1
00:00:01,480 --> 00:00:07,200
Speaker 1: Welcome to Stuff You Should Know, a production of iHeartRadio.

2
00:00:11,160 --> 00:00:13,920
Speaker 2: Hey, and welcome to the podcast. I'm Josh, and there's

3
00:00:14,080 --> 00:00:17,880
Speaker 2: Chuck and Jerry's here too, And that makes this a

4
00:00:18,480 --> 00:00:22,759
Speaker 2: timely topical Not timely is in fork times. I meant

5
00:00:22,760 --> 00:00:27,159
Speaker 2: to say timely topical episode of Stuff you should Know.

6
00:00:27,760 --> 00:00:29,520
Speaker 2: That's a great forecast of.

7
00:00:29,480 --> 00:00:32,120
Speaker 1: How it's going to go to I think fork cast.

8
00:00:33,720 --> 00:00:39,559
Speaker 1: Oh God, is this really us or is it AI

9
00:00:39,800 --> 00:00:41,520
Speaker 1: generated Josh and Chuck.

10
00:00:42,479 --> 00:00:48,159
Speaker 2: Until this year I would have been like, don't be preposterous.

11
00:00:48,240 --> 00:00:50,280
Speaker 2: Now I'm like, just give it some time.

12
00:00:51,080 --> 00:00:53,200
Speaker 1: You know how we would know as if it said

13
00:00:53,240 --> 00:00:56,200
Speaker 1: if one of us said, of course it's the real Us.

14
00:00:56,520 --> 00:00:59,480
Speaker 1: We met in the office and bonded over our Van

15
00:00:59,560 --> 00:01:01,560
Speaker 1: Halen denim vests.

16
00:01:01,840 --> 00:01:04,640
Speaker 2: Yeah, we'd be like, sucker, you just fell for the

17
00:01:04,680 --> 00:01:07,440
Speaker 2: oldest trap in the book, the Sicilian switcher.

18
00:01:08,160 --> 00:01:11,520
Speaker 1: Yeah, aka fake Wikipedia entry stuff?

19
00:01:12,080 --> 00:01:12,880
Speaker 2: Is that still up?

20
00:01:13,840 --> 00:01:16,120
Speaker 1: I haven't been to our Wikipedia page in years, so

21
00:01:16,200 --> 00:01:16,640
Speaker 1: I don't know.

22
00:01:17,480 --> 00:01:19,920
Speaker 2: Well, regardless, we're not talking about Wikipedia, although it does

23
00:01:20,040 --> 00:01:22,440
Speaker 2: kind of fall into the sure it figures in the

24
00:01:22,520 --> 00:01:25,360
Speaker 2: rubric of this. Not sure if I use that word correctly,

25
00:01:25,440 --> 00:01:29,800
Speaker 2: but it felt right. We're talking today about what are

26
00:01:30,480 --> 00:01:34,679
Speaker 2: in the biz known as large language models, but more

27
00:01:34,760 --> 00:01:38,480
Speaker 2: colloquially known by basically their their public facing names, things

28
00:01:38,480 --> 00:01:44,360
Speaker 2: like chat, GPT or barred or being AI. But essentially

29
00:01:44,400 --> 00:01:49,920
Speaker 2: what they all are are algorithms, artificially intelligent algorithms that

30
00:01:50,440 --> 00:01:54,720
Speaker 2: are trained on text, tons and tons and tons of

31
00:01:54,760 --> 00:02:00,200
Speaker 2: text written English language stuff, that are so good at

32
00:02:00,280 --> 00:02:04,200
Speaker 2: recognizing patterns in those things that they can actually simulate

33
00:02:04,280 --> 00:02:07,080
Speaker 2: a conversation with you, the person on the other side

34
00:02:07,080 --> 00:02:09,320
Speaker 2: of the computer, asking them questions.

35
00:02:10,120 --> 00:02:13,400
Speaker 1: Yeah, this is it's gonna be fun doing this episode

36
00:02:13,880 --> 00:02:19,120
Speaker 1: over every six months, right until we're replaced totally.

37
00:02:19,600 --> 00:02:21,520
Speaker 2: So I think we should say, though, like this is

38
00:02:21,600 --> 00:02:24,320
Speaker 2: we're going to like this is such a huge wide

39
00:02:24,960 --> 00:02:30,120
Speaker 2: topic that is just we're in the ignition phase, like

40
00:02:30,120 --> 00:02:34,480
Speaker 2: the fuse just caught, right, Yeah, that we're going to

41
00:02:34,560 --> 00:02:36,640
Speaker 2: really try to keep it narrow just strictly to large

42
00:02:36,680 --> 00:02:39,760
Speaker 2: language models and the immediate effect they're planning on having

43
00:02:39,880 --> 00:02:42,480
Speaker 2: or they're going to have, hopefully not planning on anything yet,

44
00:02:43,680 --> 00:02:46,160
Speaker 2: but I really would like to do one on how

45
00:02:46,240 --> 00:02:50,320
Speaker 2: to keep AI friendly and keeping it from running away. Yeah,

46
00:02:50,480 --> 00:02:52,600
Speaker 2: I say, we just kind of avoid that whole kind

47
00:02:52,639 --> 00:02:55,040
Speaker 2: of stuff. And really I'm talking to myself right now

48
00:02:55,320 --> 00:02:56,520
Speaker 2: at least for this episode.

49
00:02:56,560 --> 00:02:58,800
Speaker 1: Okay, Yeah, you know, we're going to kind of explain

50
00:02:59,320 --> 00:03:03,079
Speaker 1: how these things were and what the initial applications look

51
00:03:03,240 --> 00:03:05,040
Speaker 1: like and kind of where we are right now, and

52
00:03:05,080 --> 00:03:08,240
Speaker 1: then what it could mean for like jobs and the

53
00:03:08,280 --> 00:03:10,600
Speaker 1: economy and stuff like that. But you're right, it is

54
00:03:11,919 --> 00:03:13,600
Speaker 1: as a whole ball of wax, as you well know.

55
00:03:14,320 --> 00:03:16,560
Speaker 1: And this is a great time to plug the End

56
00:03:16,560 --> 00:03:19,000
Speaker 1: of the World with Josh Clark, which is still out there.

57
00:03:19,040 --> 00:03:19,960
Speaker 1: You can still listen to it.

58
00:03:20,600 --> 00:03:22,960
Speaker 2: The truth is out there in the form of the

59
00:03:23,040 --> 00:03:24,440
Speaker 2: End of the World with Josh Clark.

60
00:03:24,919 --> 00:03:28,680
Speaker 1: Yeah, it's a great ten part series that you did,

61
00:03:29,240 --> 00:03:34,400
Speaker 1: and AI is among those existential risks existential that you covered.

62
00:03:34,920 --> 00:03:38,040
Speaker 2: Yeah, it's episode four, I believe. And Chuck like just

63
00:03:38,120 --> 00:03:40,920
Speaker 2: from having done that research in forming my own opinions

64
00:03:40,920 --> 00:03:46,120
Speaker 2: over the years about this. Yeah, Like I'm it's staggering

65
00:03:46,160 --> 00:03:50,000
Speaker 2: to me that we're like we've just entered like what's

66
00:03:50,040 --> 00:03:58,440
Speaker 2: going to be the most revolutionary transitional phase in the

67
00:03:58,720 --> 00:04:02,120
Speaker 2: entire history of humanity. You can argue everything else took

68
00:04:02,160 --> 00:04:05,080
Speaker 2: place over very long periods of time. We started playing

69
00:04:05,120 --> 00:04:07,280
Speaker 2: with stone tools and then we started building cities. All

70
00:04:07,320 --> 00:04:09,960
Speaker 2: this stuff took place over thousands and thousands, hundreds of thousands,

71
00:04:10,000 --> 00:04:13,120
Speaker 2: millions of years. We just entered a period where stuff's

72
00:04:13,160 --> 00:04:17,120
Speaker 2: going to start happening within weeks pretty soon. As of

73
00:04:17,120 --> 00:04:19,200
Speaker 2: twenty twenty three, the whole thing just started.

74
00:04:19,680 --> 00:04:22,680
Speaker 1: Yeah, and none of this was around like this when

75
00:04:22,720 --> 00:04:24,520
Speaker 1: you did The End of the World and that was

76
00:04:24,880 --> 00:04:26,240
Speaker 1: what like five years ago.

77
00:04:26,320 --> 00:04:28,679
Speaker 2: Is Yeah, it was twenty eighteen. All this was being

78
00:04:28,720 --> 00:04:31,040
Speaker 2: worked on, but we hadn't hit that point. Like all

79
00:04:31,080 --> 00:04:33,960
Speaker 2: this was pretty much predicted and projected, and it was

80
00:04:34,200 --> 00:04:36,039
Speaker 2: clear that this was the direction people were going.

81
00:04:36,120 --> 00:04:37,640
Speaker 1: And it's here, baby, it is.

82
00:04:37,920 --> 00:04:40,760
Speaker 2: It's nuts, but it's actually here. So we're talking about

83
00:04:40,760 --> 00:04:44,400
Speaker 2: our large language models, which is a type of neural

84
00:04:44,480 --> 00:04:48,440
Speaker 2: network that are easiest to think of in terms of

85
00:04:48,880 --> 00:04:51,600
Speaker 2: like a human brain, where you have neurons that are

86
00:04:51,600 --> 00:04:53,960
Speaker 2: connected to other neurons, but they're not connected to some

87
00:04:54,080 --> 00:04:58,080
Speaker 2: other neurons, and all of those neural connections kind of

88
00:04:58,440 --> 00:05:02,800
Speaker 2: are activated by input. It's that put out something like

89
00:05:02,839 --> 00:05:05,400
Speaker 2: your conscious experience or you say a sentence or something

90
00:05:05,440 --> 00:05:09,880
Speaker 2: like that. It's it's very similar in its most basic nature.

91
00:05:10,120 --> 00:05:13,720
Speaker 1: I guess, yeah, I mean Olivia helped us out with this,

92
00:05:13,839 --> 00:05:17,120
Speaker 1: and che did a great job. I think and Google

93
00:05:17,160 --> 00:05:20,400
Speaker 1: themselves basically say, you know what it's really it's sort

94
00:05:20,400 --> 00:05:22,839
Speaker 1: of like how when you go to search for something

95
00:05:22,960 --> 00:05:28,520
Speaker 1: on our search engine tool here a weird way I

96
00:05:28,520 --> 00:05:31,120
Speaker 1: could have said that, I don't think so are handy

97
00:05:31,160 --> 00:05:34,320
Speaker 1: search for Then you know, basically what we're doing is

98
00:05:34,400 --> 00:05:40,080
Speaker 1: auto completing, like an analysis of like probability, like like

99
00:05:40,200 --> 00:05:44,479
Speaker 1: what you're typing. If you type in you know, John Coltrane,

100
00:05:44,520 --> 00:05:47,600
Speaker 1: or start to chop type in John Cole, that might

101
00:05:47,640 --> 00:05:51,120
Speaker 1: finish it out as John Coltrane a Love Supreme or

102
00:05:51,200 --> 00:05:54,719
Speaker 1: John Coltrane jazz and they're saying, you know what, what

103
00:05:54,839 --> 00:05:59,920
Speaker 1: is happening now with these lms is it's the same thing.

104
00:06:00,080 --> 00:06:03,279
Speaker 1: It's just it's got way more data, way more calculations

105
00:06:03,880 --> 00:06:07,039
Speaker 1: in the algorithm. So it's not just completing like a

106
00:06:07,080 --> 00:06:11,160
Speaker 1: word or two. It's potentially you know, hey, rewrite the

107
00:06:11,200 --> 00:06:12,760
Speaker 1: Bible or whatever you tell it to do.

108
00:06:13,279 --> 00:06:16,279
Speaker 2: Yeah, And the big difference is in the amount of

109
00:06:16,560 --> 00:06:23,479
Speaker 2: info the neural network is capable of taking into consideration. Yeah,

110
00:06:23,520 --> 00:06:26,880
Speaker 2: so may I for a minute, oh please, So imagine

111
00:06:26,880 --> 00:06:29,599
Speaker 2: with one of those autocomplete suggestion tools like they have

112
00:06:29,760 --> 00:06:34,359
Speaker 2: on Google Search. If there's five hundred thousand words in

113
00:06:34,400 --> 00:06:37,159
Speaker 2: the English language, that means that you have five hundred

114
00:06:37,200 --> 00:06:40,640
Speaker 2: thousand words that a person could possibly put in. That's

115
00:06:40,680 --> 00:06:44,159
Speaker 2: the input into the neural network, and then there's five

116
00:06:44,240 --> 00:06:47,599
Speaker 2: hundred thousand possible words that that network could put out.

117
00:06:47,640 --> 00:06:50,440
Speaker 2: So you have five hundred thousand connections to five hundred

118
00:06:50,440 --> 00:06:53,159
Speaker 2: thousand other connections. So it's like, I think two hundred

119
00:06:53,160 --> 00:06:56,119
Speaker 2: and fifty billion connections you're starting with right there. That's

120
00:06:56,279 --> 00:07:00,520
Speaker 2: just the autocomplete suggestion because it based on those connections.

121
00:07:00,560 --> 00:07:03,480
Speaker 2: In studying words in the English language and phrases in

122
00:07:03,520 --> 00:07:07,480
Speaker 2: the English language, it places emphasis more on some connections

123
00:07:07,520 --> 00:07:12,800
Speaker 2: than others. So John Coltrane is what's what's his album

124
00:07:12,800 --> 00:07:16,400
Speaker 2: I Can't Remember is a Love Supreme classic album. So

125
00:07:16,520 --> 00:07:20,240
Speaker 2: John Coltrane is much more closely related to a Love

126
00:07:20,280 --> 00:07:23,800
Speaker 2: Supreme in the mind of a neural network than John Coltrane.

127
00:07:25,720 --> 00:07:29,560
Speaker 2: Charlie Brown disco is right, just to take something off

128
00:07:29,600 --> 00:07:32,320
Speaker 2: the top of my head. And so based on that

129
00:07:32,480 --> 00:07:35,200
Speaker 2: analysis and that weight that it gives to some things

130
00:07:35,200 --> 00:07:38,400
Speaker 2: other than others, it suggests those words what the large

131
00:07:38,480 --> 00:07:41,560
Speaker 2: language models that like chat GPT, that we're seeing today.

132
00:07:42,400 --> 00:07:45,080
Speaker 2: They do the same thing. They have all those same connections,

133
00:07:45,480 --> 00:07:48,080
Speaker 2: but the analysis they do, the weight that they put

134
00:07:48,120 --> 00:07:52,560
Speaker 2: on the connections is so much more advanced, yes, and

135
00:07:52,720 --> 00:07:57,000
Speaker 2: exponential that it's it's actually not just capable of suggesting

136
00:07:57,040 --> 00:08:00,000
Speaker 2: the next word, it's capable of holding a conversation with you.

137
00:08:00,080 --> 00:08:04,480
Speaker 2: That's how much it understands how the English language works.

138
00:08:04,920 --> 00:08:08,360
Speaker 1: Yeah, Like if you said, you know, write a story

139
00:08:08,400 --> 00:08:14,320
Speaker 1: about wintertime, and you know, it got to the word snowy,

140
00:08:14,400 --> 00:08:17,160
Speaker 1: it would it would go through you know, I mean,

141
00:08:17,200 --> 00:08:20,600
Speaker 1: and this is like instantaneously it's doing these calculations. It

142
00:08:20,680 --> 00:08:24,200
Speaker 1: might say like you know, oh, hillside or winter or

143
00:08:24,240 --> 00:08:26,280
Speaker 1: snowy day, like these are all things that make sense

144
00:08:26,680 --> 00:08:29,680
Speaker 1: because I've learned that that makes sense I being you know,

145
00:08:29,800 --> 00:08:33,760
Speaker 1: the chatbot or whatever. But it probably won't be snowy

146
00:08:33,880 --> 00:08:37,599
Speaker 1: chicken wing because that doesn't seem to fit the algorithm.

147
00:08:38,120 --> 00:08:43,080
Speaker 1: And it learns all this stuff by reading the internet.

148
00:08:43,360 --> 00:08:46,680
Speaker 1: And you know, put a pin in that, because that's

149
00:08:46,720 --> 00:08:51,040
Speaker 1: pretty thorny for a whole lot of reasons, but not

150
00:08:51,120 --> 00:08:53,480
Speaker 1: the least of which is the fact that some companies,

151
00:08:53,520 --> 00:08:55,040
Speaker 1: and again we'll get to it, are starting to say

152
00:08:55,040 --> 00:08:57,440
Speaker 1: like wait a minute, Like, we created this content and

153
00:08:57,480 --> 00:09:02,320
Speaker 1: now you're just scrubbing it and then using it and

154
00:09:02,400 --> 00:09:05,240
Speaker 1: charging people to use it, and we're not getting a

155
00:09:05,240 --> 00:09:07,880
Speaker 1: piece of it. So that's just one tiny little thorn.

156
00:09:08,840 --> 00:09:10,920
Speaker 1: But in order to do this, like you said, it's

157
00:09:10,960 --> 00:09:13,760
Speaker 1: like it needs to know more, and you came up

158
00:09:13,800 --> 00:09:17,240
Speaker 1: with a great example, like the word lemon. In a

159
00:09:17,440 --> 00:09:21,760
Speaker 1: very basic way, it might understand that a lemon is

160
00:09:22,679 --> 00:09:26,280
Speaker 1: roundish and sour and yellow. But if it needs to

161
00:09:26,280 --> 00:09:29,199
Speaker 1: get smart enough to really write as if it were

162
00:09:29,240 --> 00:09:31,760
Speaker 1: a human, it needs to know that it can make lemonade,

163
00:09:31,800 --> 00:09:35,680
Speaker 1: and that it grows on a tree in these agricultural zones,

164
00:09:35,720 --> 00:09:38,040
Speaker 1: and that it's a citrus fruit, because it has to

165
00:09:38,040 --> 00:09:43,200
Speaker 1: be able to group lemon together with like things. And

166
00:09:43,280 --> 00:09:46,439
Speaker 1: those groups are either like, you know, hey, it's super

167
00:09:46,440 --> 00:09:49,960
Speaker 1: similar to this, like maybe other citrus fruits, or it's

168
00:09:50,000 --> 00:09:51,839
Speaker 1: you know, sort of similar to this but not as

169
00:09:51,840 --> 00:09:55,760
Speaker 1: similar as citrus fruits, like desserts. And then you get

170
00:09:55,760 --> 00:09:58,560
Speaker 1: to chicken wings, although actually that's not true because lemon

171
00:09:58,800 --> 00:09:59,560
Speaker 1: chicken wings.

172
00:10:00,120 --> 00:10:01,520
Speaker 2: You could have lemon pepper chicken wings.

173
00:10:01,559 --> 00:10:03,720
Speaker 1: Right, Yeah, that's what I'm saying, So yeah, Kau, But

174
00:10:04,040 --> 00:10:06,880
Speaker 1: the instance you use is like greenland, which I guess

175
00:10:06,880 --> 00:10:08,360
Speaker 1: doesn't grow lemons.

176
00:10:08,559 --> 00:10:10,800
Speaker 2: No, but I mean, I'm sure they import lemons, so

177
00:10:10,800 --> 00:10:14,720
Speaker 2: there's some connection there. But based on how connected, how

178
00:10:14,800 --> 00:10:17,880
Speaker 2: often these words show up together, and the billions and

179
00:10:17,920 --> 00:10:21,440
Speaker 2: billions of lines of texts that these language large language

180
00:10:21,440 --> 00:10:25,440
Speaker 2: models are trained on, it starts to get more and

181
00:10:25,440 --> 00:10:29,360
Speaker 2: more dimensions and making more and more connections. Right. So

182
00:10:30,400 --> 00:10:34,080
Speaker 2: as that happens, words start to cluster together, like lemon

183
00:10:34,240 --> 00:10:36,880
Speaker 2: and pie and ice box all kind of cluster together.

184
00:10:37,440 --> 00:10:42,800
Speaker 2: And by taking words and understanding how they connect to

185
00:10:42,880 --> 00:10:46,280
Speaker 2: other words, you can take the English language, just the

186
00:10:46,280 --> 00:10:48,760
Speaker 2: words of the English language, and make meaning out of it.

187
00:10:48,880 --> 00:10:51,920
Speaker 2: That's what that's all we do. And large language models

188
00:10:51,960 --> 00:10:54,800
Speaker 2: are capable of doing the same thing. But it's really

189
00:10:55,000 --> 00:10:59,640
Speaker 2: really important for you to understand that the large language

190
00:10:59,679 --> 00:11:04,880
Speaker 2: model doesn't understand what it's doing. It doesn't have any

191
00:11:04,920 --> 00:11:10,280
Speaker 2: meaning to the word lemon whatsoever. All of these dimensions

192
00:11:10,320 --> 00:11:13,679
Speaker 2: that it waits to decide whether what word it should

193
00:11:13,760 --> 00:11:18,800
Speaker 2: use next, they're called embeddings. They're just numerical representations. Yeah,

194
00:11:18,800 --> 00:11:21,040
Speaker 2: So the higher the number, the likelier it is it

195
00:11:21,080 --> 00:11:23,320
Speaker 2: goes with the word that the user just put in

196
00:11:23,400 --> 00:11:26,679
Speaker 2: or that the large language model. Just use the lower

197
00:11:26,720 --> 00:11:29,360
Speaker 2: the number, the further away it is in the cluster, right,

198
00:11:30,000 --> 00:11:33,480
Speaker 2: it doesn't understand what it's saying to you. And as

199
00:11:33,520 --> 00:11:36,240
Speaker 2: we'll see later, that accounts for a phenomenon that we're

200
00:11:36,280 --> 00:11:38,440
Speaker 2: going to have to overcome for them to get smarter,

201
00:11:38,520 --> 00:11:42,240
Speaker 2: which is called hallucinations. But that's a really critically important

202
00:11:42,280 --> 00:11:43,000
Speaker 2: thing to remember.

203
00:11:43,679 --> 00:11:48,000
Speaker 1: Yeah, Another critically important thing to remember is and you

204
00:11:48,040 --> 00:11:49,920
Speaker 1: probably get this from what we said so far if

205
00:11:49,920 --> 00:11:51,640
Speaker 1: you already know about a little bit about it, But

206
00:11:51,679 --> 00:11:55,960
Speaker 1: there's no programmer that's teaching these things and typing in

207
00:11:56,080 --> 00:11:59,840
Speaker 1: inputs and then saying here's how you learn things. Like

208
00:12:00,520 --> 00:12:03,640
Speaker 1: it's doing this on its own, and it's learning things

209
00:12:03,640 --> 00:12:07,160
Speaker 1: on its own, and what we're talking about eventually like that.

210
00:12:07,360 --> 00:12:09,719
Speaker 1: You know, where it could get super scary is when

211
00:12:09,760 --> 00:12:12,480
Speaker 1: it gets to what's called emergent abilities, where it's so

212
00:12:12,679 --> 00:12:18,160
Speaker 1: powerful and there's so much data that the nuance that's

213
00:12:18,200 --> 00:12:21,319
Speaker 1: missing now will be there right exactly.

214
00:12:21,640 --> 00:12:25,319
Speaker 2: So, yeah, that's when things are going to get even

215
00:12:25,400 --> 00:12:29,040
Speaker 2: harder to understand, you know, to remind yourself that you're

216
00:12:29,080 --> 00:12:30,320
Speaker 2: talking to a machine, you know.

217
00:12:31,120 --> 00:12:32,880
Speaker 1: Yeah, And the other thing too, though, even though I

218
00:12:32,920 --> 00:12:36,600
Speaker 1: said humans aren't in putting this data. One of the

219
00:12:36,640 --> 00:12:39,720
Speaker 1: big things that is allowing this stuff to get smarter

220
00:12:40,600 --> 00:12:45,200
Speaker 1: is human feedback. It's called r l HF, which is

221
00:12:45,760 --> 00:12:48,960
Speaker 1: reinforcement learning on human feedback. So at the end of

222
00:12:49,000 --> 00:12:52,960
Speaker 1: your whatever you've told it to create, you can go

223
00:12:53,080 --> 00:12:55,120
Speaker 1: back in and say, well, you got this wrong and

224
00:12:55,120 --> 00:12:57,600
Speaker 1: this wrong, this is what that really is, and it says,

225
00:12:58,040 --> 00:12:59,960
Speaker 1: thank you, I have now just gotten smart.

226
00:13:01,360 --> 00:13:04,079
Speaker 2: Exactly. So one of the reasons why these things are

227
00:13:04,080 --> 00:13:07,200
Speaker 2: suddenly just so smart and can say thank you, I've

228
00:13:07,240 --> 00:13:10,079
Speaker 2: just gotten so much smarter is because of a paper

229
00:13:10,440 --> 00:13:16,360
Speaker 2: that Google engineers published openly in twenty seventeen describing what's

230
00:13:16,440 --> 00:13:19,959
Speaker 2: now like the essential ingredient for a large language model

231
00:13:20,000 --> 00:13:22,520
Speaker 2: or probably any neural network from now on. It's called

232
00:13:22,520 --> 00:13:27,280
Speaker 2: a transformer. And rather than analyzing each bit of text,

233
00:13:27,360 --> 00:13:30,200
Speaker 2: let's say you say one of the very famous things.

234
00:13:30,800 --> 00:13:33,319
Speaker 2: Marvin Minsky was one of the founders of the field

235
00:13:33,320 --> 00:13:38,200
Speaker 2: of AI, and his son Henry prompted chet Gpt to

236
00:13:38,360 --> 00:13:41,600
Speaker 2: describe what losing a sock in the dryer is like

237
00:13:41,640 --> 00:13:45,120
Speaker 2: in the style of the Declaration of Independence. Right, So

238
00:13:45,240 --> 00:13:51,040
Speaker 2: depending on how Henry Minsky type that in before transformers

239
00:13:51,360 --> 00:13:55,080
Speaker 2: the neural network would analyze each word and do it

240
00:13:55,240 --> 00:13:57,920
Speaker 2: one increment and a time, maybe not even words, sometimes

241
00:13:57,960 --> 00:14:02,400
Speaker 2: strings of just letters together, phone names even if you

242
00:14:02,440 --> 00:14:07,319
Speaker 2: can believe it, phone names even And what the transformer

243
00:14:07,360 --> 00:14:09,559
Speaker 2: does is it changes that. It allows it to analyze

244
00:14:09,600 --> 00:14:12,960
Speaker 2: everything all at once. So it's so much faster, not

245
00:14:13,160 --> 00:14:16,320
Speaker 2: just in putting out a coherent answer to your question

246
00:14:16,520 --> 00:14:20,560
Speaker 2: or request, but in also training itself on that text.

247
00:14:20,600 --> 00:14:23,840
Speaker 2: So you just feed it the internet and it starts

248
00:14:23,880 --> 00:14:27,600
Speaker 2: analyzing it and self correcting. It trains itself, It learns

249
00:14:27,640 --> 00:14:32,360
Speaker 2: on its own, and that unfortunately also makes AI, including

250
00:14:32,440 --> 00:14:35,920
Speaker 2: large language models what are known as black boxes. Yeah,

251
00:14:35,960 --> 00:14:38,840
Speaker 2: we don't know how they're doing what they're doing. We

252
00:14:38,880 --> 00:14:40,560
Speaker 2: have a good idea how to make them do the

253
00:14:40,600 --> 00:14:43,800
Speaker 2: things we want, but the in between stuff, we cannot

254
00:14:43,800 --> 00:14:47,000
Speaker 2: one hundred percent say what they're doing. How they come

255
00:14:47,120 --> 00:14:51,480
Speaker 2: up with these conclusions, which also explains hallucinations in them

256
00:14:51,520 --> 00:14:53,000
Speaker 2: not really making sense to us.

257
00:14:53,560 --> 00:14:57,360
Speaker 1: Yeah, and you know, the T in GPT stands for transformer.

258
00:14:57,520 --> 00:15:02,880
Speaker 1: It's generative pre trained transformer. And the reason they call

259
00:15:02,920 --> 00:15:05,680
Speaker 1: it GPT for short is because if they call it

260
00:15:05,760 --> 00:15:10,000
Speaker 1: generative pre trained transformer, everybody would be scared out of

261
00:15:10,040 --> 00:15:11,000
Speaker 1: their minds.

262
00:15:10,880 --> 00:15:14,359
Speaker 2: We just start running around to nowhere in particular.

263
00:15:14,840 --> 00:15:17,840
Speaker 1: Yeah, should we take a break, I say, we do.

264
00:15:17,960 --> 00:15:20,560
Speaker 2: I think that we kind of explained that fairly well.

265
00:15:20,600 --> 00:15:50,640
Speaker 1: Yeah, fairly robust beginning, my friend. All right, So open

266
00:15:50,680 --> 00:15:54,760
Speaker 1: Ai launched their chat GPT and very recently in November

267
00:15:54,760 --> 00:15:58,600
Speaker 1: of twenty twenty two, and just in that brief window

268
00:15:59,160 --> 00:16:02,240
Speaker 1: was like six or eight months ago. Things are kind

269
00:16:02,280 --> 00:16:06,400
Speaker 1: of flying high, and all kinds of companies are launching

270
00:16:06,560 --> 00:16:10,320
Speaker 1: their own stuff. Some of it is well. First of all,

271
00:16:10,320 --> 00:16:14,160
Speaker 1: open ai is now at chat GPT four, yes, and

272
00:16:14,200 --> 00:16:17,640
Speaker 1: I'm sure you know more will be coming in quick succession.

273
00:16:18,320 --> 00:16:20,920
Speaker 1: But companies are launching, and we're going to talk about

274
00:16:20,920 --> 00:16:23,120
Speaker 1: all of them kind of like broad stuff like chat,

275
00:16:23,160 --> 00:16:28,280
Speaker 1: BT GPT and really specific stuff like well, hey, I'm

276
00:16:28,320 --> 00:16:31,600
Speaker 1: in the banking business. Can we just design something for

277
00:16:31,680 --> 00:16:34,960
Speaker 1: banking or just something for real estate? So they're also

278
00:16:35,000 --> 00:16:38,640
Speaker 1: getting specific on a smaller level in addition to these

279
00:16:38,720 --> 00:16:41,640
Speaker 1: large like Google and Microsoft and Being and all that stuff.

280
00:16:41,920 --> 00:16:44,160
Speaker 2: Yeah, And to get specific, all you have to do

281
00:16:44,240 --> 00:16:49,640
Speaker 2: is take an existing GPT, a large language model, and

282
00:16:50,240 --> 00:16:54,520
Speaker 2: add some software that helps guide it a little more. Yeah,

283
00:16:54,600 --> 00:16:57,440
Speaker 2: and there you go or just train it on specific

284
00:16:57,480 --> 00:17:01,800
Speaker 2: stuff like medical notes. That's another one one of the

285
00:17:01,800 --> 00:17:05,159
Speaker 2: other things that's that's changed very quickly between November of

286
00:17:05,200 --> 00:17:07,600
Speaker 2: twenty twenty two in March of twenty twenty three, when

287
00:17:08,200 --> 00:17:13,440
Speaker 2: I think GPT four became available. Just think about that.

288
00:17:13,440 --> 00:17:16,240
Speaker 2: That's that's such a short amount of time. Yeah, all

289
00:17:16,240 --> 00:17:19,760
Speaker 2: of a sudden. Now you can take a picture and

290
00:17:19,800 --> 00:17:22,480
Speaker 2: feed it into a large language model and it will

291
00:17:22,840 --> 00:17:26,720
Speaker 2: describe the picture. It will look at the picture essentially

292
00:17:26,960 --> 00:17:31,120
Speaker 2: and describe what's going on. There's a there's a demonstration

293
00:17:31,280 --> 00:17:34,440
Speaker 2: from one of the guys from open Ai who doodles

294
00:17:34,440 --> 00:17:37,600
Speaker 2: like on a little like scrapbook piece of paper some

295
00:17:37,760 --> 00:17:41,120
Speaker 2: ideas for a website. He takes a picture of that

296
00:17:41,240 --> 00:17:45,359
Speaker 2: paper that he's written on, feeds it into chat GPT four,

297
00:17:46,359 --> 00:17:49,080
Speaker 2: and it builds a website for him in a couple

298
00:17:49,119 --> 00:17:51,480
Speaker 2: of minutes. That functions the way he was thinking of

299
00:17:51,800 --> 00:17:53,760
Speaker 2: the on the doodle scratch pad.

300
00:17:54,240 --> 00:17:55,920
Speaker 1: I wonder if the only way to slow this stuff

301
00:17:55,960 --> 00:17:59,280
Speaker 1: down is to literally slow down the Internet again. Go

302
00:17:59,359 --> 00:18:02,160
Speaker 1: back to like the old days when a picture would

303
00:18:02,200 --> 00:18:05,399
Speaker 1: load like three lines at a time, right, and they

304
00:18:05,440 --> 00:18:09,760
Speaker 1: describe a picture will be like someone's hair, someone's nose,

305
00:18:10,560 --> 00:18:14,320
Speaker 1: someone's gin. Don't forget between Yeah, an hour later you

306
00:18:14,320 --> 00:18:15,160
Speaker 1: have a complete picture.

307
00:18:15,320 --> 00:18:17,359
Speaker 2: Right. I don't think there's any way to slow this

308
00:18:17,440 --> 00:18:21,800
Speaker 2: down because we're in not to be alarmist, but we're

309
00:18:21,800 --> 00:18:26,760
Speaker 2: in a second worst case scenario for introducing AI to

310
00:18:26,840 --> 00:18:30,400
Speaker 2: the world, which is, rather than state actors doing this,

311
00:18:30,640 --> 00:18:34,680
Speaker 2: which would be really bad, we have private companies doing it,

312
00:18:34,720 --> 00:18:38,879
Speaker 2: which is just slightly less bad. But they're competing in

313
00:18:38,920 --> 00:18:42,480
Speaker 2: an arms race to get the best, brightest, smartest AI

314
00:18:42,640 --> 00:18:45,600
Speaker 2: out there as fast as they can, and they're not

315
00:18:45,680 --> 00:18:48,760
Speaker 2: taking into account like all of the downsides to it.

316
00:18:48,760 --> 00:18:50,800
Speaker 2: They're just throwing it out there as much as they can.

317
00:18:51,040 --> 00:18:53,639
Speaker 2: Because one of the ways that these things get smarter

318
00:18:53,960 --> 00:18:56,760
Speaker 2: is by interacting with the public. They get better and

319
00:18:56,800 --> 00:19:01,280
Speaker 2: better at what they do from getting feedback from people

320
00:19:01,400 --> 00:19:02,280
Speaker 2: using them.

321
00:19:02,520 --> 00:19:04,960
Speaker 1: Yeah, even if it's for just some goofy fun thing

322
00:19:05,000 --> 00:19:08,440
Speaker 1: you're doing, it's learning from that. And you talked about

323
00:19:08,440 --> 00:19:12,560
Speaker 1: the advancements made between the launch of three point five

324
00:19:12,680 --> 00:19:17,560
Speaker 1: and GPT four, and three point five scored in the

325
00:19:17,600 --> 00:19:21,960
Speaker 1: tenth percentile when it took the uniform bar exam, and

326
00:19:22,320 --> 00:19:26,159
Speaker 1: four has already scored in the ninetieth percentile, and they

327
00:19:26,200 --> 00:19:29,399
Speaker 1: found that chet GPT four is really it's great at

328
00:19:29,800 --> 00:19:34,040
Speaker 1: taking tests, and it's scoring really well on tests, particularly

329
00:19:34,440 --> 00:19:37,760
Speaker 1: you know, standardized tests. All. I think it basically aced

330
00:19:38,000 --> 00:19:41,520
Speaker 1: all of the AP tests that you would take to

331
00:19:41,520 --> 00:19:45,960
Speaker 1: get into AP classes except well, it took a couple

332
00:19:45,960 --> 00:19:49,359
Speaker 1: of AP class but the max score is five, and

333
00:19:49,440 --> 00:19:52,480
Speaker 1: I think it got five's kind of on everything except

334
00:19:52,560 --> 00:19:59,160
Speaker 1: for math. It got a four and math it's kind

335
00:19:59,160 --> 00:20:02,560
Speaker 1: of it's we it's kind of weirdly counterintuitive because it's

336
00:20:02,600 --> 00:20:05,919
Speaker 1: a number space thing, but it has more trouble with math,

337
00:20:06,640 --> 00:20:10,040
Speaker 1: like rudimentary math, than it does with like constructing a

338
00:20:10,080 --> 00:20:13,720
Speaker 1: paragraph on you know, Shakespeare or something, or as Shakespeare

339
00:20:14,800 --> 00:20:18,320
Speaker 1: does better with like math word problems and more advanced

340
00:20:18,359 --> 00:20:20,760
Speaker 1: math than it does just at basic math apparently, or.

341
00:20:20,760 --> 00:20:25,440
Speaker 2: Like describing how a formula functions using you know, writing.

342
00:20:26,200 --> 00:20:28,320
Speaker 2: The thing is though, and this is another great example

343
00:20:28,320 --> 00:20:31,440
Speaker 2: of how fast this is moving. They've already figured out

344
00:20:31,440 --> 00:20:33,040
Speaker 2: that all you have to do is do what's called

345
00:20:33,040 --> 00:20:38,760
Speaker 2: prompting where you where you basically take the answer that

346
00:20:38,840 --> 00:20:42,960
Speaker 2: the the incorrect answer that the large language model gives you,

347
00:20:43,320 --> 00:20:45,840
Speaker 2: and then basically re explain it by breaking it down

348
00:20:45,880 --> 00:20:48,760
Speaker 2: into different parts, and it learns as you're doing that,

349
00:20:48,880 --> 00:20:50,560
Speaker 2: and then all of a sudden it comes up with

350
00:20:50,760 --> 00:20:53,479
Speaker 2: it gets better at math. So they've figured out tools,

351
00:20:53,520 --> 00:20:57,800
Speaker 2: extra software you can lay over a GPT that basically

352
00:20:58,200 --> 00:21:00,959
Speaker 2: teach it to do math or prompt it in the

353
00:21:01,000 --> 00:21:03,160
Speaker 2: correct way so that you get the answer you're looking

354
00:21:03,200 --> 00:21:04,640
Speaker 2: for that's based on math.

355
00:21:05,520 --> 00:21:08,720
Speaker 1: Yeah. I mean, every time I read something that said, well,

356
00:21:08,800 --> 00:21:11,240
Speaker 1: right now, it's not so great at this, I just

357
00:21:11,359 --> 00:21:14,119
Speaker 1: assume that meant and we'll have that worked out in

358
00:21:14,160 --> 00:21:15,119
Speaker 1: the next few weeks.

359
00:21:15,320 --> 00:21:17,880
Speaker 2: Yeah, pretty much. I mean, because as these things get

360
00:21:18,760 --> 00:21:21,040
Speaker 2: like bigger and smarter, in the data sets that they're

361
00:21:21,040 --> 00:21:24,080
Speaker 2: trained on get wider, they're just going to get better

362
00:21:24,119 --> 00:21:26,480
Speaker 2: and better at this because they again they learned from

363
00:21:26,480 --> 00:21:27,280
Speaker 2: their mistakes.

364
00:21:29,800 --> 00:21:34,439
Speaker 1: Yeah, just like humans, right, exactly like humans. So you

365
00:21:34,480 --> 00:21:38,000
Speaker 1: mentioned these hallucinations kind of briefly, and this this is

366
00:21:38,040 --> 00:21:40,000
Speaker 1: one of the big problems with them so far that

367
00:21:40,560 --> 00:21:42,720
Speaker 1: again I'm sure they will figure this out in due time.

368
00:21:43,119 --> 00:21:47,480
Speaker 1: But one example that Livia found was to prompt it

369
00:21:47,520 --> 00:21:51,359
Speaker 1: with what mammal lays the largest eggs. And one of

370
00:21:51,400 --> 00:21:54,720
Speaker 1: the problems is when it gives hallucinations or wrong answers it.

371
00:21:55,160 --> 00:21:57,520
Speaker 1: You know, it's not saying like, well, I'm not so

372
00:21:57,600 --> 00:22:00,800
Speaker 1: sure about this. It's saying this is true, just like

373
00:22:00,840 --> 00:22:03,879
Speaker 1: anything else. I'm spitting out right with a lot of confidence.

374
00:22:03,880 --> 00:22:06,280
Speaker 1: So the answer there was the mammal that lays the

375
00:22:06,359 --> 00:22:08,720
Speaker 1: largest eggs is the elephant, and elephant's eggs are so

376
00:22:08,760 --> 00:22:11,320
Speaker 1: small that they are often visible to the naked eye,

377
00:22:11,520 --> 00:22:14,359
Speaker 1: so they're not commonly known to lay eggs at all. However,

378
00:22:14,400 --> 00:22:17,040
Speaker 1: in terms of sheer size, and elephant's eggs are the

379
00:22:17,160 --> 00:22:19,040
Speaker 1: largest of any mammal.

380
00:22:18,800 --> 00:22:20,760
Speaker 2: Which makes sense in a really weird way if you

381
00:22:20,760 --> 00:22:21,800
Speaker 2: think about it.

382
00:22:21,920 --> 00:22:23,399
Speaker 1: Sure, those little invisible eggs.

383
00:22:23,480 --> 00:22:26,400
Speaker 2: Yeah, because mammals don't lay eggs obviously, But the way

384
00:22:26,440 --> 00:22:28,560
Speaker 2: that it put it was if you didn't know that

385
00:22:28,600 --> 00:22:31,200
Speaker 2: mammals don't lay eggs, or you didn't know anything about elephants,

386
00:22:31,520 --> 00:22:33,679
Speaker 2: you'd be like, oh, that's interesting, and take that as

387
00:22:33,680 --> 00:22:36,480
Speaker 2: a fact, because it's saying this confidently. And I saw

388
00:22:36,760 --> 00:22:40,200
Speaker 2: written somewhere that one GPT actually argued with the user

389
00:22:40,240 --> 00:22:41,919
Speaker 2: and told them they were wrong when they told the

390
00:22:41,960 --> 00:22:46,119
Speaker 2: GPRIT that it was wrong, Yeah, which is not a

391
00:22:46,160 --> 00:22:50,239
Speaker 2: behavior you want at all. But that's what's termed as

392
00:22:50,240 --> 00:22:53,560
Speaker 2: a hallucination, and a hallucination is a good way to

393
00:22:53,800 --> 00:23:00,600
Speaker 2: understand it. That I saw is that again, this GPT,

394
00:23:00,720 --> 00:23:04,200
Speaker 2: this large language model, doesn't have any idea what it's

395
00:23:04,200 --> 00:23:09,640
Speaker 2: saying means. It's just picked up it's noticed patterns that

396
00:23:09,720 --> 00:23:13,000
Speaker 2: we've not noticed before, and it's putting them together in

397
00:23:13,119 --> 00:23:17,280
Speaker 2: nonsensical ways. But they're still sensible if you read them.

398
00:23:17,320 --> 00:23:20,600
Speaker 2: It's just factually they're not sensible because it doesn't have

399
00:23:20,640 --> 00:23:24,600
Speaker 2: any fact checking necessarily, it just knows what it's finding

400
00:23:25,440 --> 00:23:28,080
Speaker 2: kind of correlates with other things. So there's some sensible

401
00:23:28,119 --> 00:23:31,560
Speaker 2: stuff in there, like the phrase invisible to the naked eye,

402
00:23:32,000 --> 00:23:35,959
Speaker 2: or laying eggs or elephants in mammals. Like this stuff

403
00:23:36,280 --> 00:23:38,840
Speaker 2: all makes sense. It's not like these are just strings

404
00:23:38,840 --> 00:23:42,520
Speaker 2: of letters. Yeah, it's just putting them together in ways

405
00:23:42,560 --> 00:23:46,440
Speaker 2: that are not true. They're factually incorrect, and that's a hallucination.

406
00:23:46,560 --> 00:23:50,840
Speaker 2: It's not like the computer is thinking that this is true.

407
00:23:50,880 --> 00:23:57,560
Speaker 2: It doesn't understand things like truth in falsehood. It just creates,

408
00:23:58,000 --> 00:24:00,440
Speaker 2: and some of the time it gets it really wrong.

409
00:24:00,920 --> 00:24:02,439
Speaker 1: Yeah, I didn't know what an elephant is.

410
00:24:03,160 --> 00:24:06,400
Speaker 2: No, it just knows that it correlates to in some

411
00:24:06,640 --> 00:24:11,320
Speaker 2: really small way. That we've never noticed before the word eggs.

412
00:24:11,560 --> 00:24:15,280
Speaker 1: Yeah, and this is uh that that's a problem if

413
00:24:15,440 --> 00:24:17,440
Speaker 1: if it's just like, oh, well, this thing isn't quite

414
00:24:17,440 --> 00:24:19,400
Speaker 1: where it needs to be at because it thinks elephants

415
00:24:19,440 --> 00:24:22,280
Speaker 1: lay eggs. But there have already been plenty of real

416
00:24:22,440 --> 00:24:25,800
Speaker 1: world examples where people are using this and it's screwing

417
00:24:25,840 --> 00:24:28,160
Speaker 1: things up for their business or for commerce or something

418
00:24:28,240 --> 00:24:31,200
Speaker 1: where the yeah, or their client, well, that's that's one.

419
00:24:31,240 --> 00:24:35,440
Speaker 1: There was an attorney who was representing a passenger who

420
00:24:35,520 --> 00:24:39,720
Speaker 1: was suing an airline and used chat bt to do research,

421
00:24:40,320 --> 00:24:42,640
Speaker 1: and it came up with a bunch of fake cases

422
00:24:42,960 --> 00:24:46,560
Speaker 1: that this attorney didn't bother to fact check, I guess.

423
00:24:47,160 --> 00:24:50,199
Speaker 1: And there were like a dozen fake cases that this

424
00:24:50,280 --> 00:24:51,840
Speaker 1: attorney submitted in his brief.

425
00:24:52,080 --> 00:24:54,360
Speaker 2: And it wasn't like so like from what I understand,

426
00:24:54,359 --> 00:24:57,800
Speaker 2: like the the brief was largely compiled from what the

427
00:24:57,960 --> 00:25:01,480
Speaker 2: GPT spit out. The It wasn't like the GPT just

428
00:25:01,520 --> 00:25:03,240
Speaker 2: made up the names of cases. It made up the

429
00:25:03,280 --> 00:25:06,119
Speaker 2: names of cases and then described the background of the

430
00:25:06,160 --> 00:25:09,560
Speaker 2: case and how they related to the case at hand, right,

431
00:25:09,640 --> 00:25:12,440
Speaker 2: So it just completely made these up out of out

432
00:25:12,440 --> 00:25:15,359
Speaker 2: of the blue. And yeah, that lawyer had no idea.

433
00:25:15,400 --> 00:25:17,800
Speaker 2: He said in a brief later that he had no

434
00:25:17,880 --> 00:25:20,439
Speaker 2: idea that this thing was capable of being incorrect. So

435
00:25:20,760 --> 00:25:22,399
Speaker 2: it was like one of the first times he used it,

436
00:25:22,480 --> 00:25:24,280
Speaker 2: and he threw himself on the mercy of the court.

437
00:25:24,600 --> 00:25:26,520
Speaker 2: And I'm not quite sure exactly what happened. I think

438
00:25:26,520 --> 00:25:28,440
Speaker 2: they're still figuring out what to do about it.

439
00:25:29,480 --> 00:25:31,679
Speaker 1: Maybe just go spend some quality time with your little

440
00:25:31,800 --> 00:25:33,919
Speaker 1: chat butt exactly.

441
00:25:34,960 --> 00:25:39,600
Speaker 2: Similarly, Meta had a large language model that basically got

442
00:25:39,680 --> 00:25:42,040
Speaker 2: laughed off of the Internet because it was very science

443
00:25:42,040 --> 00:25:45,800
Speaker 2: focused and it would make up things that just didn't exist,

444
00:25:45,840 --> 00:25:49,879
Speaker 2: like mathematical formula, like there was one that called the

445
00:25:50,000 --> 00:25:54,399
Speaker 2: yoko or no, the lenin Ono correlation or something like

446
00:25:54,440 --> 00:25:57,480
Speaker 2: that completely made up this thing that I read, and

447
00:25:57,520 --> 00:25:59,960
Speaker 2: I was like, oh, that's interesting. I had no idea.

448
00:26:00,320 --> 00:26:02,840
Speaker 2: I have never heard this stuff before, and I would

449
00:26:02,880 --> 00:26:04,479
Speaker 2: have just thought that it was real had I not

450
00:26:04,520 --> 00:26:07,080
Speaker 2: realized and known ahead of time that it was a

451
00:26:07,080 --> 00:26:10,800
Speaker 2: hallucination that this math thing does not exist anywhere. And

452
00:26:10,800 --> 00:26:15,080
Speaker 2: it even attributed it to a live mathematician said that

453
00:26:15,160 --> 00:26:18,479
Speaker 2: this was the guy who discovered it. So like, it

454
00:26:18,560 --> 00:26:22,320
Speaker 2: really can get hard to discern what's true and what's not,

455
00:26:22,640 --> 00:26:25,440
Speaker 2: which again is a really big problem if we haven't

456
00:26:25,440 --> 00:26:26,439
Speaker 2: gotten that cross yet.

457
00:26:26,680 --> 00:26:33,560
Speaker 1: Did they say that mathematician's name was math be calculus. Yeah.

458
00:26:33,640 --> 00:26:35,639
Speaker 1: Another example, and this is, you know, we're going to

459
00:26:35,680 --> 00:26:38,600
Speaker 1: talk a little bit about you know, replacing jobs in

460
00:26:38,640 --> 00:26:42,679
Speaker 1: the various ways that can and already is happening. But

461
00:26:42,760 --> 00:26:46,120
Speaker 1: seeing that, for instance, said oh, you know what, let

462
00:26:46,119 --> 00:26:47,479
Speaker 1: me try this thing out and see if we can

463
00:26:47,480 --> 00:26:51,200
Speaker 1: get it to write an actual story. And so they

464
00:26:51,240 --> 00:26:54,199
Speaker 1: got an AI tool to write one on what is

465
00:26:54,240 --> 00:26:57,239
Speaker 1: compound interest, and it was just there was a lot

466
00:26:57,280 --> 00:27:01,040
Speaker 1: of stuff wrong in it. There was some plagiarism, you know,

467
00:27:01,280 --> 00:27:04,359
Speaker 1: directly lifted. So there's you know, these things aren't fool

468
00:27:04,359 --> 00:27:06,800
Speaker 1: proof yet, and it's definitely not something that should be

469
00:27:08,080 --> 00:27:12,040
Speaker 1: utilized for like a public facing website that's supposed to

470
00:27:12,680 --> 00:27:17,719
Speaker 1: have like really solid vetted articles about uh well, especially

471
00:27:17,760 --> 00:27:21,080
Speaker 1: seen it about a tech right of all things.

472
00:27:21,320 --> 00:27:24,840
Speaker 2: That's something that the National Eating Disorder Association found out

473
00:27:24,840 --> 00:27:30,200
Speaker 2: the hard way. They apparently replaced entirely it's human staffed

474
00:27:30,240 --> 00:27:34,520
Speaker 2: hotline with the chatbot, and supposedly they were accused of

475
00:27:34,520 --> 00:27:37,320
Speaker 2: doing this to bust the union that had formed there

476
00:27:38,520 --> 00:27:41,200
Speaker 2: and so when they released the chatpot into the world

477
00:27:41,440 --> 00:27:44,639
Speaker 2: and it started offering advice to people suffering from eating disorders.

478
00:27:44,880 --> 00:27:49,000
Speaker 2: It gave standard, you know, weight loss advice, which you

479
00:27:49,119 --> 00:27:51,840
Speaker 2: probably get from your doctor who didn't realize you had

480
00:27:51,840 --> 00:27:54,480
Speaker 2: an eating disorder, but in the context of an eating disorder,

481
00:27:54,720 --> 00:27:57,720
Speaker 2: it was all like trigger, trigger, trigger, one right after

482
00:27:57,800 --> 00:28:00,240
Speaker 2: the other. Right, Like, it was telling these people with

483
00:28:00,280 --> 00:28:03,000
Speaker 2: eating disorders to like, weigh yourself every week and try

484
00:28:03,040 --> 00:28:05,399
Speaker 2: to cut out five hundred to one thousand calories a

485
00:28:05,480 --> 00:28:08,879
Speaker 2: day and you'll lose some weight, and just stuff that

486
00:28:08,880 --> 00:28:11,320
Speaker 2: that would set everybody off. And very quickly they took

487
00:28:11,359 --> 00:28:14,959
Speaker 2: it offline and I guess brought their humans back, hopefully

488
00:28:15,000 --> 00:28:15,880
Speaker 2: at double the pay.

489
00:28:16,920 --> 00:28:19,920
Speaker 1: Yeah. But I mean this stuff is that's already being

490
00:28:19,960 --> 00:28:22,560
Speaker 1: solved as well, because they point out that GPT four

491
00:28:23,359 --> 00:28:27,240
Speaker 1: has already scored forty percent higher than three point five,

492
00:28:27,880 --> 00:28:32,159
Speaker 1: again just a handful of months ago on these accuracy tests,

493
00:28:32,720 --> 00:28:36,800
Speaker 1: so that that is even getting better. And you know,

494
00:28:36,840 --> 00:28:38,800
Speaker 1: where where I guess people want it to get to

495
00:28:39,760 --> 00:28:42,680
Speaker 1: is to the point where it doesn't need human supervision

496
00:28:42,760 --> 00:28:45,800
Speaker 1: to spit out really really accurate stuff exactly.

497
00:28:45,920 --> 00:28:48,479
Speaker 2: That's pretty much where they're hoping to get it. And

498
00:28:48,520 --> 00:28:51,600
Speaker 2: I mean it's just they have the model, they have

499
00:28:51,840 --> 00:28:53,800
Speaker 2: everything they need. They just it just has to be

500
00:28:53,840 --> 00:28:54,360
Speaker 2: tinkered with.

501
00:28:54,440 --> 00:28:58,440
Speaker 1: Now, should we take another break? I think so, all right,

502
00:28:58,480 --> 00:29:00,880
Speaker 1: we'll take another break and then get into sort of

503
00:29:00,920 --> 00:29:04,040
Speaker 1: the the economics of it and whether or not your

504
00:29:04,200 --> 00:29:05,680
Speaker 1: job may be at risk right after this.

505
00:29:29,640 --> 00:29:33,240
Speaker 2: So one of the astounding things about this that it

506
00:29:33,360 --> 00:29:37,320
Speaker 2: really caught everybody off guard is that these large language models,

507
00:29:37,800 --> 00:29:41,520
Speaker 2: the jobs they're coming after are white collar knowledge jobs.

508
00:29:42,320 --> 00:29:45,920
Speaker 2: They're so good at things like writing, they're good at researching,

509
00:29:45,960 --> 00:29:50,200
Speaker 2: they're good at analyzing photos. Now and that's a huge

510
00:29:50,680 --> 00:29:53,760
Speaker 2: sea change from what it's been like traditionally. Right wherever

511
00:29:54,000 --> 00:29:57,920
Speaker 2: whenever we've automated things, it's usually replaced manual labor. Now

512
00:29:57,920 --> 00:30:00,400
Speaker 2: it's the manual labor that's safe in this Yeah, this

513
00:30:00,640 --> 00:30:05,200
Speaker 2: generation of automation, it's the white collar knowledge shops that

514
00:30:05,280 --> 00:30:08,200
Speaker 2: are at risk. And not just white collar job but

515
00:30:08,480 --> 00:30:12,320
Speaker 2: artists in yeah, like just who have nothing to do

516
00:30:12,360 --> 00:30:16,480
Speaker 2: with white collar or jobs, they're at risk as well.

517
00:30:17,200 --> 00:30:20,880
Speaker 1: Yeah, I'm sure the farmers are all sitting around going,

518
00:30:21,480 --> 00:30:22,280
Speaker 1: how's that going for you?

519
00:30:22,360 --> 00:30:24,680
Speaker 2: Yeah, how's that taste? Uh?

520
00:30:24,880 --> 00:30:28,160
Speaker 1: So? Yeah, art art is when when d A L L. E.

521
00:30:28,320 --> 00:30:31,040
Speaker 1: Doll E came out, that was an art tool where

522
00:30:31,880 --> 00:30:33,280
Speaker 1: a lot of people, a lot of people I know,

523
00:30:33,360 --> 00:30:36,120
Speaker 1: would input. I guess I never did it. I never

524
00:30:36,280 --> 00:30:38,920
Speaker 1: do anything like that, not because I'm afraid or anything,

525
00:30:39,000 --> 00:30:44,160
Speaker 1: but I just just not interested basically. But I guess

526
00:30:44,160 --> 00:30:46,160
Speaker 1: you would submit, like a photograph of yourself and then

527
00:30:46,160 --> 00:30:48,120
Speaker 1: it would say, well, here's you as a superhero or

528
00:30:48,160 --> 00:30:52,800
Speaker 1: here's you as a Renaissance painting or whatever. And you know,

529
00:30:52,920 --> 00:30:57,280
Speaker 1: it's sourcing images from real artists throughout history, from Getty

530
00:30:57,320 --> 00:31:00,280
Speaker 1: Images and places like that, and there are are ready

531
00:31:00,360 --> 00:31:04,880
Speaker 1: artists that are suing for infringement. Getty Images is suing

532
00:31:04,920 --> 00:31:09,520
Speaker 1: for infringement and saying you can't even if you're mixing

533
00:31:09,640 --> 00:31:13,120
Speaker 1: up things and it's not like a Rembrandt. Let's say

534
00:31:13,560 --> 00:31:16,239
Speaker 1: you're using all of the artists from that era and

535
00:31:16,360 --> 00:31:19,600
Speaker 1: mashing it up together in a way that, like we

536
00:31:19,720 --> 00:31:21,160
Speaker 1: think basically is illegal.

537
00:31:21,400 --> 00:31:25,080
Speaker 2: Yeah, they say this doesn't count as transformative use, which

538
00:31:25,120 --> 00:31:30,400
Speaker 2: is typically protected under the law. Right, this is instead

539
00:31:31,080 --> 00:31:34,640
Speaker 2: just some sort of mash up that that a machine

540
00:31:34,680 --> 00:31:39,040
Speaker 2: is doing. It's to me, it's almost splitting hairs. But

541
00:31:39,080 --> 00:31:41,560
Speaker 2: I also very much get where they're coming from not

542
00:31:41,600 --> 00:31:44,720
Speaker 2: just a place of panic, but like they're a real

543
00:31:46,280 --> 00:31:48,240
Speaker 2: like they have a basis in fact that these things

544
00:31:48,280 --> 00:31:51,640
Speaker 2: are not transforming because they don't understand what they're doing.

545
00:31:52,520 --> 00:31:56,840
Speaker 1: Yeah, and companies are taking notice very quickly. There are

546
00:31:56,840 --> 00:31:59,280
Speaker 1: some companies I'm sure everyone's going to kind of fall

547
00:31:59,320 --> 00:32:01,840
Speaker 1: in line, that are already saying, well, no, you got

548
00:32:01,880 --> 00:32:04,880
Speaker 1: to start paying us for access to this stuff. We

549
00:32:05,120 --> 00:32:09,400
Speaker 1: paid human beings to create this content for lack of

550
00:32:09,400 --> 00:32:12,760
Speaker 1: a better word, and put it online for people to access.

551
00:32:13,360 --> 00:32:15,440
Speaker 1: But you can't come in here now and access it

552
00:32:15,480 --> 00:32:18,960
Speaker 1: with a bot and use it and charge for it

553
00:32:19,400 --> 00:32:21,840
Speaker 1: without giving us a little juice. And there are a

554
00:32:21,880 --> 00:32:24,960
Speaker 1: lot of companies that are already saying like, you can't

555
00:32:25,120 --> 00:32:27,760
Speaker 1: use this. If you're an employee of our company, you

556
00:32:27,800 --> 00:32:32,080
Speaker 1: can't use chatbots at all because some of our company's

557
00:32:32,080 --> 00:32:37,480
Speaker 1: secrets might end up being spilled somehow, or you know,

558
00:32:37,520 --> 00:32:40,960
Speaker 1: our databases are all of a sudden exposed. So companies

559
00:32:40,960 --> 00:32:43,920
Speaker 1: are really moving fast to trying to protect their ip

560
00:32:44,280 --> 00:32:45,120
Speaker 1: I guess.

561
00:32:44,920 --> 00:32:48,440
Speaker 2: Well, yeah, and one of the I mean some of

562
00:32:48,480 --> 00:32:51,840
Speaker 2: the companies that are behind the GPTs that are out

563
00:32:51,920 --> 00:32:54,320
Speaker 2: right now, the large language models that are out right

564
00:32:54,360 --> 00:32:58,720
Speaker 2: now are well known for not only not protecting their

565
00:32:58,840 --> 00:33:03,280
Speaker 2: users information, but for rating it for its own use. Like,

566
00:33:03,360 --> 00:33:06,080
Speaker 2: for example, Meta is one of the ones with They

567
00:33:06,160 --> 00:33:11,280
Speaker 2: have their large language models called Lama, and there's a

568
00:33:11,360 --> 00:33:15,240
Speaker 2: chatbot called Alpaca. And it makes total sense that you

569
00:33:15,480 --> 00:33:18,920
Speaker 2: are probably signing away your right to protect your information

570
00:33:19,000 --> 00:33:21,520
Speaker 2: when you use those things on whatever computer you're using

571
00:33:21,520 --> 00:33:24,080
Speaker 2: it on or whatever network you're using it on. I

572
00:33:24,120 --> 00:33:27,000
Speaker 2: don't understand exactly. I haven't seen anything that says this

573
00:33:27,160 --> 00:33:29,160
Speaker 2: is how they're doing it, or even that they are

574
00:33:29,160 --> 00:33:32,640
Speaker 2: definitely doing this. I think it's just that the powers

575
00:33:32,680 --> 00:33:35,400
Speaker 2: that be no, like they would totally do this if

576
00:33:35,440 --> 00:33:37,400
Speaker 2: they can, and they probably are, so we should just

577
00:33:37,400 --> 00:33:40,440
Speaker 2: stay keep our employees away from it, you know, as

578
00:33:40,520 --> 00:33:41,200
Speaker 2: much as we can.

579
00:33:42,040 --> 00:33:46,080
Speaker 1: Yeah, it's like we said, it's being used on smaller

580
00:33:46,160 --> 00:33:50,600
Speaker 1: levels by One of the uses that Livia dug up

581
00:33:50,680 --> 00:33:53,200
Speaker 1: was like, let's say a real estate agent, instead of

582
00:33:53,200 --> 00:33:56,959
Speaker 1: taking time to write up listings, has a chatbot to it,

583
00:33:57,000 --> 00:33:59,400
Speaker 1: and then they can go through afterward and make adjustments

584
00:33:59,440 --> 00:34:01,080
Speaker 1: to it as needed.

585
00:34:01,360 --> 00:34:05,280
Speaker 2: Well, in exchange, that database now knows exactly what you

586
00:34:05,320 --> 00:34:07,480
Speaker 2: think of that one ugly bathroom.

587
00:34:07,440 --> 00:34:12,360
Speaker 1: That's right, or doctors may be using it to compile

588
00:34:12,440 --> 00:34:17,400
Speaker 1: lists of possible diseases or conditions that someone might have

589
00:34:17,440 --> 00:34:23,359
Speaker 1: based on symptoms. These all sound like uses that are like, hey,

590
00:34:23,640 --> 00:34:25,080
Speaker 1: this sounds like it could be a good thing in

591
00:34:25,080 --> 00:34:28,200
Speaker 1: some ways, and it can be in some ways. But

592
00:34:28,880 --> 00:34:31,400
Speaker 1: it's the wild West right now, so it's not like

593
00:34:31,520 --> 00:34:35,319
Speaker 1: there's any there's anyone saying, well, you can't use it

594
00:34:35,360 --> 00:34:37,160
Speaker 1: for that, you can only use it for this, you

595
00:34:37,200 --> 00:34:37,759
Speaker 1: know what I'm saying.

596
00:34:37,840 --> 00:34:40,560
Speaker 2: Plus, also, everything that we've come up with as just

597
00:34:41,080 --> 00:34:44,840
Speaker 2: Internet users in the general public has been what we

598
00:34:44,880 --> 00:34:48,400
Speaker 2: could come up with in given three months, with no

599
00:34:48,560 --> 00:34:51,399
Speaker 2: warning that we should start thinking about this. It's just like, hey,

600
00:34:51,440 --> 00:34:53,080
Speaker 2: this is here, what are you going to do with it?

601
00:34:53,120 --> 00:34:55,640
Speaker 2: And people are just finding new things to do with

602
00:34:55,680 --> 00:34:58,840
Speaker 2: it every day, And yeah, some of them are benign,

603
00:34:58,960 --> 00:35:02,880
Speaker 2: like having a draft a blog post for your business.

604
00:35:03,280 --> 00:35:05,279
Speaker 2: I thought they were already doing that based on some

605
00:35:05,360 --> 00:35:08,920
Speaker 2: of the emails that I get from like businesses, right, Yeah,

606
00:35:08,920 --> 00:35:11,719
Speaker 2: but they definitely are now if they weren't before. And

607
00:35:11,760 --> 00:35:16,000
Speaker 2: that's totally cool because there's there's a it's just taking

608
00:35:16,040 --> 00:35:18,960
Speaker 2: some of the weight off of the humans that are

609
00:35:19,000 --> 00:35:23,759
Speaker 2: already doing this. Work. Right, what's going to be problematic

610
00:35:24,080 --> 00:35:28,200
Speaker 2: is when it comes for the full job, or enough

611
00:35:28,239 --> 00:35:32,839
Speaker 2: of the job that the company can transfer whatever's left

612
00:35:32,840 --> 00:35:35,839
Speaker 2: of that person's job to other people and make them

613
00:35:35,920 --> 00:35:39,120
Speaker 2: just work a little harder while they're supported by the AI.

614
00:35:39,840 --> 00:35:43,400
Speaker 1: Yeah, here's some stats that they were pretty shocking to me.

615
00:35:43,480 --> 00:35:47,239
Speaker 1: I didn't know it was moving this fast. But there's

616
00:35:47,239 --> 00:35:50,400
Speaker 1: a networking app called Fishbowl and in twenty twenty three,

617
00:35:50,560 --> 00:35:54,520
Speaker 1: just earlier this year, they found that forty percent of

618
00:35:55,040 --> 00:35:58,759
Speaker 1: what they call working professionals are already using some kind

619
00:35:58,800 --> 00:36:01,799
Speaker 1: of either chat, GPT or some kind of AI tool

620
00:36:01,960 --> 00:36:07,200
Speaker 1: while they work, whether it's generating idealists or brainstorming lists,

621
00:36:07,320 --> 00:36:11,799
Speaker 1: or actually writing stuff or maybe looking at code. And

622
00:36:12,120 --> 00:36:15,080
Speaker 1: this is the troubling part. Forty of those forty percent,

623
00:36:16,000 --> 00:36:19,240
Speaker 1: almost seventy percent are doing that in secret and hadn't

624
00:36:19,239 --> 00:36:20,920
Speaker 1: told their bosses that they were doing that.

625
00:36:21,040 --> 00:36:23,360
Speaker 2: Right, those are just working professionals. We haven't even started

626
00:36:23,360 --> 00:36:24,600
Speaker 2: talking about students yet.

627
00:36:25,200 --> 00:36:27,680
Speaker 1: Yeah. I mean you combine that with work from home,

628
00:36:27,760 --> 00:36:29,200
Speaker 1: you got a real racket going on.

629
00:36:29,719 --> 00:36:33,120
Speaker 2: For sure, you know. Yeah, no, totally again though, I mean,

630
00:36:33,160 --> 00:36:35,200
Speaker 2: like if you can use it to do good work,

631
00:36:35,239 --> 00:36:38,040
Speaker 2: and you can now do more work. I think you

632
00:36:38,040 --> 00:36:40,880
Speaker 2: should be paid for more work. Like if your productivity's

633
00:36:40,920 --> 00:36:44,239
Speaker 2: gone through the roof, great, you figured it out. I've

634
00:36:44,280 --> 00:36:46,799
Speaker 2: got no problem with that. It's the opposite that I

635
00:36:46,800 --> 00:36:47,600
Speaker 2: have the problem with.

636
00:36:48,280 --> 00:36:50,920
Speaker 1: Well, let's skip students for a second and then and

637
00:36:51,000 --> 00:36:54,160
Speaker 1: talk about that since you brought it up, Because here's

638
00:36:54,239 --> 00:36:59,840
Speaker 1: the thing this is the United States doesn't have a

639
00:36:59,840 --> 00:37:04,040
Speaker 1: G eight track record of ignoring the bottom line in

640
00:37:04,080 --> 00:37:09,839
Speaker 1: favor of just keeping hardworking humans at their jobs. So

641
00:37:10,920 --> 00:37:13,239
Speaker 1: I think it was a Goldman. Sachs said that they

642
00:37:13,320 --> 00:37:17,840
Speaker 1: found that there could actually be an increase in the

643
00:37:17,880 --> 00:37:21,160
Speaker 1: annual GDP by about seven percent over ten years because

644
00:37:21,680 --> 00:37:26,200
Speaker 1: productivity increases. And I guess the idea is that productivity

645
00:37:26,280 --> 00:37:29,720
Speaker 1: is increasing because let's say you've got twenty to thirty

646
00:37:29,719 --> 00:37:32,680
Speaker 1: percent of stuff being done by AI. That opens up

647
00:37:32,680 --> 00:37:36,000
Speaker 1: twenty to thirty percent of your time for your employees

648
00:37:36,040 --> 00:37:41,720
Speaker 1: to maybe innovate or you know, get do other capitalistic things.

649
00:37:42,960 --> 00:37:45,240
Speaker 1: But what it to me, and this is just my opinion,

650
00:37:45,280 --> 00:37:48,000
Speaker 1: and again we're really early in all this, but it's

651
00:37:48,000 --> 00:37:51,440
Speaker 1: a bottom line world, and especially a bottom line country

652
00:37:51,440 --> 00:37:55,279
Speaker 1: that we live in, and I imagine what it would

653
00:37:55,360 --> 00:37:59,319
Speaker 1: likely mean is by by jobs more than it means, well, hey,

654
00:37:59,360 --> 00:38:01,840
Speaker 1: you've got more time, and why don't you innovate it

655
00:38:01,920 --> 00:38:06,200
Speaker 1: your job, because for most jobs, it'll probably be like, oh,

656
00:38:06,200 --> 00:38:07,840
Speaker 1: wait a minute, if we can teach it to do

657
00:38:07,880 --> 00:38:10,640
Speaker 1: forty center your job, I bet we could train it

658
00:38:10,640 --> 00:38:12,400
Speaker 1: to do a one hundred percent.

659
00:38:12,239 --> 00:38:15,080
Speaker 2: Yeah, or we can get rid of, you know, a

660
00:38:15,120 --> 00:38:16,759
Speaker 2: bunch of you and just keep some of you to

661
00:38:17,080 --> 00:38:18,440
Speaker 2: do the other sixty percent.

662
00:38:19,000 --> 00:38:20,880
Speaker 1: You know. But now see these people are out of

663
00:38:20,960 --> 00:38:22,839
Speaker 1: jobs that it's going to bite them in the rear though,

664
00:38:22,880 --> 00:38:26,640
Speaker 1: because it's not ultimately going to be well, who knows.

665
00:38:26,719 --> 00:38:28,120
Speaker 1: It doesn't seem like it could be good for the

666
00:38:28,160 --> 00:38:30,000
Speaker 1: overall economy if all of a sudden, all these people

667
00:38:30,000 --> 00:38:32,800
Speaker 1: are out of jobs. Because people being out of jobs

668
00:38:32,880 --> 00:38:35,640
Speaker 1: mean they're not that means the economy is going to tank.

669
00:38:35,680 --> 00:38:38,680
Speaker 1: They're not spending. And it's not like a situation where

670
00:38:40,440 --> 00:38:44,960
Speaker 1: you know, the tractor replaced the plow and then the

671
00:38:45,080 --> 00:38:48,480
Speaker 1: robot tractor replaced the tractor. But hey, now we've got

672
00:38:48,480 --> 00:38:51,400
Speaker 1: these better jobs where you're designing and building these robot

673
00:38:51,440 --> 00:38:55,160
Speaker 1: tractors and they're higher paying and they're great. It's not

674
00:38:55,400 --> 00:38:59,520
Speaker 1: like that because you know, the farmer was replaced who

675
00:38:59,600 --> 00:39:02,840
Speaker 1: drove that tractor and isn't skilled in the practice of

676
00:39:02,920 --> 00:39:07,640
Speaker 1: designing robot tractors. And in this case, in most cases,

677
00:39:08,480 --> 00:39:12,400
Speaker 1: they're not being there's not some other job waiting for

678
00:39:12,480 --> 00:39:15,560
Speaker 1: someone who got fired in the world of designing AI.

679
00:39:16,680 --> 00:39:17,399
Speaker 1: Does that make sense?

680
00:39:17,440 --> 00:39:19,799
Speaker 2: No, it makes total sense. But yeah, and in this case,

681
00:39:19,840 --> 00:39:22,640
Speaker 2: one of the big differences is instead of the farmer

682
00:39:22,719 --> 00:39:24,840
Speaker 2: having to go figure out how to work a computer,

683
00:39:25,120 --> 00:39:27,319
Speaker 2: that people working computers now have to go figure out

684
00:39:27,320 --> 00:39:30,719
Speaker 2: how to be farmers in order to sustain themselves. Right,

685
00:39:31,160 --> 00:39:33,960
Speaker 2: But you're right, we don't have a track record of

686
00:39:34,040 --> 00:39:38,120
Speaker 2: taking care of people very well, at least who are

687
00:39:38,200 --> 00:39:40,600
Speaker 2: out of a job. And I mean, without getting on

688
00:39:40,640 --> 00:39:44,319
Speaker 2: a soapbox here, what's either going to come out of this,

689
00:39:44,360 --> 00:39:45,840
Speaker 2: because there's going to be one or the other. The

690
00:39:45,840 --> 00:39:48,400
Speaker 2: status quo as it is now or as it wasn't

691
00:39:48,600 --> 00:39:51,080
Speaker 2: as of up to twenty twenty two. We don't know

692
00:39:51,160 --> 00:39:55,120
Speaker 2: that that's going to be around anymore. Instead, we'll either

693
00:39:55,200 --> 00:39:59,120
Speaker 2: do something like create universal basic income for people to

694
00:39:59,239 --> 00:40:03,720
Speaker 2: be like, hey, your industry literally does not exist anymore

695
00:40:03,960 --> 00:40:07,400
Speaker 2: and it just happened overnight. Basically, we're just gonna make

696
00:40:07,400 --> 00:40:09,759
Speaker 2: sure that everybody's at least minimally taken care of while

697
00:40:09,800 --> 00:40:13,120
Speaker 2: we're figuring out what comes next, or it's gonna be

698
00:40:13,160 --> 00:40:17,200
Speaker 2: like good luck, chump, you're fired, You're out on your own. Instead,

699
00:40:17,239 --> 00:40:19,960
Speaker 2: we're gonna take all this extra wealth, this extra two

700
00:40:20,080 --> 00:40:23,400
Speaker 2: trillion dollars that's gonna be generated, and push it upward

701
00:40:23,840 --> 00:40:27,279
Speaker 2: toward the wealthy instead and everybody else. Is just the

702
00:40:28,280 --> 00:40:32,000
Speaker 2: divide between wealthy and not wealthy is just going to

703
00:40:32,480 --> 00:40:35,120
Speaker 2: exponentially grow. One of those two things is gonna happen,

704
00:40:35,200 --> 00:40:37,560
Speaker 2: because I don't see how there's just gonna be a

705
00:40:37,600 --> 00:40:40,360
Speaker 2: regular middle ground like there is now where it's kind

706
00:40:40,360 --> 00:40:44,600
Speaker 2: of shaky, and how we're taking care of people, because

707
00:40:44,600 --> 00:40:48,600
Speaker 2: there's just gonna be so many layoffs and fairly skilled

708
00:40:48,640 --> 00:40:52,280
Speaker 2: workers being laid off too. We've just never encountered that before.

709
00:40:52,800 --> 00:40:56,320
Speaker 1: Yeah. I mean, that's the thing that these the largest

710
00:40:56,320 --> 00:40:59,839
Speaker 1: corporations might want to think about. Is all that's gonna

711
00:40:59,880 --> 00:41:05,880
Speaker 1: take is one CEO of a huge corporation to say,

712
00:41:06,000 --> 00:41:09,160
Speaker 1: wait a minute, it I think I can get rid

713
00:41:09,200 --> 00:41:12,319
Speaker 1: of seventy five percent of the vps in my in

714
00:41:12,360 --> 00:41:18,120
Speaker 1: my company, right, and like who unless who accept the

715
00:41:18,160 --> 00:41:20,399
Speaker 1: person at the very very top of that food chain

716
00:41:21,239 --> 00:41:23,880
Speaker 1: is protected. And the answer is nobody.

717
00:41:24,680 --> 00:41:26,919
Speaker 2: No, No, the offense essentially.

718
00:41:26,719 --> 00:41:27,920
Speaker 1: At the end of the day, because they make a

719
00:41:28,000 --> 00:41:29,879
Speaker 1: lot of money. If you it's one thing to lay

720
00:41:29,880 --> 00:41:32,279
Speaker 1: off a bunch of you know, technical writers that are

721
00:41:32,640 --> 00:41:35,000
Speaker 1: all sitting in their cubicles, But if you start laying

722
00:41:35,000 --> 00:41:38,680
Speaker 1: off those those vps who get those big bonuses, that's

723
00:41:38,880 --> 00:41:41,239
Speaker 1: more bonus money. And you know, are we looking at

724
00:41:41,280 --> 00:41:44,240
Speaker 1: a situation where a corporation is run by one human?

725
00:41:44,760 --> 00:41:47,480
Speaker 2: I mean, it's entirely possible, Like you can make a

726
00:41:47,520 --> 00:41:49,239
Speaker 2: really good case that what it is going to wipe

727
00:41:49,280 --> 00:41:53,120
Speaker 2: out is the middle management, right, vps, This is exactly

728
00:41:53,200 --> 00:41:55,400
Speaker 2: like you said, and that we still will need some

729
00:41:55,520 --> 00:41:57,160
Speaker 2: humans to do some stuff.

730
00:41:57,160 --> 00:41:59,320
Speaker 1: Like the board take care of the board, right, sure.

731
00:41:59,280 --> 00:42:02,279
Speaker 2: Of course, yeah, but yes, I mean who knows, we

732
00:42:02,880 --> 00:42:07,439
Speaker 2: have no idea at this point. Ultimately, it could very

733
00:42:07,520 --> 00:42:13,799
Speaker 2: easily provide for a much better healthier society, at least

734
00:42:13,840 --> 00:42:17,759
Speaker 2: financially speaking, it could do that, especially given a long

735
00:42:17,840 --> 00:42:18,760
Speaker 2: enough period of time.

736
00:42:19,160 --> 00:42:21,040
Speaker 1: I'm a cynic when it comes to that kind of

737
00:42:21,040 --> 00:42:21,600
Speaker 1: trust though.

738
00:42:21,680 --> 00:42:23,560
Speaker 2: I am as well for sure. But if you look

739
00:42:23,600 --> 00:42:28,080
Speaker 2: back in history at the history of technology overall, especially

740
00:42:28,160 --> 00:42:30,280
Speaker 2: if you just turn a blind eye to human suffering

741
00:42:30,280 --> 00:42:32,719
Speaker 2: for a second and you just look at the progress

742
00:42:32,719 --> 00:42:36,359
Speaker 2: of society. Right in a lot of ways it has

743
00:42:36,400 --> 00:42:39,840
Speaker 2: has gotten better and better things to technology. There's also

744
00:42:39,880 --> 00:42:42,640
Speaker 2: a lot of downsides to it. Nothing's black and white,

745
00:42:43,120 --> 00:42:46,200
Speaker 2: it's just not that's just not how things are. So

746
00:42:46,239 --> 00:42:48,680
Speaker 2: there's of course going to be problems. There's going to

747
00:42:48,719 --> 00:42:50,759
Speaker 2: be suffering, there's going to be people left behind, there's

748
00:42:50,800 --> 00:42:52,800
Speaker 2: going to be people that fall through the cracks. It's

749
00:42:52,880 --> 00:42:56,360
Speaker 2: just inevitable. We just don't know how many people for

750
00:42:56,480 --> 00:42:59,400
Speaker 2: how long and what will happen to those people on

751
00:42:59,440 --> 00:43:01,280
Speaker 2: the other of this transition.

752
00:43:02,320 --> 00:43:06,120
Speaker 1: Yeah, I was talking with somebody the other day about

753
00:43:06,880 --> 00:43:11,800
Speaker 1: the writer's strike in Hollywood. The WGA is striking right now.

754
00:43:12,520 --> 00:43:14,160
Speaker 1: For those of you who don't know, it's kind of

755
00:43:14,160 --> 00:43:16,680
Speaker 1: all over the place. But one of the things that

756
00:43:16,719 --> 00:43:21,359
Speaker 1: they have argued for in this round of negotiations is, hey,

757
00:43:21,440 --> 00:43:25,520
Speaker 1: you can't replace us with AI, and the studios all

758
00:43:25,560 --> 00:43:28,880
Speaker 1: came back and said, well, how about this, We'll assess

759
00:43:28,960 --> 00:43:33,240
Speaker 1: that on a year to year basis. And that's frightening

760
00:43:33,320 --> 00:43:36,520
Speaker 1: if you're if you're either a writer in Hollywood or

761
00:43:36,680 --> 00:43:41,239
Speaker 1: you're somebody who loves TV and films and quality TV

762
00:43:41,360 --> 00:43:46,120
Speaker 1: in films, because I don't know if I think ideation

763
00:43:46,800 --> 00:43:52,319
Speaker 1: and initial scripts maybe even right now, could I could

764
00:43:52,320 --> 00:43:54,359
Speaker 1: see that happening where they said they're like, all right now,

765
00:43:54,400 --> 00:43:58,120
Speaker 1: we'll bring in a human to refine this thing at

766
00:43:58,120 --> 00:44:01,000
Speaker 1: a much lower wage. That it's probably what they're most

767
00:44:01,000 --> 00:44:04,680
Speaker 1: afraid of, rather than being wholesale replaced, because like you said,

768
00:44:04,920 --> 00:44:09,160
Speaker 1: these programs are they're all about just data and numbers.

769
00:44:09,200 --> 00:44:12,120
Speaker 1: They're not They don't have human feelings, and that's what

770
00:44:12,560 --> 00:44:15,319
Speaker 1: art is. And so I think I would be more

771
00:44:15,320 --> 00:44:19,880
Speaker 1: concerned if I was writing pamphlets for Verizon or something,

772
00:44:20,719 --> 00:44:21,759
Speaker 1: or if I was.

773
00:44:22,040 --> 00:44:23,800
Speaker 2: Some pamphlet writer for Verizon.

774
00:44:23,880 --> 00:44:28,000
Speaker 1: Just went in gulp, No, I'm so sorry. But like

775
00:44:28,239 --> 00:44:30,720
Speaker 1: BuzzFeed back in the day, instead of having a dozen

776
00:44:30,760 --> 00:44:34,320
Speaker 1: writers writing clickbait articles, why not have just one human

777
00:44:34,680 --> 00:44:38,560
Speaker 1: that is a prompt engineer that's managing a virtual AI

778
00:44:39,600 --> 00:44:43,799
Speaker 1: clickbait room that's just pumping out these articles that you

779
00:44:43,840 --> 00:44:46,160
Speaker 1: know they were paying someone down to forty grand a

780
00:44:46,200 --> 00:44:47,160
Speaker 1: year to write previously.

781
00:44:47,760 --> 00:44:50,160
Speaker 2: Yeah, I mean it's a great question, like that was

782
00:44:50,200 --> 00:44:54,160
Speaker 2: a horrific, horrible job to have. Not too many years ago.

783
00:44:54,320 --> 00:44:56,319
Speaker 2: So it's great to have a computer do it, but

784
00:44:56,719 --> 00:44:58,839
Speaker 2: that means that we need these other people to go

785
00:44:58,920 --> 00:45:02,040
Speaker 2: on to be to have writing jobs that are more

786
00:45:02,080 --> 00:45:05,560
Speaker 2: satisfying to them than that. But that's not necessarily the case,

787
00:45:05,640 --> 00:45:10,360
Speaker 2: because as these things get smarter and better, they're just

788
00:45:10,400 --> 00:45:12,120
Speaker 2: going to be relied upon more. We're not going to

789
00:45:12,160 --> 00:45:15,160
Speaker 2: go back. There's no going back now. It just happened

790
00:45:15,200 --> 00:45:18,719
Speaker 2: like it just happened basically as of March twenty twenty three.

791
00:45:19,480 --> 00:45:23,600
Speaker 2: And one of the big problems that people have already

792
00:45:23,640 --> 00:45:31,280
Speaker 2: projected running into is if computers replace humans, say writers,

793
00:45:33,200 --> 00:45:36,480
Speaker 2: basically entirely. Eventually all the stuff that humans have written

794
00:45:36,480 --> 00:45:38,799
Speaker 2: on the Internet is going to become dated. It's going

795
00:45:38,840 --> 00:45:41,560
Speaker 2: to stop, yeah, and it will have been replaced and

796
00:45:41,640 --> 00:45:48,640
Speaker 2: picked up on by generative, pre trained transformers. Right, And

797
00:45:48,680 --> 00:45:52,040
Speaker 2: eventually all the writing on the Internet after a certain

798
00:45:52,120 --> 00:45:54,480
Speaker 2: date will have been written by computers, but will be

799
00:45:54,480 --> 00:45:57,600
Speaker 2: being scraped by computers. When humans go ask the computer

800
00:45:57,680 --> 00:46:01,160
Speaker 2: a question, the computer then goes in reference is something

801
00:46:01,200 --> 00:46:04,040
Speaker 2: written by a computer, So humans will be completely taken

802
00:46:04,080 --> 00:46:06,239
Speaker 2: out of the equation in that in that respect, we'll

803
00:46:06,239 --> 00:46:09,360
Speaker 2: be getting all of our information, at least non historical

804
00:46:09,400 --> 00:46:13,279
Speaker 2: information from non humans, and that could be a really

805
00:46:13,320 --> 00:46:15,839
Speaker 2: big problem, not just in the fact that we're losing

806
00:46:16,000 --> 00:46:18,959
Speaker 2: jobs or in the fact that computers are now telling

807
00:46:19,040 --> 00:46:21,600
Speaker 2: us all of our information, but also that there's some

808
00:46:23,400 --> 00:46:27,520
Speaker 2: there's some part of what humans put into things that

809
00:46:27,600 --> 00:46:31,120
Speaker 2: will be lost that I think we're going to demand.

810
00:46:31,239 --> 00:46:34,319
Speaker 2: I saw somebody put it like, I think, I can't

811
00:46:34,320 --> 00:46:37,000
Speaker 2: remember who it was, but they said, we're going like

812
00:46:37,280 --> 00:46:39,960
Speaker 2: people will go seek out human written stuff. There will

813
00:46:40,000 --> 00:46:43,040
Speaker 2: always be audiences for human written stuff. Yeah. Maybe, like

814
00:46:43,080 --> 00:46:47,120
Speaker 2: you said, will rely on computers to write the Verizon pamphlets,

815
00:46:47,480 --> 00:46:49,439
Speaker 2: but we're not going to rely on computers to write

816
00:46:49,440 --> 00:46:52,200
Speaker 2: great works of literature or to create great works of art,

817
00:46:52,320 --> 00:46:55,120
Speaker 2: Like we're just not going to They'll still do that.

818
00:46:55,160 --> 00:46:56,840
Speaker 2: They're going to be writing books and movies and all that,

819
00:46:56,880 --> 00:46:59,640
Speaker 2: but there will always be a taste in a market

820
00:46:59,680 --> 00:47:03,399
Speaker 2: for human created stuff. This guy said, I think he's right.

821
00:47:04,400 --> 00:47:08,440
Speaker 1: Yeah. And Justine Bateman, I don't know if you saw that.

822
00:47:09,560 --> 00:47:10,800
Speaker 1: I don't know if it was a blog poster.

823
00:47:11,880 --> 00:47:14,080
Speaker 2: Are you having a hallucination right now? Did you mean

824
00:47:14,200 --> 00:47:14,880
Speaker 2: Justine Bateman?

825
00:47:15,400 --> 00:47:20,560
Speaker 1: Yeah, yeah, Justin Bateman, the Jason Bateman's sister the actor,

826
00:47:20,680 --> 00:47:22,960
Speaker 1: and she's done all kinds of things since then. I

827
00:47:23,000 --> 00:47:26,440
Speaker 1: know she has a computer science degree, so she's very

828
00:47:26,440 --> 00:47:28,880
Speaker 1: smart and knows a lot about this stuff. But she

829
00:47:29,000 --> 00:47:32,080
Speaker 1: basically said, and this is beyond just the chatbot stuff.

830
00:47:32,520 --> 00:47:36,400
Speaker 1: But she was like, right now, there are major Hollywood

831
00:47:36,400 --> 00:47:41,719
Speaker 1: stars being scanned, and there may be a brand new

832
00:47:41,760 --> 00:47:45,600
Speaker 1: Tom Cruise movie in sixty years. Yeah, long after he's

833
00:47:45,680 --> 00:47:49,239
Speaker 1: dead starring Tom Cruise. He may be making movies for

834
00:47:49,239 --> 00:47:52,480
Speaker 1: the next two hundred years. And like, is this what

835
00:47:52,520 --> 00:47:55,279
Speaker 1: you want? Actors? Do you want to be scanned and

836
00:47:55,400 --> 00:47:58,880
Speaker 1: have them use your image like this in perpetuity for

837
00:47:59,160 --> 00:48:02,120
Speaker 1: you know, there will be money involved. It's not like

838
00:48:02,160 --> 00:48:03,640
Speaker 1: they can just say, Okay, we can just do whatever

839
00:48:03,640 --> 00:48:06,360
Speaker 1: we want. But what if they're like, here's a billion dollars,

840
00:48:06,480 --> 00:48:11,520
Speaker 1: Tom Cruise, just for the use of your image in perpetuity,

841
00:48:11,760 --> 00:48:15,239
Speaker 1: because we will be able to duplicate that so realistically

842
00:48:15,800 --> 00:48:20,320
Speaker 1: that people won't know human voices, same thing that's already happening.

843
00:48:20,880 --> 00:48:26,400
Speaker 1: What yeah, it is. So that stuff is kind of scary.

844
00:48:26,560 --> 00:48:29,400
Speaker 1: And you know when you read I didn't really know

845
00:48:29,440 --> 00:48:32,960
Speaker 1: this was kind of already happening in companies. But Olivia

846
00:48:33,040 --> 00:48:37,080
Speaker 1: found this stuff IBM CEO Arvin Krishna said just last

847
00:48:37,080 --> 00:48:39,879
Speaker 1: month in May that he believed thirty percent of back

848
00:48:39,880 --> 00:48:44,719
Speaker 1: office jobs could be replaced over five years, and it

849
00:48:44,800 --> 00:48:48,760
Speaker 1: was pausing hiring for close to eight thousand positions because

850
00:48:48,800 --> 00:48:52,160
Speaker 1: they might be able to use AI instead. And then

851
00:48:52,280 --> 00:48:56,759
Speaker 1: Dropbox talked about the AI era when they announced a

852
00:48:56,840 --> 00:49:00,239
Speaker 1: round of layoffs. So it is happening right now in

853
00:49:00,320 --> 00:49:00,719
Speaker 1: real time.

854
00:49:00,800 --> 00:49:04,680
Speaker 2: Pretty amazing. Yeah, that's I mean, there's proof positive right there,

855
00:49:05,040 --> 00:49:08,479
Speaker 2: Like that guy couldn't even wait a couple months a year,

856
00:49:09,120 --> 00:49:12,280
Speaker 2: Like this really started up in March, and he's saying

857
00:49:12,320 --> 00:49:14,719
Speaker 2: this in May. Already in May, they're like, wait, wait,

858
00:49:14,760 --> 00:49:17,759
Speaker 2: stop hiring. We're gonna eventually replace these guys with AI

859
00:49:18,160 --> 00:49:21,480
Speaker 2: so soon that we're going to stop hiring those positions

860
00:49:21,480 --> 00:49:24,040
Speaker 2: for now until the AI is competent enough to take over.

861
00:49:24,600 --> 00:49:27,319
Speaker 1: I mean, how many people does ib employ? What, what's

862
00:49:27,360 --> 00:49:28,160
Speaker 1: thirty percent of that?

863
00:49:28,400 --> 00:49:30,960
Speaker 2: I don't know. I would say at least at least

864
00:49:31,000 --> 00:49:34,879
Speaker 2: one hundred people, right, So, yeah, like you said, it's

865
00:49:34,880 --> 00:49:37,040
Speaker 2: happening already. And then one other thing to look out

866
00:49:37,040 --> 00:49:40,600
Speaker 2: for too, that's I believe is already at least theoretically

867
00:49:40,640 --> 00:49:45,879
Speaker 2: possible since AI can write code. Now they'll be able

868
00:49:45,920 --> 00:49:50,640
Speaker 2: to create new large language models themselves, So the computers

869
00:49:50,640 --> 00:49:53,400
Speaker 2: will be able to create new AI.

870
00:49:54,080 --> 00:49:56,160
Speaker 1: Well, that's the singularity, right.

871
00:49:56,320 --> 00:50:00,799
Speaker 2: No, the singularity is when one of them a understands

872
00:50:00,880 --> 00:50:05,480
Speaker 2: what it is and become Yes, that's the singularity.

873
00:50:05,840 --> 00:50:07,560
Speaker 1: But this leads to that though, doesn't it It does.

874
00:50:07,840 --> 00:50:11,600
Speaker 2: It's hypothetically yes, but we just understand what's going on

875
00:50:11,680 --> 00:50:13,880
Speaker 2: so little that you just can't say either way. Really,

876
00:50:13,920 --> 00:50:16,239
Speaker 2: you definitely can't say that it no, it won't happen.

877
00:50:16,280 --> 00:50:18,359
Speaker 2: It's just fantasy. And you also can't say yes, it's

878
00:50:18,360 --> 00:50:19,280
Speaker 2: definitely gonna happen.

879
00:50:19,960 --> 00:50:25,520
Speaker 1: Yeah. And here's the thing, man, I'm not a paranoid technophobe.

880
00:50:26,080 --> 00:50:27,960
Speaker 2: You don't any measure foiled cap on.

881
00:50:28,719 --> 00:50:33,120
Speaker 1: No, by any measure. I'm a pretty positive thinker, and

882
00:50:33,200 --> 00:50:35,239
Speaker 1: this this is pretty scary to me.

883
00:50:37,560 --> 00:50:44,160
Speaker 2: I'm just gonna leave that there, agreed, Chuck. Okay, if

884
00:50:44,200 --> 00:50:47,520
Speaker 2: you want to know more about large language models, everybody

885
00:50:47,640 --> 00:50:51,279
Speaker 2: just to start looking around Earth and when you see

886
00:50:51,320 --> 00:50:54,000
Speaker 2: people running from explosions, go toward it and ask what's

887
00:50:54,040 --> 00:50:54,440
Speaker 2: going on?

888
00:50:55,040 --> 00:50:56,800
Speaker 1: You almost said, type it into a search engine.

889
00:50:57,080 --> 00:50:59,520
Speaker 2: Yeah, steer clear of those. Yeah, there's so much more,

890
00:50:59,520 --> 00:51:02,120
Speaker 2: we could have talked about. But this is if you

891
00:51:02,160 --> 00:51:04,520
Speaker 2: ask me, this is round one. I think we definitely

892
00:51:04,560 --> 00:51:06,000
Speaker 2: need to do at least one or so.

893
00:51:06,080 --> 00:51:08,680
Speaker 1: More on this, Okay, Yeah, and then one day, like

894
00:51:08,719 --> 00:51:11,600
Speaker 1: I said, Ai, Josh and Chuckle, just wrap it all

895
00:51:11,680 --> 00:51:13,480
Speaker 1: up and spank it on the bottom and say no

896
00:51:13,640 --> 00:51:14,279
Speaker 1: problems here.

897
00:51:14,400 --> 00:51:17,120
Speaker 2: Hopefully they'll give us a billion dollars rather than like

898
00:51:17,160 --> 00:51:20,040
Speaker 2: a month free of blue Apron instead.

899
00:51:20,680 --> 00:51:22,359
Speaker 1: Yeah, I mean we could talk here.

900
00:51:23,840 --> 00:51:27,000
Speaker 2: Well, since Chuck said we can talk real confidential like

901
00:51:27,400 --> 00:51:28,760
Speaker 2: that means it's time for listener.

902
00:51:28,800 --> 00:51:34,880
Speaker 1: May I'm gonna call this conception, not inception, but conception.

903
00:51:35,040 --> 00:51:36,680
Speaker 2: Oh I saw this one. I don't know how I

904
00:51:36,719 --> 00:51:37,319
Speaker 2: feel about this.

905
00:51:39,000 --> 00:51:41,200
Speaker 1: Hey, guys. Last year, my wife and I were attempting

906
00:51:41,239 --> 00:51:43,040
Speaker 1: to get pregnant. A couple of months in we made

907
00:51:43,040 --> 00:51:45,040
Speaker 1: plans to stay with some friends in another town for

908
00:51:45,080 --> 00:51:47,600
Speaker 1: a weekend, and when we can arrive, it happened to

909
00:51:47,640 --> 00:51:52,279
Speaker 1: coincide with my wife's ovulation cycle. As shy people, we

910
00:51:52,320 --> 00:51:55,800
Speaker 1: both felt a little bit awkward about, you know, hugging

911
00:51:55,880 --> 00:51:58,960
Speaker 1: and kissing in a friend's guest room, but we really

912
00:51:59,040 --> 00:52:01,360
Speaker 1: didn't want to miss that chance and that time of

913
00:52:01,360 --> 00:52:04,319
Speaker 1: the month, so we went about getting in the mood

914
00:52:04,320 --> 00:52:07,080
Speaker 1: as quietly as possible, and my wife suggested we play

915
00:52:07,080 --> 00:52:09,600
Speaker 1: a podcast from my phone so that, you know, if

916
00:52:09,640 --> 00:52:12,440
Speaker 1: any noise is made outside the room, it would sound

917
00:52:12,480 --> 00:52:14,560
Speaker 1: like we were just doing a little pre bedtime listening.

918
00:52:15,920 --> 00:52:18,360
Speaker 1: I knew I needed something with nice, simple production values,

919
00:52:18,440 --> 00:52:21,120
Speaker 1: so we wouldn't get distracted, of course, by the whiz

920
00:52:21,200 --> 00:52:23,640
Speaker 1: bang sounds and whatnot. And since you were my intro

921
00:52:23,760 --> 00:52:26,040
Speaker 1: to the world of podcast, I've always had a steady

922
00:52:26,040 --> 00:52:29,840
Speaker 1: supply of yours downloaded. I picked the least interesting sounding

923
00:52:29,880 --> 00:52:33,640
Speaker 1: one in the feed at the time, How Coal Works.

924
00:52:33,960 --> 00:52:38,600
Speaker 2: Okay, I thought that one turned out to be surprisingly interesting. Yeah,

925
00:52:38,640 --> 00:52:40,239
Speaker 2: I could see how we would have thought that though.

926
00:52:40,800 --> 00:52:43,359
Speaker 1: Yeah, for sure. We put that on and we did

927
00:52:43,360 --> 00:52:47,000
Speaker 1: our business. Six weeks later we got a positive pregnancy test,

928
00:52:47,880 --> 00:52:50,439
Speaker 1: and now over a year later, we've welcomed our son

929
00:52:50,520 --> 00:52:53,640
Speaker 1: into the world. Name is Cole, and of course we

930
00:52:53,719 --> 00:52:57,279
Speaker 1: named him Cole. That is what this person said.

931
00:52:57,400 --> 00:52:59,200
Speaker 2: Wait a minute, Wait a minute, they really did name

932
00:52:59,239 --> 00:52:59,680
Speaker 2: them Cole.

933
00:53:00,600 --> 00:53:04,319
Speaker 1: No, he said it as a joke. Oh but great

934
00:53:04,360 --> 00:53:07,400
Speaker 1: minds right, good joke for both of them. It's almost

935
00:53:07,440 --> 00:53:13,120
Speaker 1: like you're both chatbox. And this person said You're fine

936
00:53:13,160 --> 00:53:15,600
Speaker 1: to read this, but give me a fake name. And

937
00:53:15,640 --> 00:53:18,360
Speaker 1: so I just want to say thanks to Gene for

938
00:53:18,440 --> 00:53:19,520
Speaker 1: writing in about this.

939
00:53:20,040 --> 00:53:21,840
Speaker 2: Gene is in Gene Transfer.

940
00:53:23,000 --> 00:53:25,400
Speaker 1: Sure, thanks a lot, Gene.

941
00:53:25,400 --> 00:53:29,759
Speaker 2: We appreciate that. I think again, I'm still figuring that

942
00:53:29,800 --> 00:53:32,719
Speaker 2: one out. And if you want to be like Gene,

943
00:53:32,840 --> 00:53:36,360
Speaker 2: I'm making air quotes here, you can send us an

944
00:53:36,400 --> 00:53:39,160
Speaker 2: email to wrap it up. Spank it on the bottom.

945
00:53:39,320 --> 00:53:40,400
Speaker 2: Only humans can do that.

946
00:53:41,120 --> 00:53:43,160
Speaker 1: I wonder if when you said spanking on the bottom,

947
00:53:43,239 --> 00:53:44,840
Speaker 1: if that created any issues.

948
00:53:45,160 --> 00:53:49,760
Speaker 2: Yeah, I hadn't thought about that. Maybe playfully how about that? Sure,

949
00:53:51,080 --> 00:53:58,680
Speaker 2: and send it off to stuff podcast at iHeartRadio dot com.

950
00:53:58,920 --> 00:54:01,799
Speaker 1: Stuff you Should Know is a production of iHeartRadio. For

951
00:54:01,880 --> 00:54:06,080
Speaker 1: more podcasts my heart Radio, visit the iHeartRadio app, Apple Podcasts,

952
00:54:06,200 --> 00:54:08,040
Speaker 1: or wherever you listen to your favorite shows.