --> 00:30:44,480
S1: million fine from the FCC, and they were basically forced

521
00:30:44,480 --> 00:30:47,750
S1: to spend another 16 million on security as part of

522
00:30:47,750 --> 00:30:51,560
S1: the settlement. And massive love to my buddy Matt Johannson,

523
00:30:51,560 --> 00:30:55,820
S1: this story right here goes out to his site, which

524
00:30:55,820 --> 00:30:59,000
S1: is vulnerable. You. So now a lot more of my stories.

525
00:30:59,000 --> 00:31:02,330
S1: He's going to start making actual cybersecurity stories. So instead

526
00:31:02,330 --> 00:31:05,810
S1: of going to like Bleeping Computer and uh, what are

527
00:31:05,810 --> 00:31:09,470
S1: some of the other ones, a bunch of security news ones, right. Um,

528
00:31:09,470 --> 00:31:13,280
S1: but I'm like helping him with like some of the summarization,

529
00:31:13,280 --> 00:31:16,010
S1: different pieces, like what a good summary looks like, big

530
00:31:16,010 --> 00:31:20,120
S1: picture impact, why it matters. So I'm giving a bunch

531
00:31:20,120 --> 00:31:22,640
S1: of feedback on that kind of stuff. And, uh, yeah,

532
00:31:22,670 --> 00:31:24,950
S1: I really love the idea of being able to point

533
00:31:24,950 --> 00:31:29,150
S1: to my buddy's site. Who can do ultimately, to do this? Well,

534
00:31:29,150 --> 00:31:31,380
S1: you have to find the stories and you have to

535
00:31:31,410 --> 00:31:34,080
S1: summarize them well. And the reason he's going to be

536
00:31:34,080 --> 00:31:36,780
S1: so good at this is because he is an actual

537
00:31:36,780 --> 00:31:41,610
S1: cybersecurity expert. He runs like Appsec, Reddit, and he has

538
00:31:41,610 --> 00:31:44,130
S1: been in this field, I think like almost as long

539
00:31:44,130 --> 00:31:47,520
S1: as me. Uh, or if not the same level. I

540
00:31:47,520 --> 00:31:49,050
S1: don't think so. I don't think he's as old as me,

541
00:31:49,050 --> 00:31:52,200
S1: but he's been doing this for so long. One of

542
00:31:52,200 --> 00:31:55,170
S1: the longest people in the industry, and he's an absolute

543
00:31:55,200 --> 00:31:58,050
S1: guru at it. And he's good at writing at the

544
00:31:58,050 --> 00:32:01,560
S1: same time, and he's learning AI. So it's like, this

545
00:32:01,560 --> 00:32:04,950
S1: is the perfect combination, and I can't wait to start

546
00:32:04,950 --> 00:32:07,830
S1: pointing more and more of my stories to his site

547
00:32:07,830 --> 00:32:10,830
S1: and spreading the love there as a buddy as well. U.S.

548
00:32:10,830 --> 00:32:15,570
S1: Commerce Department proposing a ban on Chinese, Russian, automotive, auto,

549
00:32:15,600 --> 00:32:19,530
S1: moto automotive hardware and software. Basically saying it's a way

550
00:32:19,530 --> 00:32:24,690
S1: for adversaries to get footholds into U.S. systems and infrastructure, which, yeah,

551
00:32:24,690 --> 00:32:28,140
S1: that's what we've been talking about. Ukraine reportedly found Starlink

552
00:32:28,170 --> 00:32:34,130
S1: terminals in a downed Russian Shahed drone. LM based hacking assistance.

553
00:32:34,160 --> 00:32:36,500
S1: I think we all get that lots of different tools

554
00:32:36,500 --> 00:32:40,220
S1: here that actually help you attack and defend with Llms.

555
00:32:40,250 --> 00:32:43,730
S1: Telegram is now starting to hand over phone numbers and

556
00:32:43,760 --> 00:32:48,230
S1: IP addresses of suspects. This is all part of their

557
00:32:48,230 --> 00:32:51,650
S1: CEO getting arrested. Security researchers found a way to remotely

558
00:32:51,650 --> 00:32:55,430
S1: control millions of Kia vehicles by only having the license

559
00:32:55,430 --> 00:32:58,580
S1: plate number, which allows you to look up and basically

560
00:32:58,580 --> 00:33:02,150
S1: gain controls. Pretty nasty. One so the biggest hype switching

561
00:33:02,180 --> 00:33:04,400
S1: to AI in tech, biggest hype in the last week

562
00:33:04,400 --> 00:33:07,490
S1: has been around notebook LM, which is, by the way,

563
00:33:07,490 --> 00:33:10,250
S1: what I use to create that, um, that podcast I

564
00:33:10,250 --> 00:33:14,060
S1: played for you earlier. So it's a big Google project.

565
00:33:14,060 --> 00:33:16,100
S1: I don't know why they're not talking about it more.

566
00:33:16,310 --> 00:33:18,380
S1: Never mind. Yes, I do, because they're really bad at

567
00:33:18,380 --> 00:33:22,850
S1: product and marketing and everything like that. So basically it's

568
00:33:22,850 --> 00:33:25,820
S1: this tiny little project which they're not announcing, they're not

569
00:33:25,820 --> 00:33:29,250
S1: putting out anywhere. Like you learn about it on Twitter? Whatever.

570
00:33:29,280 --> 00:33:32,310
S1: It'll probably be dead in a week anyway. Uh, to

571
00:33:32,340 --> 00:33:36,420
S1: the graveyard, along with everything else. But anyway, fun while

572
00:33:36,420 --> 00:33:40,140
S1: it lasted. It's super fun. And oh yeah, here's an example.

573
00:33:40,140 --> 00:33:42,750
S1: So I uploaded an Alma file, which is part of

574
00:33:42,750 --> 00:33:44,760
S1: a Telos project, which I'm going to be talking about,

575
00:33:44,790 --> 00:33:48,540
S1: hopefully making content on tomorrow. Anyway, it's this giant context

576
00:33:48,540 --> 00:33:52,320
S1: file I uploaded it. It is actually the context file

577
00:33:52,320 --> 00:33:56,610
S1: in the Telos format for a security program. So managing

578
00:33:56,610 --> 00:33:59,430
S1: a cyber security program for a company, it's for this

579
00:33:59,430 --> 00:34:02,250
S1: fake company called alma. I've uploaded it. It's in the

580
00:34:02,250 --> 00:34:05,490
S1: fabric repo right now, but it's soon going to be

581
00:34:05,490 --> 00:34:08,670
S1: in the Telos repo as well, which is designed to

582
00:34:08,670 --> 00:34:12,000
S1: do specifically this. But check this out. You put it

583
00:34:12,000 --> 00:34:15,750
S1: in there. Look at this generate deep dive conversation. You

584
00:34:15,750 --> 00:34:19,200
S1: can actually create a podcast based on that security program.

585
00:34:19,200 --> 00:34:22,020
S1: And they'll just talk about the security program back and forth.

586
00:34:22,050 --> 00:34:25,470
S1: That's crazy by itself. But most importantly, look, you can

587
00:34:25,469 --> 00:34:28,340
S1: ask these questions about the security program. And you got

588
00:34:28,340 --> 00:34:31,610
S1: this typing field down here where you can interact and

589
00:34:31,610 --> 00:34:35,330
S1: ask questions about the program. So super cool. Newsom. Yeah,

590
00:34:35,360 --> 00:34:40,310
S1: recommend checking out notebook LM Google because they definitely needed

591
00:34:40,310 --> 00:34:45,050
S1: a Google domain. Whatever. Newsom vetoed the California AI bill,

592
00:34:45,080 --> 00:34:48,230
S1: saying it was too restrictive. I agree with him on this.

593
00:34:48,260 --> 00:34:51,620
S1: A good, good move, though. I want to see another

594
00:34:51,620 --> 00:34:54,860
S1: follow up bill. California made it illegal to use AI

595
00:34:54,890 --> 00:34:59,000
S1: to impersonate actors without their consent. Somebody found out their

596
00:34:59,000 --> 00:35:02,839
S1: voice was being AI cloned by a company and they're

597
00:35:02,840 --> 00:35:08,180
S1: mad about it. Convergence AI, co-founded by Marvin Porter and

598
00:35:08,180 --> 00:35:13,130
S1: Andy Tallis, has raised 12 million to develop proxy agents

599
00:35:13,130 --> 00:35:16,700
S1: with long term memory. New task of agent, new class

600
00:35:16,700 --> 00:35:19,970
S1: of agents. Yeah, uh, this is going to be huge.

601
00:35:19,969 --> 00:35:24,470
S1: Huge meta launched three two. They can now handle images

602
00:35:24,469 --> 00:35:28,440
S1: and text competing with OpenAI and anthropic on that ChatGPT

603
00:35:28,469 --> 00:35:32,070
S1: advanced mode. Definitely, definitely. I use it all the time

604
00:35:32,070 --> 00:35:34,950
S1: as all these examples we just went through. Hugging face

605
00:35:34,950 --> 00:35:41,220
S1: now has 1 million AI model listings. 1 million models, all.

606
00:35:41,250 --> 00:35:43,890
S1: Are they all open source? Pretty sure they're all open source.

607
00:35:43,920 --> 00:35:45,900
S1: Are they all open source? I think so. I think

608
00:35:45,900 --> 00:35:48,690
S1: they're all open source. They're definitely all there for you

609
00:35:48,690 --> 00:35:52,710
S1: to use. I think that implies that they're open source. Um, yeah.

610
00:35:52,739 --> 00:35:56,700
S1: Anthropic is in early talks with investors. Could value the

611
00:35:56,700 --> 00:35:59,580
S1: company at 30 to 40 billion. Oh, by the way,

612
00:35:59,610 --> 00:36:03,480
S1: OpenAI just closed its funding, so I think it's worth

613
00:36:03,480 --> 00:36:06,600
S1: some crazy amount now. They just raised like more than

614
00:36:06,630 --> 00:36:09,600
S1: Elon did for a lot of his projects. Yeah. Crazy

615
00:36:09,630 --> 00:36:12,779
S1: crazy amount of money. They just raised enclosed. Google's rolled

616
00:36:12,780 --> 00:36:17,760
S1: out two new production Gemini models, constant stream model updates

617
00:36:17,760 --> 00:36:21,420
S1: for the main players, much smarter models or models that

618
00:36:21,420 --> 00:36:24,480
S1: can do new stuff like vision or whatever, multimodal, and

619
00:36:24,480 --> 00:36:28,830
S1: then models that are like two 500 times cheaper. One

620
00:36:28,830 --> 00:36:31,169
S1: of the cool things that was just mentioned is like

621
00:36:31,200 --> 00:36:35,190
S1: in general, models are 99% cheaper than a year ago

622
00:36:35,190 --> 00:36:40,439
S1: for the same functionality 99%. Ilyas I reading list Agile's

623
00:36:40,440 --> 00:36:44,700
S1: original principles have been lost in translation. AI is disrupting

624
00:36:44,700 --> 00:36:48,840
S1: the traditional ad supported internet model. Yea yea yea yea yea.

625
00:36:48,870 --> 00:36:52,770
S1: This one's crucial. I did a piece a little while

626
00:36:52,770 --> 00:36:56,640
S1: back called what was that called? I don't remember. It

627
00:36:56,640 --> 00:37:00,450
S1: doesn't matter. Um, real internet things. I've got this part

628
00:37:00,480 --> 00:37:04,950
S1: called AI mediation. Oh yeah. Here's the AI mediation piece. Yeah.

629
00:37:04,980 --> 00:37:08,550
S1: Das will mediate everything. We'll say what we want. Our

630
00:37:08,580 --> 00:37:12,780
S1: Das will manipulate external demons and APIs to make it happen.

631
00:37:12,780 --> 00:37:17,040
S1: So this is from, uh, I think the very beginning

632
00:37:17,070 --> 00:37:20,370
S1: of the year I put this out, but it's actually

633
00:37:20,670 --> 00:37:24,510
S1: from my book from 2016. And this is just an updated,

634
00:37:24,520 --> 00:37:28,480
S1: illustrated version. I've been talking about this. You should check

635
00:37:28,510 --> 00:37:30,730
S1: out this book, by the way. It's called The Real

636
00:37:30,730 --> 00:37:33,640
S1: Internet of Things. It's actually a little bit crappy of

637
00:37:33,640 --> 00:37:37,180
S1: a book. It's more like a blog post or like

638
00:37:37,180 --> 00:37:40,690
S1: an essay, but it's pretty well structured. But most importantly,

639
00:37:40,690 --> 00:37:43,600
S1: all these things that I've been saying on the podcast

640
00:37:43,600 --> 00:37:45,700
S1: that you've been listening to for like 2 or 3

641
00:37:45,700 --> 00:37:48,100
S1: years or five years or however long it was all

642
00:37:48,100 --> 00:37:51,520
S1: in that book in 2016, all these predictions about AI,

643
00:37:51,550 --> 00:37:54,040
S1: what's about to happen to AR? It's all in there.

644
00:37:54,040 --> 00:37:56,890
S1: And that's why I've been consistent about what I've been

645
00:37:56,890 --> 00:38:00,310
S1: saying and predicting all along. So I highly recommend you

646
00:38:00,310 --> 00:38:02,830
S1: check out the source material, which is that first book.

647
00:38:03,010 --> 00:38:05,500
S1: And yes, I was even talking about AI back in

648
00:38:05,500 --> 00:38:08,620
S1: 2016 and that that I was the source of all

649
00:38:08,620 --> 00:38:10,900
S1: these different predictions. Obviously, I didn't know it was going

650
00:38:10,930 --> 00:38:13,600
S1: to happen in 2022. I thought it might take longer,

651
00:38:13,600 --> 00:38:18,280
S1: but yeah. Anyway, go check that out. Um, SpaceX Starlink

652
00:38:18,310 --> 00:38:22,089
S1: is about to hit 4 million subscribers. Um, yeah, I

653
00:38:22,090 --> 00:38:25,950
S1: base a lot of my investment in prediction on leaders

654
00:38:25,950 --> 00:38:31,050
S1: and ideas. Elon jobs Jenson Zuckerberg doesn't mean they're going

655
00:38:31,080 --> 00:38:33,630
S1: to win, but I think it's bad idea to bet

656
00:38:33,630 --> 00:38:38,310
S1: against them because they are so locked in on vision

657
00:38:38,310 --> 00:38:42,120
S1: and also able to execute. Smart TVs from Samsung and

658
00:38:42,120 --> 00:38:45,360
S1: LG are taking snapshots of what you're watching multiple times

659
00:38:45,360 --> 00:38:49,020
S1: per second. Plus energy started a $1 billion solar and

660
00:38:49,020 --> 00:38:54,420
S1: battery storage project. The white collar apocalypse is nigh. The

661
00:38:54,420 --> 00:38:57,900
S1: decline of status, economy and the impact of AI on

662
00:38:57,900 --> 00:39:01,290
S1: job markets. Yeah, pretty cool one. Definitely go check this

663
00:39:01,290 --> 00:39:05,070
S1: one out. Meta. Orion's AR glasses are a big upgrade.

664
00:39:05,070 --> 00:39:09,600
S1: I don't think Mark is right about replacing phones by 2030.

665
00:39:09,630 --> 00:39:13,470
S1: I think that's too ambitious. I think phones offer too

666
00:39:13,469 --> 00:39:16,560
S1: many things that glasses don't, and I just don't think

667
00:39:16,560 --> 00:39:20,910
S1: the physical miniaturization tech is there to go into the

668
00:39:20,910 --> 00:39:24,060
S1: glasses to replicate what's on the phone Hacker News thread,

669
00:39:24,090 --> 00:39:26,820
S1: where people are sharing their experiences about whether having a

670
00:39:26,820 --> 00:39:30,509
S1: personal website helped get them hired. Some people said yes,

671
00:39:30,540 --> 00:39:32,460
S1: a lot of people said yes. Some people said no.

672
00:39:32,460 --> 00:39:36,300
S1: All right, humans. California is taking phones out of schools,

673
00:39:36,300 --> 00:39:40,440
S1: basically limiting cell phone use. So massive props to Jonathan

674
00:39:40,440 --> 00:39:43,950
S1: Hite on this. Uh, someone found a method that uses

675
00:39:43,950 --> 00:39:48,629
S1: stem cells to regenerate insulin producing cells. ExxonMobil is facing

676
00:39:48,630 --> 00:39:53,100
S1: a lawsuit for coming up with a supposed fraud saying

677
00:39:53,130 --> 00:39:57,720
S1: or a myth saying that recycling could help solve the

678
00:39:57,719 --> 00:40:00,899
S1: problem with plastic waste. California is looking to make it

679
00:40:00,930 --> 00:40:04,830
S1: one click. Get off of subscriptions just the same way.

680
00:40:04,830 --> 00:40:07,920
S1: One click sign ups. So if you offer a one

681
00:40:07,950 --> 00:40:09,930
S1: click sign up, you have to offer a one way

682
00:40:09,930 --> 00:40:12,630
S1: to get out of it as well. Over 10,000 books

683
00:40:12,630 --> 00:40:17,100
S1: were banned in US public schools from 2023 to 2024.

684
00:40:17,130 --> 00:40:20,880
S1: Insane Steve Jobs had a thing called a ten minute rule,

685
00:40:20,880 --> 00:40:23,400
S1: just ten minutes to tackle a problem or a task.

686
00:40:23,430 --> 00:40:25,680
S1: Check this out. That's why I love doing video. I've

687
00:40:25,680 --> 00:40:28,589
S1: got one of these. I actually use it. It's called pomodoro,

688
00:40:28,590 --> 00:40:32,160
S1: but you can set it for any time. Right. So

689
00:40:32,190 --> 00:40:35,640
S1: 30 minutes is like the classic one. It's inverted. But

690
00:40:35,640 --> 00:40:37,710
S1: but you can put it down to like. There you go.

691
00:40:37,739 --> 00:40:43,379
S1: Ten minutes. So I think it's cool. Constraints. Constraints inspire creativity.

692
00:40:43,530 --> 00:40:48,509
S1: And Maria Popova explores profound curiosity and love of Henry

693
00:40:48,540 --> 00:40:54,299
S1: David Thoreau as he encounters a screech owl in the wild. Ideas.

694
00:40:54,300 --> 00:40:57,930
S1: Everything is going behind the paywall. Yeah, I'm not going

695
00:40:57,960 --> 00:41:01,410
S1: to read this. I'm just going to give you the overview. Freeform.

696
00:41:01,440 --> 00:41:05,400
S1: I think it's really interesting to think about how there's

697
00:41:05,430 --> 00:41:08,790
S1: the open internet is going away. Everything's going to be

698
00:41:08,790 --> 00:41:12,210
S1: behind auth not only to block bots but also to

699
00:41:12,239 --> 00:41:16,439
S1: get micropayments. So rich people who have money for their

700
00:41:16,469 --> 00:41:18,780
S1: their eyes. Okay I'm going to give you a little

701
00:41:18,810 --> 00:41:20,549
S1: bit of a I also have to make this piece

702
00:41:20,550 --> 00:41:24,089
S1: of content, but a tiny little view of the future. Okay,

703
00:41:24,120 --> 00:41:26,640
S1: you're going to have all your personal context in your

704
00:41:26,640 --> 00:41:29,760
S1: personal AI. It's with you all the time. Your personal

705
00:41:29,760 --> 00:41:32,610
S1: AI is the one going out and reading everything for you.

706
00:41:32,610 --> 00:41:36,029
S1: It's consuming everything. It's consuming all the APIs, all the

707
00:41:36,030 --> 00:41:40,830
S1: RSS feeds, all the videos. It's making transcripts. It's translating them,

708
00:41:40,830 --> 00:41:43,680
S1: it's summarizing them. It's doing everything for you. And as

709
00:41:43,680 --> 00:41:46,319
S1: you're moving through your day, when you have a little

710
00:41:46,350 --> 00:41:49,560
S1: bit of extra time, it'll create a custom piece of

711
00:41:49,560 --> 00:41:52,740
S1: content for you. Actually, I'm merging both of these. Cool.

712
00:41:52,739 --> 00:41:55,650
S1: It'll create a custom piece of content for you. Okay,

713
00:41:55,680 --> 00:41:59,310
S1: so there's this paper that I just summarized. Where is

714
00:41:59,310 --> 00:42:01,770
S1: that thing? Oh, it's right here. So my buddy just

715
00:42:01,770 --> 00:42:05,310
S1: sent me this thing. It's this giant PDF. Um, my

716
00:42:05,340 --> 00:42:07,230
S1: AI is going to go read this thing. It's going

717
00:42:07,260 --> 00:42:09,270
S1: to give me a summary based on how much time

718
00:42:09,270 --> 00:42:11,489
S1: I have before my next meeting. It's like, hey, you

719
00:42:11,489 --> 00:42:13,290
S1: want to hear about this thing? Or do you want

720
00:42:13,320 --> 00:42:15,450
S1: to see a video? I'm like, yeah, give me a video.

721
00:42:15,450 --> 00:42:19,830
S1: So it's Richard Feynman explaining this paper to me in 30s,

722
00:42:19,830 --> 00:42:23,340
S1: which means the content. Watch this. The content duration, the

723
00:42:23,340 --> 00:42:27,870
S1: content format, the content style, all of. Oh and the performer.

724
00:42:27,900 --> 00:42:31,080
S1: The person actually giving the content. So it could be

725
00:42:31,080 --> 00:42:34,169
S1: a graphic novel. It could be a video, it could

726
00:42:34,170 --> 00:42:37,590
S1: be an essay, it could be a video, it could

727
00:42:37,590 --> 00:42:40,470
S1: be text, it could be whatever, but it could shrink

728
00:42:40,469 --> 00:42:45,810
S1: from 30s down to 10s do a giant set of books,

729
00:42:45,810 --> 00:42:50,310
S1: and the AI will dynamically recreate that for you, which

730
00:42:50,310 --> 00:42:54,569
S1: means this is actually the future of education. And this

731
00:42:54,570 --> 00:42:57,750
S1: is going to be a whole separate conversation. In fact,

732
00:42:57,750 --> 00:42:59,790
S1: we're going to break this out as a separate clip

733
00:42:59,790 --> 00:43:02,100
S1: to talk about this. This is going to be the

734
00:43:02,100 --> 00:43:05,100
S1: future of education, because think about this. The AI can

735
00:43:05,100 --> 00:43:10,379
S1: also adjust based on somebody's limitations. Okay. They haven't learned

736
00:43:10,380 --> 00:43:12,810
S1: this topic yet. They're not so great at English yet.

737
00:43:12,810 --> 00:43:15,690
S1: They have a learning disability okay. The AI is going

738
00:43:15,719 --> 00:43:18,180
S1: to be an expert in learning disabilities, which means it

739
00:43:18,180 --> 00:43:21,060
S1: can repeat things, or it could avoid certain ways of

740
00:43:21,060 --> 00:43:23,969
S1: explaining things, or it could double down on other ways

741
00:43:23,969 --> 00:43:27,930
S1: of explaining things. Dynamic adjustment to the best way to

742
00:43:27,960 --> 00:43:31,290
S1: teach you at that given time. Maybe if you're a kid,

743
00:43:31,290 --> 00:43:33,840
S1: you want to hear this lesson as a story. You

744
00:43:33,840 --> 00:43:36,000
S1: want to watch it as a video. The whole thing

745
00:43:36,030 --> 00:43:40,050
S1: could be framed as Taylor Swift or Pokémon or Star Wars. Okay,

746
00:43:40,080 --> 00:43:43,350
S1: so your eye is going to be able to shape

747
00:43:43,380 --> 00:43:48,270
S1: teaching you anything using any format, in any duration, in

748
00:43:48,270 --> 00:43:52,440
S1: any style of teaching. Okay. It could be a podcast

749
00:43:52,440 --> 00:43:55,020
S1: between your two favorite people, and they're talking about the

750
00:43:55,020 --> 00:43:58,319
S1: ideas debating back and forth. It could be a graphic novel,

751
00:43:58,320 --> 00:44:00,719
S1: it could be a TV show, it could be a cartoon,

752
00:44:00,719 --> 00:44:02,850
S1: it could be an anime. It could be a YouTube

753
00:44:02,850 --> 00:44:07,140
S1: video of your favorite creator just doing a YouTube video

754
00:44:07,140 --> 00:44:09,270
S1: for you about that thing, and you're like, nope, I

755
00:44:09,270 --> 00:44:12,270
S1: want it in 30s nope. I want it in one

756
00:44:12,270 --> 00:44:15,750
S1: hour episodes, five of them per week. So it's like

757
00:44:15,750 --> 00:44:17,819
S1: I sit down for an hour and I learn this thing.

758
00:44:17,820 --> 00:44:22,799
S1: This is going to massively disrupt education. Okay, we already

759
00:44:22,800 --> 00:44:25,739
S1: knew I was going to disrupt education. We already know

760
00:44:25,770 --> 00:44:28,890
S1: technology is going to disrupt education. Like this is all known.

761
00:44:28,890 --> 00:44:32,370
S1: But this particular formula of being able to adjust all

762
00:44:32,370 --> 00:44:37,620
S1: those variables and create dynamic content, that is the game

763
00:44:37,620 --> 00:44:42,990
S1: that that is the education game, like 100%. Okay. Now

764
00:44:42,989 --> 00:44:45,420
S1: now think about this. All you need to do is

765
00:44:45,420 --> 00:44:48,180
S1: come up with the absolute best content, okay? Once you

766
00:44:48,180 --> 00:44:50,850
S1: have the best content, which which I hate to tell you,

767
00:44:50,850 --> 00:44:53,190
S1: it's kind of already out there. Okay. What what are

768
00:44:53,190 --> 00:44:58,440
S1: the best concepts in journalism or biology or whatever? It's

769
00:44:58,440 --> 00:45:01,739
S1: in there. It's in the textbooks already. Right? The smartest

770
00:45:01,739 --> 00:45:03,870
S1: people have already been talking about it. It's in all

771
00:45:03,900 --> 00:45:07,620
S1: of their books. Well, go read their books. We've already

772
00:45:07,620 --> 00:45:10,229
S1: seen their lectures. Half of those lectures are already on

773
00:45:10,230 --> 00:45:15,839
S1: open MOOCs on, uh, you know, Harvard or MIT or whatever. Like,

774
00:45:15,840 --> 00:45:18,770
S1: these people aren't being quiet. It's not a secret what

775
00:45:18,770 --> 00:45:21,049
S1: they know about the world. So all the stuff's out there,

776
00:45:21,050 --> 00:45:24,050
S1: so the content's solved. The only question is, how are

777
00:45:24,050 --> 00:45:27,560
S1: we getting that content to the people? Now, you take

778
00:45:27,560 --> 00:45:31,370
S1: something like an extremely expensive university like Harvard or Stanford

779
00:45:31,370 --> 00:45:35,270
S1: or something like 11 people go per year and it's

780
00:45:35,270 --> 00:45:38,509
S1: horribly expensive. How many people are on the planet? Oh,

781
00:45:38,540 --> 00:45:42,590
S1: 8 billion, 8 billion people need that education and don't

782
00:45:42,590 --> 00:45:46,820
S1: have the time. And they have language, you know, disabilities

783
00:45:46,820 --> 00:45:51,020
S1: or language like, um, mismatches with the way the original

784
00:45:51,020 --> 00:45:55,400
S1: content was taught. Doesn't matter. Instant translation to their language

785
00:45:55,400 --> 00:45:58,040
S1: while on ramping them onto English if they want to

786
00:45:58,040 --> 00:46:02,120
S1: learn English. 8 billion people getting exactly the education they

787
00:46:02,120 --> 00:46:05,719
S1: need in the topics that they need in any format

788
00:46:05,719 --> 00:46:09,259
S1: that is best for them, represented by deepfakes of exactly

789
00:46:09,260 --> 00:46:12,950
S1: the best people or in the format of whatever a show,

790
00:46:12,980 --> 00:46:14,750
S1: whatever is the best way to get it to them.

791
00:46:14,750 --> 00:46:18,189
S1: Like how is that not the way people are going

792
00:46:18,190 --> 00:46:20,410
S1: to learn in the future? That's what I'm asking you.

793
00:46:20,440 --> 00:46:24,040
S1: How is that not the future? Now, again, how fast

794
00:46:24,070 --> 00:46:25,960
S1: are we going to get there? Are there lots of

795
00:46:25,960 --> 00:46:30,430
S1: different ways that will be opposed? Of course, of course. Right.

796
00:46:30,460 --> 00:46:34,030
S1: So when is it going to start? Today. Yesterday. Tomorrow.

797
00:46:34,060 --> 00:46:36,730
S1: That's when it's going to start. It's already starting. When

798
00:46:36,730 --> 00:46:39,430
S1: will it be finished? Never. There will always be a

799
00:46:39,430 --> 00:46:41,620
S1: forces that push it back and say no, you've got

800
00:46:41,620 --> 00:46:44,259
S1: to dress up. You've got to go to this special school.

801
00:46:44,290 --> 00:46:47,169
S1: Because remember, a lot of education is actually not about

802
00:46:47,170 --> 00:46:49,870
S1: the content and about the the learning. A lot of

803
00:46:49,870 --> 00:46:52,750
S1: the education is actually about the stamp that you get afterwards,

804
00:46:52,750 --> 00:46:55,330
S1: which we can solve a different way, which we'll be

805
00:46:55,330 --> 00:46:58,270
S1: talking about at some point to and kind of already

806
00:46:58,270 --> 00:47:00,489
S1: have talked about. All right. Let's keep going. I think

807
00:47:00,489 --> 00:47:03,730
S1: we covered both of those topics pretty well. Oh, yeah.

808
00:47:04,120 --> 00:47:09,160
S1: Future of education. Exactly. Okay. Discovery acts framework extension of

809
00:47:09,160 --> 00:47:13,750
S1: Axios for bug hunters and penetration testers. So it's basically

810
00:47:13,750 --> 00:47:17,350
S1: a way to do large distributed scans as part of

811
00:47:17,380 --> 00:47:20,739
S1: like Bug Bounty, the Russian APT tool matrix. New tool

812
00:47:20,739 --> 00:47:26,680
S1: matrix focused on Russian APT groups GRU, SVR, FSB affiliated

813
00:47:26,680 --> 00:47:31,719
S1: threat actors. Reuse of tools like Mimikatz, impact, exact, Metasploit

814
00:47:31,719 --> 00:47:36,340
S1: Re George Re George being particularly notable for limited use

815
00:47:36,340 --> 00:47:40,960
S1: by by ransomware gangs okay Merkle map subdomain search engine

816
00:47:40,960 --> 00:47:45,070
S1: helps you discover subdomains. Gun gear is a command line tool.

817
00:47:45,070 --> 00:47:48,070
S1: What's with this font? Do Not like is a command

818
00:47:48,070 --> 00:47:51,430
S1: line tool written by my buddy gunner and go. Microservices

819
00:47:51,430 --> 00:47:55,180
S1: are technical debt. Really really cool. Take there. Be somebody

820
00:47:55,180 --> 00:47:59,770
S1: who does things. That was a fantastic essay. It's lists

821
00:47:59,800 --> 00:48:03,310
S1: all the way down. Another fantastic one. US government is

822
00:48:03,310 --> 00:48:07,569
S1: now publishing tide sensor data to show sea level rise.

823
00:48:07,600 --> 00:48:11,740
S1: Love the transparency here. Comedy is search, critical mass and

824
00:48:11,739 --> 00:48:16,870
S1: tipping points. Semantic chunking for rag flipper zero. Firmware. Open

825
00:48:16,900 --> 00:48:20,170
S1: source firmware unlocks a lot more of your flipper device.

826
00:48:20,200 --> 00:48:24,250
S1: MTA Open Data Challenge Paul Graham addresses the question of

827
00:48:24,250 --> 00:48:28,540
S1: whether to follow your passion or not. Compressed jpg, compressed

828
00:48:28,540 --> 00:48:33,880
S1: jpeg move fast and abandon things. BBC Sound Effects Library

829
00:48:33,910 --> 00:48:36,940
S1: what I learned in the past years spent building an

830
00:48:36,940 --> 00:48:40,900
S1: AI video editor. Danger zone is teamed up with Google's.

831
00:48:40,930 --> 00:48:44,950
S1: Gvisor to enhance its security, allowing journalists to open suspicious

832
00:48:44,950 --> 00:48:48,640
S1: documents safely. RM Creative Console might be getting one of

833
00:48:48,640 --> 00:48:54,160
S1: these actually need to go order that advanced structure output tutorial.

834
00:48:54,160 --> 00:48:58,540
S1: Logitech's MX Creative Console is a game changer for photo

835
00:48:58,540 --> 00:49:02,620
S1: and video editors. Dieter Rams inspired iPhone dock. This thing

836
00:49:02,650 --> 00:49:04,359
S1: looks really cool. I don't think I would get one,

837
00:49:04,360 --> 00:49:07,990
S1: but isn't that cool? Yeah, that looks cool. And automatic

838
00:49:07,989 --> 00:49:10,660
S1: litter boxes because I want a cat, but I'm allergic

839
00:49:10,660 --> 00:49:13,780
S1: to cats, so probably not getting one recommendation of the week.

840
00:49:13,810 --> 00:49:16,629
S1: Take a document or something you know really well and

841
00:49:16,630 --> 00:49:19,960
S1: upload it to notebook LM and tinker around with it.

842
00:49:19,989 --> 00:49:22,840
S1: It is super cool and do it quickly because they

843
00:49:22,840 --> 00:49:27,130
S1: might put the project in the Google graveyard. Um. Aphorism

844
00:49:27,130 --> 00:49:30,160
S1: of the week. I want so much that is not

845
00:49:30,160 --> 00:49:33,009
S1: here and do not know where to go. I want

846
00:49:33,010 --> 00:49:35,710
S1: so much that is not here and do not know

847
00:49:35,710 --> 00:49:41,230
S1: where to go. Charles Bukowski. Unsupervised learning is produced and

848
00:49:41,230 --> 00:49:44,800
S1: edited by Daniel Miessler on a Neumann U87 AI microphone

849
00:49:44,800 --> 00:49:48,760
S1: using Hindenburg. Intro and outro music is by zombie with

850
00:49:48,760 --> 00:49:51,580
S1: a Y. And to get the text and links from

851
00:49:51,580 --> 00:49:53,710
S1: this episode, sign up for the newsletter version of the

852
00:49:53,710 --> 00:49:59,290
S1: show at Daniel missler.com/newsletter. We'll see you next time.