1 00:00:01,000 --> 00:00:04,759 How'd you like to listen to dot net rocks with no ads? Easy? 2 00:00:05,320 --> 00:00:09,400 Become a patron For just five dollars a month, you get access to a 3 00:00:09,480 --> 00:00:14,199 private RSS feed where all the shows have no ADS. Twenty dollars a month 4 00:00:14,240 --> 00:00:18,600 will get you that and a special dot net Rocks patron mug. Sign up 5 00:00:18,640 --> 00:00:23,760 now at Patreon dot dot net rocks dot com. Hey Carlin Richard Here, 6 00:00:24,079 --> 00:00:29,280 As you may have heard, NDC is back offering their incredible in person conferences 7 00:00:29,320 --> 00:00:33,240 around the world, and we'd like to tell you about them. NDC Oslow 8 00:00:33,240 --> 00:00:37,039 will be made twenty first through the twenty fifth. Go to NDC Oslo dot 9 00:00:37,119 --> 00:00:42,759 com to register. NDC Copenhagen is happening August twenty seventh through the thirty first. 10 00:00:43,320 --> 00:00:48,960 Go to NDC Copenhagen dot com for more information. NDC Porto is happening 11 00:00:49,000 --> 00:00:53,280 October sixteenth through the twentieth. The early bird discount for DC Porto ends July 12 00:00:53,560 --> 00:00:58,280 twenty first. Go to Eddcporto dot com to register and check out the full 13 00:00:58,320 --> 00:01:15,239 lineup of conferences at DC conferences dot com. Hey, welcome back to dot 14 00:01:15,280 --> 00:01:18,760 net Rocks. This Carl Franklin and this is Richard Campbell. I'm very excited, 15 00:01:18,879 --> 00:01:23,079 Richard, because my one of my heroes of podcasting, Grant Barrett, 16 00:01:23,159 --> 00:01:26,239 is going to be with us in just a few minutes. Um. But 17 00:01:26,480 --> 00:01:30,680 first let's do better, no framework and get out of the way. Awesome 18 00:01:38,400 --> 00:01:41,040 man would have got Believe it or not, I started a new podcast. 19 00:01:41,359 --> 00:01:45,959 Are you kidding? I'm stupid? I don't know, because both put several 20 00:01:46,000 --> 00:01:49,680 hours of podcasts a weekend, so it's not surprising. Yeah. So this 21 00:01:49,719 --> 00:01:53,480 one is a personal more like a personal journal. Oh yeah, I'm just 22 00:01:53,640 --> 00:01:57,879 finding people that I've met in my travels and I've worked with that I find 23 00:01:57,920 --> 00:02:01,079 really interesting and I think would also their stories would be very interesting to my 24 00:02:01,200 --> 00:02:05,840 listeners. And so it's called no Excuse for boredom. I love it. 25 00:02:06,760 --> 00:02:09,240 It's a no excuse for boredom dot com. You can't believe you got that 26 00:02:09,360 --> 00:02:15,719 domain name. I have a way with getting domain names. I don't know. 27 00:02:15,879 --> 00:02:19,759 Yeah, yep, no excuse for boredom dot com. So there's only 28 00:02:19,759 --> 00:02:23,080 two episodes up there. Um. The first one is Dan Grief. He's 29 00:02:23,120 --> 00:02:27,000 a guy from England who I know, you don't know, but he I 30 00:02:27,080 --> 00:02:31,680 basically talked to him about the difference between new England and Old England in terms 31 00:02:31,719 --> 00:02:36,360 of language and slang, and great, you probably like this too. I 32 00:02:36,360 --> 00:02:40,360 asked him to define a whole bunch of British slang words and phrases that confused 33 00:02:40,360 --> 00:02:45,639 me when I first heard them, like oh uh, okay, are you 34 00:02:45,639 --> 00:02:50,120 trying to take the piss out of me? The classic? I never The 35 00:02:50,159 --> 00:02:53,199 only time I ever heard that was like on a Gordon Ramsey show, right, 36 00:02:53,599 --> 00:02:57,240 and I'm like, what the heck are they talking about. I kind 37 00:02:57,280 --> 00:03:00,000 of got from the context that it means like are you kidding with me? 38 00:03:00,120 --> 00:03:04,199 Or are you are you just trying to pull a fast one or whatever, 39 00:03:04,680 --> 00:03:07,800 but I'd never heard it before. So there's just a long list. Yeah, 40 00:03:07,800 --> 00:03:14,039 there's something very cathe or esque about that. It really put in a 41 00:03:14,159 --> 00:03:19,400 stint exactly. I just didn't get it. So there's it's just funny. 42 00:03:19,439 --> 00:03:23,039 And then the second one is with Randy Judkins. Oh man, I remember 43 00:03:23,120 --> 00:03:25,319 him. Yeah, my friend from in the early days, my friend from 44 00:03:25,319 --> 00:03:29,759 Maine who did that first dot net what is dot net video? And he's 45 00:03:29,800 --> 00:03:34,159 a jungling magician, comedian. He was so funny too. Oh had me 46 00:03:34,280 --> 00:03:36,400 rolling and stitches. All right, Well, enough of that. That's what 47 00:03:36,520 --> 00:03:38,800 I got who was talking to us today? Richard grabbed a comment off a 48 00:03:38,840 --> 00:03:43,159 show sixteen o five, the one we did back in twenty eighteen, so 49 00:03:43,199 --> 00:03:46,240 that's a while ago. And it was actually about IoT and edge competing. 50 00:03:46,240 --> 00:03:49,360 But we were taught and this is Jared Rhodes we were talking to. I 51 00:03:49,400 --> 00:03:52,439 don't even remember this because of course you remember every show. It was only 52 00:03:52,560 --> 00:03:58,400 five years ago. Wait, oh no, that's me. But the latter 53 00:03:58,439 --> 00:04:01,360 half of the show we were talking about data collection a lot because these you 54 00:04:01,400 --> 00:04:03,680 know, we were talking about how small the hardware gotten and you can put 55 00:04:03,680 --> 00:04:08,319 it just about anywhere and nothing. It's always uploading to the cloud. You 56 00:04:08,360 --> 00:04:12,560 know, there's consequences to all of that. And Mike comment on the show, 57 00:04:12,599 --> 00:04:15,000 and admittedly he commented, you know, within a month of it being 58 00:04:15,000 --> 00:04:16,879 published several years ago, he said, hey, this is a great ethical 59 00:04:16,879 --> 00:04:21,560 discussion about collecting data. I always revert back to Kransberg's first law tech it 60 00:04:21,720 --> 00:04:26,839 is neither good or bad, nor is it neutral. I think that's actually 61 00:04:26,920 --> 00:04:30,319 his quote, but okay, I don't disagree with the quote. Mentoring a 62 00:04:30,439 --> 00:04:34,240 store gathering data is troubling. You obviously opt in. And this is when 63 00:04:34,240 --> 00:04:38,519 we were talking about some of the projects I've been involved with where they're actually 64 00:04:38,560 --> 00:04:41,920 instrumenting shelves to see what you look at, what you pick up, what 65 00:04:42,040 --> 00:04:45,360 you put down, all of those kinds of things. And so the point 66 00:04:45,399 --> 00:04:47,199 of you obviously opt in by seeing the monitor continue to enter, But is 67 00:04:47,199 --> 00:04:51,800 that enough? Shouldn't the customer not to what extent their behavior will be analyzed? 68 00:04:51,800 --> 00:04:56,800 And for what purpose? What are the patrons who beyond their control only 69 00:04:56,839 --> 00:04:59,720 have access to that store, are you know, aliening their business and taking 70 00:04:59,720 --> 00:05:03,040 a v to the system to gather data? Programming ethics have always fascinated me. 71 00:05:03,120 --> 00:05:08,439 Recent protests by Google employees now again twenty eighteen. Yeah, and that 72 00:05:08,519 --> 00:05:13,439 was when Google was going into China and they were demanding the Great Firewall stuff 73 00:05:13,480 --> 00:05:17,879 and censoring and so forth. Recent protests by Google employees over developing censoring products 74 00:05:17,879 --> 00:05:21,920 for the Chinese government and the military machine learning for the US government highlight the 75 00:05:21,959 --> 00:05:27,800 backlashes these companies faced with adopting technologies for different applications. And I'm like man 76 00:05:28,000 --> 00:05:32,959 five years later and there's no protests than ever. But it always boils down 77 00:05:33,000 --> 00:05:36,399 to this, you will be able to find a developer to do what you 78 00:05:36,480 --> 00:05:41,160 want to have done. There will always be someone with the ethics that aligned 79 00:05:41,199 --> 00:05:44,480 to yours or what desperate enough to do that work. This is unique to 80 00:05:44,519 --> 00:05:47,040 programming. However, programming has a unique advantage to being cheap to scale, 81 00:05:47,360 --> 00:05:51,720 so one developer can have a profound impact on humanity versus one unethical police officer 82 00:05:51,959 --> 00:05:57,040 or one unethical investor. Had Bertie made Off, I love this reference. 83 00:05:57,079 --> 00:06:00,240 Had Bertie made Off known how to make an ASP website, he could have 84 00:06:00,319 --> 00:06:04,600 offered his services anonymously to many more people. It was a great discussion on 85 00:06:04,720 --> 00:06:09,839 what to two about this responsibility. How do companies talk about this? My 86 00:06:09,959 --> 00:06:15,000 experience is most don't, don't yeah, but we will eventually need to Yeah. 87 00:06:15,040 --> 00:06:18,079 Wow, thanks you, Thank Mike, you killed it. I was 88 00:06:18,079 --> 00:06:21,560 so glad to find this comment as I dug through knowing where we were going 89 00:06:21,639 --> 00:06:27,079 to go today, and so thanks thanks for this. It is absolutely the 90 00:06:27,079 --> 00:06:30,720 conversations we need to have and are having. And a copy of music Codebuy 91 00:06:30,839 --> 00:06:33,199 is on its way to un If you'd like a copy of Music Code by 92 00:06:33,480 --> 00:06:36,920 write a comment on the website at dondt Rocks dot com or on the facebooks 93 00:06:36,959 --> 00:06:40,600 we publish every show there, and if you comment there and everybody in the 94 00:06:40,600 --> 00:06:44,120 show. We'll send you a copy of music Code Buy and definitely follow us 95 00:06:44,160 --> 00:06:47,600 on mastadon. I'm at Carl Franklin at tech Hub dot social, and I'm 96 00:06:47,800 --> 00:06:53,040 Rich Campbell at mastadon dot social. Send us a two Rudy too too. 97 00:06:53,040 --> 00:06:56,759 It's still funny, Yeah, it's still funny. Okay, we're still in 98 00:06:56,879 --> 00:07:00,480 grade seven. I know I'm always going to be that way. This is 99 00:07:00,519 --> 00:07:05,199 a really exciting time to have Grant Barrett on the show because I think Grant 100 00:07:05,680 --> 00:07:15,720 exists in the intersection of language and ethics and tach. So let me read 101 00:07:15,759 --> 00:07:19,319 his bio for you. Away with Words, Wayward Radio dot org co host 102 00:07:19,360 --> 00:07:26,160 and co producer, Grant Barrett is an American lexicographer and linguist specializing in slang 103 00:07:26,240 --> 00:07:29,959 and new words. The public radio show has heard from coast to coast in 104 00:07:29,959 --> 00:07:34,360 the United States. It's a Wayward Radio dot org slash radio and around the 105 00:07:34,399 --> 00:07:40,759 world by podcasts Wayward Radio dot org slash Podcasts. He has helped produce dozens 106 00:07:40,759 --> 00:07:45,920 of dictionaries and has authored three books of his own. He's a voracious reader, 107 00:07:46,279 --> 00:07:48,519 reads and speaks a bit of French and Spanish, and spent years working 108 00:07:48,560 --> 00:07:54,199 in the jargon rich jungles of information technology. Though born and raised in Missouri 109 00:07:54,279 --> 00:07:57,639 and having been a long time resident of New York City, Grant now lives 110 00:07:57,639 --> 00:08:01,399 in San Diego, California, with his wife, also linguist and lexicographer, 111 00:08:01,439 --> 00:08:07,439 and their young son. His personal website is Grant Barrett with two tis dot 112 00:08:07,439 --> 00:08:11,560 com. Welcome Grant, Hello, how you doing Hey in the bio and 113 00:08:11,680 --> 00:08:15,600 the u URLs. Yeah. Absolutely. First of all, I just got 114 00:08:15,600 --> 00:08:18,519 to say your podcast as a staple in my house. You know, it's 115 00:08:18,519 --> 00:08:20,560 a radio show too, heard coast to coast in the United States as you 116 00:08:20,600 --> 00:08:26,319 just read, Yes, you're right, and also independently, voraciously independent, 117 00:08:26,519 --> 00:08:28,240 right yeah, yeah, not a part of NPR, but heard on a 118 00:08:28,279 --> 00:08:31,759 lot of NPR stations. It's hard to explain that to people, but it's 119 00:08:31,799 --> 00:08:37,480 a it's an one, tiny little nonprofit. We just the radio stations are 120 00:08:37,559 --> 00:08:39,879 clients of ours, so we just so cool. I talked to the radio 121 00:08:39,879 --> 00:08:43,879 stations because you have an MPR sound, well you could easily be MPR. 122 00:08:45,000 --> 00:08:48,080 Yeah, we are on NPR stations, so we take the NPR sound. 123 00:08:48,080 --> 00:08:50,720 But we were just added to a bunch of new stations Miami and Tucson. 124 00:08:50,919 --> 00:08:58,799 And actually we're now heard in Roswell, New Mexico. Wow, Aliensis are 125 00:08:58,840 --> 00:09:03,480 listening. Excellent, it's the caller there. Let me tell you where that 126 00:09:03,519 --> 00:09:09,200 comes from, you know, Beetle Juice. Yeah, that stuff a great 127 00:09:09,200 --> 00:09:11,720 movie, all right. So the reason that I got in touch with you 128 00:09:11,840 --> 00:09:16,919 is because I was listening to your show and Martha read a story or a 129 00:09:16,960 --> 00:09:24,639 comment or an anecdote about She went to chat GPT and told it to give 130 00:09:24,720 --> 00:09:31,759 her a one paragraph description of a way with words, and it came up 131 00:09:31,759 --> 00:09:35,759 with this what she thought was a completely unique paragraph. She said, she 132 00:09:35,840 --> 00:09:39,519 googled each one of them, each one of the sentences, and they could 133 00:09:39,559 --> 00:09:46,720 not be found elsewhere on the Internet, and then started this moral discussion about 134 00:09:46,799 --> 00:09:50,440 you know, there's people who are really afraid of this, all right, 135 00:09:50,519 --> 00:09:54,519 of chat GPT, and of language evolving in general. And I think one 136 00:09:54,559 --> 00:09:58,679 of you, maybe it was you, said that, well, back in 137 00:09:58,720 --> 00:10:05,240 the days of what was it Socrates or ancient Greece, they were concerned that 138 00:10:05,360 --> 00:10:09,679 if they taught, if they educated people, and books, you know, 139 00:10:09,720 --> 00:10:13,039 people were writing books. Obviously they weren't writing, you know, printing books, 140 00:10:13,039 --> 00:10:16,360 but if they were writing, that was going to make people stupid, 141 00:10:16,440 --> 00:10:18,799 or because they wouldn't have to remember so much. Yeah, at the time 142 00:10:20,159 --> 00:10:26,120 and the ancient Greeks, you were expected to memorize and recite allowed, and 143 00:10:26,519 --> 00:10:31,919 once writing became more standard, you might put it on a sheepskin, vellum 144 00:10:31,480 --> 00:10:37,000 or papyrus. People are like, wait a second, you're not memorizing this, 145 00:10:37,080 --> 00:10:41,279 You're just gonna go and read it. It's not in your head. 146 00:10:41,559 --> 00:10:45,879 How do you actually know anything? Right, If you just know where to 147 00:10:45,919 --> 00:10:48,399 get it, that's not knowledge. It's stored out of your brain. You 148 00:10:48,480 --> 00:10:52,600 got to think that was a technological advance, that better writing implements had come 149 00:10:52,639 --> 00:10:56,639 along. You weren't chiseling in stone or pressing in clay. And well, 150 00:10:56,639 --> 00:11:01,679 it's the evidence of every technological advance is as a of resisting force, there's 151 00:11:01,679 --> 00:11:05,759 always opposition. This was the same screaming match that happened when Gutenberg made the 152 00:11:05,799 --> 00:11:11,919 press. Well, everything, all technological advances, has pushed back. There's 153 00:11:11,960 --> 00:11:18,440 not a one. I mean, somebody probably argued against fire. It burns 154 00:11:16,120 --> 00:11:20,159 somebody, somebody, nothing good will come of it. Somebody argued against the 155 00:11:20,159 --> 00:11:24,720 wheel. But wait, it doesn't have a stopping point. It can just 156 00:11:24,799 --> 00:11:26,639 keep going. How well, I know that when I go to sleep, 157 00:11:26,679 --> 00:11:30,679 it'll step, You'll still be there in the morning. Yeah, right, 158 00:11:30,840 --> 00:11:33,480 roll away. Yeah, yeah, it's circular, that's satanic. I don't 159 00:11:33,519 --> 00:11:37,799 I don't know. We can't trust people with new technology that I think that's 160 00:11:37,799 --> 00:11:41,159 what it comes down to, right, and that's why people. But on 161 00:11:41,240 --> 00:11:43,519 a larger point, what's funny about this is, you know, since we've 162 00:11:43,519 --> 00:11:48,759 booked this interview for the three of us to talk, the larger conversation has 163 00:11:48,759 --> 00:11:56,080 continued as fast as all the GPT out growths and companies. And I mean, 164 00:11:56,279 --> 00:12:01,480 how many how many companies were founded in Delaware? Since so much of 165 00:12:01,559 --> 00:12:05,919 this conversation hasn't moved on since we just booked this interview, Margat and I 166 00:12:05,919 --> 00:12:11,080 talked about it on the air. I think the way to discuss this isn't 167 00:12:11,080 --> 00:12:16,840 necessarily from an ethical standpoint so much as it is from a personal standpoint. 168 00:12:16,039 --> 00:12:22,080 It's Martha took. Her approach was really interesting as a former magazine writer, 169 00:12:24,039 --> 00:12:28,720 so know she were for Glamor and for the Washington Post and some other things. 170 00:12:28,759 --> 00:12:31,279 She just said, well, if I was still in the trenches, 171 00:12:33,960 --> 00:12:37,399 what to stop my competitor from taking all my gigs and using this to write 172 00:12:37,480 --> 00:12:41,200 rough drafts and then tuning them up a little bit and turning them in and 173 00:12:41,600 --> 00:12:45,720 beating me to every deadline, taking all of the greatest gigs, and nobody 174 00:12:45,720 --> 00:12:48,200 being the wiser. Now we know that there are tools that can help detect 175 00:12:48,200 --> 00:12:54,080 the GPT derived you know articles. Yeah, but that's today. Yeah, 176 00:12:54,080 --> 00:12:56,279 she's got a good point. Yeah, except that quality is really not that 177 00:12:56,320 --> 00:13:00,159 good. Like, is it ultimately saving you time? I don't even I'm 178 00:13:00,200 --> 00:13:01,960 not even convinced. Well, that's the question you still have, Like I 179 00:13:01,960 --> 00:13:05,279 said, you have to tune it still, right? You still? I 180 00:13:05,279 --> 00:13:09,320 mean what if you what if your machine learning was based on your own writing 181 00:13:11,000 --> 00:13:13,200 and not the larger world. Yeah, it's totally going to be as smart 182 00:13:13,240 --> 00:13:18,480 as you and invariably not as right. What if what if it was a 183 00:13:18,639 --> 00:13:22,919 two set process. The first draft was just from the larger chat GTPT and 184 00:13:22,960 --> 00:13:26,960 the second draft said, now run it through the corpus of my own writing. 185 00:13:28,080 --> 00:13:30,960 So it sounds like me. I read an interesting opinion I think it 186 00:13:31,000 --> 00:13:35,840 was in Washington Post about from a teacher who said teachers need not worry about 187 00:13:35,960 --> 00:13:43,360 AI, and the pushback was, there's a simple solution, just make tests 188 00:13:43,080 --> 00:13:48,960 in how in person or even oral exams. Yeah. Handwriting, Yeah, 189 00:13:48,960 --> 00:13:54,279 handwriting or you know, Oh you mentioned that things have progressed. I've noticed 190 00:13:54,320 --> 00:14:00,840 that there's a word Press Chat GPT plug in. Now, oh, because 191 00:14:00,879 --> 00:14:05,399 that's going to work. Well, all the SEO animals love this crap, 192 00:14:05,519 --> 00:14:07,360 right, How many happy SEO blogs are we going to see with this? 193 00:14:07,720 --> 00:14:11,120 But I just gon alert them up with Google ads. Yeah, well they 194 00:14:11,120 --> 00:14:15,399 were doing that anyway. This is, I know, but now more and 195 00:14:15,559 --> 00:14:20,559 easier. Yeah. I would wonder if it'll actually be better. I don't 196 00:14:20,559 --> 00:14:22,519 know, because it might actually be worse. So I want to talk about 197 00:14:24,080 --> 00:14:26,360 So we're at a bridging point here where there's a before and and after, 198 00:14:26,639 --> 00:14:30,879 but I wanted to talk about and compare this to another bridging point. I 199 00:14:30,919 --> 00:14:33,759 went back to school as an older student in his twenties, and I went 200 00:14:33,759 --> 00:14:37,799 to Columbia University to get a degree. And I was right there at the 201 00:14:37,879 --> 00:14:43,000 point where universities weren't really decided whether or not laptops should be allowed in all 202 00:14:43,039 --> 00:14:48,039 classrooms. So this was two thousand so and I was at Columbia University, 203 00:14:48,080 --> 00:14:52,240 and I have horrific handwriting. So I had one professor who allowed me to 204 00:14:52,279 --> 00:14:56,720 take my exams on a laptop, and a number of the other students were 205 00:14:56,759 --> 00:15:00,360 kind of resentfull of this. But most students didn't even have a laptop class 206 00:15:00,360 --> 00:15:03,720 to take notes. But now it is really standard to have a laptop and 207 00:15:03,759 --> 00:15:07,120 class to take notes. And the difficulty is you don't know what a student 208 00:15:07,200 --> 00:15:11,879 is doing unless you can see their screen. Yeah, and now the difficulties 209 00:15:11,879 --> 00:15:15,600 that we saw a lot of this during the pandemic. The difficulty is if 210 00:15:15,639 --> 00:15:18,360 you're taking an exam remotely, you have to lock it down as all this 211 00:15:18,480 --> 00:15:22,879 software and all these businesses that are designed to track eye movement and mouse movement, 212 00:15:24,000 --> 00:15:26,559 and they'll even pan the room with your camera to know what else and 213 00:15:26,600 --> 00:15:31,759 who else is in the room. So we've kind of leaped this whole thing 214 00:15:31,159 --> 00:15:35,600 where it's just assumed you have a laptop, and now they're worried about the 215 00:15:35,639 --> 00:15:37,720 laptop doing too much, rather than it shouldn't even be in the room. 216 00:15:39,120 --> 00:15:43,080 I have friends who are professors who say, you have no idea how rampant 217 00:15:43,120 --> 00:15:46,600 cheating is, Like it's unbelievable, no doubt. And you know, it's 218 00:15:46,600 --> 00:15:50,639 funny because there's another bridge point that I want to talk about when we talk 219 00:15:50,679 --> 00:15:54,279 about technology kind of going from oh my god, this is a problem too. 220 00:15:54,399 --> 00:15:58,080 Of course it's here, and I think GPT and things like this, 221 00:15:58,159 --> 00:16:00,519 all this machine learning will go of course it's here, or we'll just at 222 00:16:00,519 --> 00:16:04,159 some point wonder and laugh about this panic. UM I remember as an IT 223 00:16:04,440 --> 00:16:08,919 guy, I was largely on the support side for ad agencies and publishing companies. 224 00:16:10,240 --> 00:16:15,799 UM. I remember teaching people to use the internet. Right where we 225 00:16:15,799 --> 00:16:19,919 went from a completely unnetworked office with standalone computers on desks and the most networking 226 00:16:19,960 --> 00:16:25,799 we have was maybe to a printer, to hooking up email for the first 227 00:16:25,840 --> 00:16:26,720 time, and people going, what am I going to use this for? 228 00:16:29,000 --> 00:16:32,399 What does this do? I don't know anybody outside the office, what is 229 00:16:32,399 --> 00:16:34,240 it? What's an email address? Why is it that this AT symbol kind 230 00:16:34,240 --> 00:16:38,240 of put another symbol there instead. I don't like the AT symbol, you 231 00:16:38,279 --> 00:16:42,120 know. Um lots of stuff like that. One woman printing out all of 232 00:16:42,120 --> 00:16:47,039 her emails and filing them because she's at She's like, oh, you know, 233 00:16:47,240 --> 00:16:52,799 things like that, And now emails emails actually almost day pass a emails 234 00:16:52,799 --> 00:16:55,159 like kind of like people like, oh I don't want email? Who uses 235 00:16:55,159 --> 00:16:59,000 email anymore? So? I don't know I make notes in my contact. 236 00:16:59,039 --> 00:17:02,240 Now, what's the prime medium this person likes? Yeah, how am I 237 00:17:02,240 --> 00:17:03,279 going to reach God? Help us? I wish we had email as the 238 00:17:03,279 --> 00:17:07,880 primary one still, because I wish I have fifteen different ways to reach people. 239 00:17:07,920 --> 00:17:10,759 I don't know which one, but anyway, I just want to talk 240 00:17:10,799 --> 00:17:15,160 about this instead of getting this moral panic and an ethical panic. Really, 241 00:17:15,240 --> 00:17:18,680 let's talk about this as an inflection point. And we've seen these, whether 242 00:17:18,720 --> 00:17:26,279 it's moving from speaking or memorizing our knowledge versus writing it down, whether it's 243 00:17:26,759 --> 00:17:30,000 should we have laptops in the classroom versus having them the class expecting them in 244 00:17:30,039 --> 00:17:36,240 the classroom, to should we have email to expecting to have email. There's 245 00:17:36,240 --> 00:17:40,480 all these different times in history where technology has gone from oh my god, 246 00:17:41,119 --> 00:17:42,720 this is scary too. Of course we have this. This is scary, 247 00:17:42,880 --> 00:17:45,880 right, And I think that's where we are with this. So I know 248 00:17:45,960 --> 00:17:52,079 that you are one of your things on way with words is that you are 249 00:17:52,200 --> 00:17:57,160 very much always trying to talk people off the ledge of being like, you 250 00:17:57,200 --> 00:18:03,599 know, grammar Nazis, usage Nazis, and you're like, like, language 251 00:18:03,680 --> 00:18:06,680 is changing, you need to change with it. You know, it's not 252 00:18:06,720 --> 00:18:08,240 the world doesn't going to stop and change for you. But that does not 253 00:18:08,359 --> 00:18:14,680 excuse figuratively and literally, like I'm not okay with that. Calm yourself. 254 00:18:15,119 --> 00:18:19,119 Yeah, see see this is exactly why. Right. So yeah, you're 255 00:18:19,160 --> 00:18:25,960 like, it's changing, and people wonder why how these things work back in 256 00:18:26,079 --> 00:18:30,440 time? You know, how did something turn into something else? And you 257 00:18:30,480 --> 00:18:33,480 guys do the research. You you're like, well, it's because well maybe 258 00:18:33,480 --> 00:18:37,039 you can give me a better example. But it happens, it changes, 259 00:18:37,119 --> 00:18:42,000 and it's often based on misunderstandings, right, uh. Yeah. Sometimes sometimes 260 00:18:42,000 --> 00:18:48,400 it's just natural where one c one category that was broad becomes specific. For 261 00:18:48,440 --> 00:18:52,039 example, girl used to mean boy, right, how is that weird? 262 00:18:52,319 --> 00:18:55,799 I mean yeah, well, because it was more general, it could mean 263 00:18:56,039 --> 00:19:00,920 a young person in general, you know, or so sometimes broad topic. 264 00:19:00,599 --> 00:19:04,920 One of my favorite ones to talk about is how the very specific psychotherapy term 265 00:19:06,039 --> 00:19:10,319 of ano retentive now just means oh, they're particular. So it left the 266 00:19:10,400 --> 00:19:15,240 specific domain where it's very precise in this profession to be used an everyday language 267 00:19:15,240 --> 00:19:18,880 by people who are not you know, they're not professional at all. Do 268 00:19:18,680 --> 00:19:25,559 you find an author who use that phrase in some kind of popular media that 269 00:19:25,640 --> 00:19:29,799 then it became part of the regular lexicon, it wasn't one as many you 270 00:19:29,839 --> 00:19:34,039 know, the things often leave their specific police jargon, you know, often 271 00:19:34,119 --> 00:19:37,839 leaves, for example, to leave the specific language of the police and show 272 00:19:37,880 --> 00:19:41,039 up in hip hop and then it becomes slang in the street. Right. 273 00:19:41,119 --> 00:19:45,160 So so a lot of times the bridge there is some kind of contact, 274 00:19:45,160 --> 00:19:48,640 but it usually requires repeated contact and not one point of contact. So I 275 00:19:48,640 --> 00:19:53,559 guess, given all that history of you in Martha, are you a little 276 00:19:53,559 --> 00:19:57,839 more scared with AI than the rest of us? Are you? Are you 277 00:19:57,920 --> 00:20:02,960 like, is this just falling into this is just another inflection point and don't 278 00:20:03,000 --> 00:20:04,839 worry and everybody. Yeah, I think that's another inflection point. And you 279 00:20:04,880 --> 00:20:07,519 know, I think on the segment we aired on the show, I can't 280 00:20:07,519 --> 00:20:11,079 remember how it ended up when we edited it. I think I specifically said 281 00:20:11,079 --> 00:20:14,480 I want to avoid the term AI. I don't believe this is AI. 282 00:20:14,920 --> 00:20:17,839 I think it's just machine learning, right, and it is, And to 283 00:20:17,920 --> 00:20:21,720 call it AI is over has a lot of the AI we've been calling AI 284 00:20:21,799 --> 00:20:26,279 for decades is an AI. It's just sophisticated conditionals. Yeah. The way 285 00:20:26,319 --> 00:20:29,920 I've been describing it is artificial intellages is what you call something when it doesn't 286 00:20:29,960 --> 00:20:33,400 work. As soon as it does work, it'll get a new name. 287 00:20:33,559 --> 00:20:36,000 Yeah. Yeah, Well, although I think that some of the stupid things 288 00:20:36,000 --> 00:20:41,440 it's producing are just as stupid as humans. Potentially stupider, Yeah, potentially 289 00:20:41,480 --> 00:20:45,960 stupider. The opposite of artificial intelligence is natural stupidity, right, Yeah, 290 00:20:45,960 --> 00:20:48,240 although you know, it's a ton of operator air from what I'm saying, 291 00:20:48,240 --> 00:20:51,880 a lot of those stupid results are just because the operators don't know what the 292 00:20:51,880 --> 00:20:55,559 hell they're going. Why are you having existential debates with a search engine? 293 00:20:55,720 --> 00:20:57,240 Right? We all saw the one where we're all laughing, Like the guy 294 00:20:57,279 --> 00:21:00,480 who wrote this very ponderous piece at the Washington Boats were like, yeah, 295 00:21:00,559 --> 00:21:06,559 dude, you're like completely treating it like an entity. It's literally every single 296 00:21:06,599 --> 00:21:10,240 time you ask it a question, it's only uh, I know, it 297 00:21:10,519 --> 00:21:12,839 doesn't have a brain, you're acting like as a brain. It's software. 298 00:21:14,240 --> 00:21:18,200 But here's the thing, though, Um, companies are going to get dinged 299 00:21:18,279 --> 00:21:22,359 for these things representing them in ways they don't want to be represented. Yeah. 300 00:21:22,599 --> 00:21:26,799 Um, this has just happened with with being right. Microsoft, they 301 00:21:26,880 --> 00:21:34,240 basically allowed open ended discussions with their AI, and what happens they've noticed is 302 00:21:34,279 --> 00:21:40,640 after maybe twenty iterations back and forth on the same subject, the box starts 303 00:21:40,680 --> 00:21:44,599 to lose its You know what, Now I now know why I actually was 304 00:21:44,640 --> 00:21:47,359 able to talk to some folks on the inside. Yeah, because it's a 305 00:21:47,400 --> 00:21:52,720 software problem. It's a bug interesting like a bug bug, or like it's 306 00:21:52,839 --> 00:21:56,160 it's it is solvable. They brought but they solved it by not hitting the 307 00:21:56,240 --> 00:22:00,559 limits. Look this whole mechanism about context, right, yeah. Yeah, 308 00:22:00,640 --> 00:22:06,079 So your parser reads what you've written and it tokenizes it, so it knows 309 00:22:06,079 --> 00:22:07,839 how to respond to it, and it puts it into cash. So the 310 00:22:07,920 --> 00:22:11,759 next thing, then it writes a response, which is also tokenized. Then 311 00:22:11,799 --> 00:22:15,559 you write your next response in order to maintain context, it tokenizes that, 312 00:22:15,599 --> 00:22:18,839 then it reads them all together and creates a response. Eventually you overflow the 313 00:22:18,880 --> 00:22:22,799 cash. Oh so you've got too many layers of recursive tokenizing. That's right, 314 00:22:23,160 --> 00:22:26,440 you just run out of room. They've only allocated so much space for 315 00:22:26,480 --> 00:22:30,039 tokens. This is why you get it. Well, the oldest tokens fall 316 00:22:30,079 --> 00:22:33,000 off the back, and the oldest tokens are typically the ones shaping things. 317 00:22:33,079 --> 00:22:36,720 Yeah, because you need the umbrella thought. Because a lot of times your 318 00:22:36,720 --> 00:22:41,440 first question sets your umbrella, your umbrella. Yeah. The best demonstration I've 319 00:22:41,480 --> 00:22:42,759 done of this is I fired up chat GPT. It says, only talk 320 00:22:42,799 --> 00:22:48,240 to me an iambic pentameter, and then just chat away with it, and 321 00:22:48,319 --> 00:22:52,240 at some point it forgets. I love when you give it rules and it's 322 00:22:52,799 --> 00:22:56,720 like a four year old. Right. Yeah, I told you not to 323 00:22:56,799 --> 00:22:59,160 talk to me, but you want to know what the limits of the cash 324 00:22:59,160 --> 00:23:02,200 are. Put key instructions at the beginning of the cap. But to get 325 00:23:02,279 --> 00:23:06,400 back to the bing thing, this chatbot kept insisting to New York Times reporter 326 00:23:06,599 --> 00:23:10,400 Kevin Ruse I guess that he didn't actually love his wife and said that it 327 00:23:10,440 --> 00:23:14,480 would like to steal nuclear secret and yeah, okay, but it's a bot. 328 00:23:14,640 --> 00:23:18,079 It's not going to steal. There was another situation where somebody in one 329 00:23:18,079 --> 00:23:22,720 of our rds actually convinced the bot that, you know, fifteen plus three 330 00:23:22,839 --> 00:23:26,160 equal twenty one, and the bot said, yes, I'm sorry, you 331 00:23:26,200 --> 00:23:30,680 are right, fifteen plus three equals twenty one. And then I to test 332 00:23:30,759 --> 00:23:33,920 it, went back and asked it, what's fifteen plus three and said eighteen. 333 00:23:34,119 --> 00:23:37,359 Yeah, because it didn't remember that previous conversation, because that's not how 334 00:23:37,359 --> 00:23:41,400 it's spelled. It doesn't have the context from all the other conversations it's had 335 00:23:41,720 --> 00:23:45,839 with all these other people. It only has its context from you, because 336 00:23:45,839 --> 00:23:48,920 when it does combine all the contexts, you end up with TA. Okay, 337 00:23:48,960 --> 00:23:53,200 you got to explain that the original TA was the original Microsoft bot. 338 00:23:53,200 --> 00:23:57,799 I go back at twenty eighteen, which within ten hours had turned into a 339 00:23:57,880 --> 00:24:02,839 race of psychopath because it was combining everything that people were throwing at it, 340 00:24:02,920 --> 00:24:04,200 right, Yeah, and they were, and they were testing its limits. 341 00:24:04,200 --> 00:24:07,680 There are tons of edge cases. It wasn't when they took it down. 342 00:24:07,799 --> 00:24:10,920 And yeah, I'm still not convinced they're not going to take this down. 343 00:24:11,160 --> 00:24:14,119 I'm not. I'm not convinced either, and I don't. I think it's 344 00:24:14,119 --> 00:24:17,920 a wonderful experiment. I think it's an experiment on the people using it as 345 00:24:18,039 --> 00:24:22,160 much on the software itself. I wonder how much of this is almost another 346 00:24:22,240 --> 00:24:26,319 kind of Stanford experiment. Yeah, maybe how lonely are people that they're having 347 00:24:26,359 --> 00:24:30,279 these conversations, you know, I want to I want to tackle this from 348 00:24:30,279 --> 00:24:33,680 another angle as well. You know what, This also reminds me of a 349 00:24:33,720 --> 00:24:40,000 lot of the complaints about Mechanical Turk, when people were talking about what happens 350 00:24:40,039 --> 00:24:45,640 when you put all this effort out to people to automate stuff that you know, 351 00:24:47,240 --> 00:24:49,240 pennies on the task, right, And I've done that, and actually, 352 00:24:49,279 --> 00:24:52,400 you know, one of my other lives, one of the reasons I 353 00:24:52,440 --> 00:24:56,079 do the radio show, it was as a dictionary editor. So I actually 354 00:24:56,160 --> 00:25:00,400 did a paper for the Dictionary Society of North America and presented it what when 355 00:25:00,400 --> 00:25:03,559 you ask people on mechanical turk to write dictionary entries? And so I found 356 00:25:03,599 --> 00:25:08,160 a bunch of fairly common words that weren't in any major dictionary, and I 357 00:25:08,599 --> 00:25:12,599 created a bunch of tasks. And first I had people find sample sentences use 358 00:25:12,680 --> 00:25:15,920 these words, and then I had another whole bunch of people use those sample 359 00:25:15,960 --> 00:25:19,519 senses to approximate dictionary style entry. And then I had another batch of people 360 00:25:21,279 --> 00:25:25,480 trim them down and make them look like ractual dictionary entries with parts of speech. 361 00:25:25,599 --> 00:25:29,559 And it actually worked pretty well. It was more expensive than having an 362 00:25:29,559 --> 00:25:33,440 actual dictionary editor to do it. But and now I have a paper proposal 363 00:25:33,519 --> 00:25:38,480 before that same Dictionary Society to see what chat GPT does. And I'm also 364 00:25:38,519 --> 00:25:45,759 going to use stable diffusion or a variety of some other AI image generator to 365 00:25:45,799 --> 00:25:53,720 create dictionary style illustrations for those definitions. See if I can't scare the lexicographers, 366 00:25:53,880 --> 00:26:00,119 nice do blaying up against the plagiarism aspect too. I really that Martha 367 00:26:00,200 --> 00:26:04,440 did that work to check all the sentences. Just because you had you scrape 368 00:26:04,480 --> 00:26:07,599 something off the web and run it through a synonym engine, you know, 369 00:26:07,680 --> 00:26:11,559 doesn't make it original work, right, it just doesn't search well, yeah, 370 00:26:11,559 --> 00:26:15,799 because it's it's even though it's not good texts that generates it's plausible right, 371 00:26:15,880 --> 00:26:18,400 Yes, Now it's definitely a Dunning Krueger amplifier. If you know nothing 372 00:26:18,440 --> 00:26:22,799 about a subject, all this stuff looks great. Well, that's Wikipedia, 373 00:26:23,119 --> 00:26:30,440 indeed. Yeah. I love going to Wikipedia and finding like something that's been 374 00:26:30,440 --> 00:26:33,839 in there for years and I know it is fundamentally wrong. Yeah. I 375 00:26:33,160 --> 00:26:38,319 go to Wikipedia for the footnotes. It gives me interesting places to go look 376 00:26:38,359 --> 00:26:42,160 for information. It's not information itself. Well, that's the researcher's attitude. 377 00:26:42,200 --> 00:26:45,240 I go to I go find a paper and I look at its bibliography and 378 00:26:45,400 --> 00:26:49,400 suddenly I know how the path to become an expert on the thing. Sure 379 00:26:49,799 --> 00:26:53,880 you at least know what those authors thought was important, whether they found or 380 00:26:53,920 --> 00:27:00,400 at least you know what other papers they pillaged for. Right. So another 381 00:27:00,400 --> 00:27:03,920 point of inflection that you guys talked about on the show and still talk about 382 00:27:03,000 --> 00:27:10,440 quite a bit is texting language and how you know people are afraid because kids 383 00:27:10,440 --> 00:27:12,319 have their phones and their you know, their heads down in their phones. 384 00:27:12,359 --> 00:27:17,480 And then we don't talk about it that much anymore. But you you talked 385 00:27:17,480 --> 00:27:19,880 about that book and I went and read the book because internet is at it. 386 00:27:21,440 --> 00:27:23,559 Yeah. Yeah, but Greta McCulloch, it's a fantastic book. It's 387 00:27:23,559 --> 00:27:27,759 one of the best books written on internet language ever. It's being used in 388 00:27:27,759 --> 00:27:30,440 classrooms now, that's how good it is. It's awesome, and that the 389 00:27:30,440 --> 00:27:36,000 basic tenant is, look, you know these this is definitely changing the language. 390 00:27:36,599 --> 00:27:40,200 You know, law is a word now, not just land of lakes. 391 00:27:40,720 --> 00:27:42,799 Yeah, it's in context. Yeah, and it doesn't mean lots of 392 00:27:42,839 --> 00:27:47,519 love. It depends on the context. But yeah, it's context sensitive. 393 00:27:47,799 --> 00:27:52,319 What the reason I said we don't talk about texting languages for some people that 394 00:27:52,880 --> 00:27:57,480 in their mind it's just these very specific things like b R B or LLL 395 00:27:57,839 --> 00:28:02,680 or OMG. But there's so much more to it. And Gretchen gets into 396 00:28:02,720 --> 00:28:10,880 that about all these different ways that we express emphasis and laughter and sarcasm or 397 00:28:11,240 --> 00:28:17,799 community or insiderness or outsiderness and hesitation. There's so much more to there's so 398 00:28:17,880 --> 00:28:22,160 many more layers, and what you think of as an error maybe having may 399 00:28:22,200 --> 00:28:26,519 have a message attached to it. Lack of capitalization or punctuation may have a 400 00:28:26,640 --> 00:28:32,160 point, and you need to understand that. But what's really important is that 401 00:28:32,240 --> 00:28:37,240 again the moral panic about what does it all mean when people don't capitalize and 402 00:28:37,279 --> 00:28:42,279 punctuate their text messages has vanished. It's mostly gone. And what we discovered 403 00:28:42,400 --> 00:28:47,880 was that people really do care about their speech. And then when we all 404 00:28:47,960 --> 00:28:52,599 moved from those bar phones that had we only had ten nine entry of our 405 00:28:52,759 --> 00:28:59,440 text messages and went to touch phones with haptic response, with full keyboards, 406 00:29:00,119 --> 00:29:03,279 we could finally type whole sentences. We did, right, most of us 407 00:29:03,359 --> 00:29:07,079 do. Most of us type the best that we can. But then it 408 00:29:07,200 --> 00:29:10,960 also then we also were brought in a chart of emojis, of ever ever 409 00:29:11,000 --> 00:29:15,240 expanding emojis. Yeah, and those actually have expanded our ability to communicate. 410 00:29:15,400 --> 00:29:19,599 They're the new hieroglyphs. Well, but they're necessary. There's something I talk 411 00:29:19,680 --> 00:29:23,559 about and I don't have given this presentation a while, but I'll the version 412 00:29:23,599 --> 00:29:30,279 of it I'll give to you is there's something called parallinguistic restitution. The vocal 413 00:29:30,440 --> 00:29:36,400 version of speech is speech. The written version is only a pale invitation of 414 00:29:36,400 --> 00:29:41,279 what we speak aloud, and it'll never catch up with the spoken version ever, 415 00:29:41,720 --> 00:29:45,200 And so we're constantly trying to make the written version accommodate all the things 416 00:29:45,240 --> 00:29:49,359 we can do with our spoken language, and it really never catches up. 417 00:29:49,400 --> 00:29:56,000 But paral linguistic restitution is what we do when we make things like emoji, 418 00:29:56,400 --> 00:30:00,000 or when we make things like emoticons, or when we do clever little things 419 00:30:00,039 --> 00:30:03,960 with punctuation. It's when we're adding back in gestures by making a smiley face, 420 00:30:04,440 --> 00:30:08,400 or we're adding back in a grimace by putting the right emojians. Still, 421 00:30:08,440 --> 00:30:11,640 I never felt older than when my teenager came up to me and said, 422 00:30:12,119 --> 00:30:18,920 hashtag I'm hungry. I don't know about old, but I've always thought, 423 00:30:18,000 --> 00:30:21,839 you don't have to say hashtag. You don't have to say hashtag yeah 424 00:30:21,880 --> 00:30:25,680 okay. But also every emotions no word following it. When it's just the 425 00:30:25,720 --> 00:30:27,400 hash symbol, it's not a hashtag. It has to have a word following 426 00:30:27,400 --> 00:30:32,240 it, so to stop that. Yeah, it's you can call it a 427 00:30:32,279 --> 00:30:34,759 hash. You can remember. There are some things that really Bud Graham Barrett, 428 00:30:34,799 --> 00:30:40,000 this is good to know a few. What are some of the others. 429 00:30:40,759 --> 00:30:44,839 I'm not telling you have to listen to all five hundred episodes. Well 430 00:30:44,880 --> 00:30:47,599 I have. I appreciate that. I just don't remember them. I feel 431 00:30:47,599 --> 00:30:48,720 sorry for you. Who made you do that? Now? Me? I 432 00:30:48,759 --> 00:30:52,359 did it on my own. Like when he was in prison, they didn't 433 00:30:52,160 --> 00:30:55,880 play like Guns and Rosens to full Ball and they played our show. Nice 434 00:30:56,359 --> 00:31:00,759 guys made two thousand plus podcasts. He does not recommending through the whole back 435 00:31:00,799 --> 00:31:04,400 catalog, but he does it himself anyway. Oh, it's its own kind 436 00:31:04,400 --> 00:31:07,640 of OCD. And so that's those are those are prison bars I see in 437 00:31:07,680 --> 00:31:12,039 the background there absolutely okay, And I got interrupt for one moment for this 438 00:31:12,160 --> 00:31:18,039 very important message. You know, Amazon Aws is a great home for your 439 00:31:18,079 --> 00:31:22,720 dot net applications. Whether you're looking to run your apps as serverless functions, 440 00:31:23,160 --> 00:31:29,599 have them running containers, or you require the complete flexibility of virtualized hardware, 441 00:31:30,119 --> 00:31:33,400 you can do it all on Aws. Dot Net toolkits and templates guide you 442 00:31:33,440 --> 00:31:37,559 through Bill and deployment, or you can use your favorite infrastructure as code and 443 00:31:37,640 --> 00:31:45,519 CICD tools including Azure DevOps, GitHub Actions and git labs, CICD loving Visual 444 00:31:45,559 --> 00:31:48,920 Studio vs Code or Rider. There's an AWS toolkit for each of them. 445 00:31:49,359 --> 00:31:55,160 Want to use familiar Microsoft Sequel server databases, you can with the Amazon Relational 446 00:31:55,240 --> 00:32:00,319 Database service, which provides fully managed Sequel server databases, then offer auto manic 447 00:32:00,400 --> 00:32:05,359 scaling failover in snapshots. Once your dot net application is deployed, it can 448 00:32:05,440 --> 00:32:10,200 use any of the over two hundred AWS services to extend its functionality. Interested 449 00:32:10,240 --> 00:32:15,680 in AI, IoT, machine learning, video processing, or transcription, your 450 00:32:15,720 --> 00:32:21,119 dot Net on AWS application can do at all. To learn more, go 451 00:32:21,160 --> 00:32:29,960 to AWS dot Amazon dot Com, slash net and we're back. It's dot 452 00:32:29,960 --> 00:32:32,119 net Rocks. I'm Richur Campbell. That's call Franklin, Yo. Talking to 453 00:32:32,160 --> 00:32:37,960 our fellow podcaster Grant Barrett, who's also a lexographer. And I didn't really 454 00:32:37,960 --> 00:32:43,000 think that lexographers were that much fun. It's lexicographer, a lexicographer. Yeah, 455 00:32:43,000 --> 00:32:49,359 well you're you're okay with with US languages? Yeah, you're in good 456 00:32:49,359 --> 00:32:52,720 company though. The very first radio interview I ever did was with an Australian 457 00:32:52,920 --> 00:32:55,960 station promoting my first book, and the guy called me a lexiographer the whole 458 00:32:55,960 --> 00:33:00,079 time and igrapher. Yeah, when you're doing a live radio interview, you 459 00:33:00,160 --> 00:33:04,039 just don't correct people. It's like correcting your father at the dinner table. 460 00:33:04,039 --> 00:33:07,039 You just don't do it. All right, So we're not going to learn 461 00:33:07,200 --> 00:33:12,480 the other pet peeves unless you listen to a way with words. But certainly 462 00:33:12,559 --> 00:33:15,920 this sort of inflection point, as you call it, is really interesting, 463 00:33:16,039 --> 00:33:22,839 especially for programmers because this, you know, for years, other people have 464 00:33:22,960 --> 00:33:28,079 been looking at us as programmers, as people that are replacing their jobs. 465 00:33:28,359 --> 00:33:34,559 Right, we automate. That's our job to write automation software to make people's 466 00:33:34,559 --> 00:33:38,400 who do those jobs either less brain dependent, you know, push the button, 467 00:33:39,079 --> 00:33:43,160 or eliminate them all together. I mean that's the perception. I'm not 468 00:33:43,200 --> 00:33:49,160 saying it's true. I would say that software has increased the number of available 469 00:33:49,240 --> 00:33:53,960 jobs and change those jobs from sort of menial tasks to things that have meaning. 470 00:33:54,519 --> 00:33:59,039 But they're new jobs and on people to retrain. And if you don't 471 00:33:59,039 --> 00:34:01,440 want to retrain, that's on them. But now the shoes on the other 472 00:34:01,480 --> 00:34:05,720 foot. Now programmers are like, oh my god, chat GPT, you 473 00:34:05,720 --> 00:34:07,159 can tell it to write a program that does acts and it will just spit 474 00:34:07,199 --> 00:34:12,880 it out. Might not be right, but I've actually used it, and 475 00:34:13,159 --> 00:34:15,639 you know, how do how do I do that little thing again? And 476 00:34:15,679 --> 00:34:19,920 I asked it and for the most part it works, but sometimes it doesn't. 477 00:34:19,639 --> 00:34:22,199 Like I totally do that because I'm the worst. The most program I 478 00:34:22,280 --> 00:34:24,920 know is a little bit of Pearl script doing a little bit of a little 479 00:34:24,960 --> 00:34:29,800 bit of Python, and mainly I just like, yeah, but you ran 480 00:34:29,840 --> 00:34:34,320 an Exchange struggle of all these crap scripts that I do my little work with, 481 00:34:34,440 --> 00:34:37,599 but you ran Exchange, and this is where you and Richard have common 482 00:34:37,679 --> 00:34:40,880 but we both know pain. Oh god, I did pain. But also 483 00:34:40,960 --> 00:34:45,360 I got to work with I actually loved like things like Skip or Cisco servers, 484 00:34:45,400 --> 00:34:51,440 you know, blocking blocking everything coming in with us, the original iOS. 485 00:34:52,199 --> 00:34:54,000 It was just nice. It was just nice just to see all that 486 00:34:54,039 --> 00:34:59,079 stuff bounced back and never enter your network. Just watch those logs and you're 487 00:34:59,079 --> 00:35:01,840 like, they did it. I did it. All that intrusion not showing 488 00:35:01,880 --> 00:35:07,599 up into my networks weak. It is interesting that programmers now are feeling that 489 00:35:07,719 --> 00:35:14,280 existential dread of you know, Chat GPT kind of taking their jobs. You 490 00:35:14,320 --> 00:35:17,639 know, you know, it's not just an inflection point, it's I want 491 00:35:17,679 --> 00:35:22,039 to emphasize that point about it being a bridge. I've had the good fortune 492 00:35:22,159 --> 00:35:27,639 to be on a lot of different bridge points in history. For example, 493 00:35:27,679 --> 00:35:32,000 when I worked in newspapers, I bridged the print paste up era where you 494 00:35:32,039 --> 00:35:37,880 actually paste it up the paper and shot it with a camera. But that's 495 00:35:37,880 --> 00:35:40,079 how you went to press to the digital era, where you did it all 496 00:35:40,199 --> 00:35:44,960 on a you know, a Mac see with a floppy test and PageMaker. 497 00:35:45,920 --> 00:35:49,480 And I remember going to the press and there was a guy working in the 498 00:35:49,519 --> 00:35:53,079 press room who was very disgruntled that I was taking his job because we were 499 00:35:53,119 --> 00:35:55,559 doing it on a Mac and we weren't doing it the old fashioned way. 500 00:35:55,559 --> 00:36:00,840 Were used a line of type machine to output the ages. There was this 501 00:36:00,239 --> 00:36:06,639 giant thing that involved chemicals and this weird custom keyboard, and he was very 502 00:36:06,639 --> 00:36:08,719 angry about it that his job was gone. You know, there's all of 503 00:36:08,880 --> 00:36:14,880 these technological innovations. Somebody's going to do it shifted, but if they can't 504 00:36:14,960 --> 00:36:17,400 keep up, it's on them. That's how I feel. Yeah, I 505 00:36:17,440 --> 00:36:20,679 mean, I want to have some compassion for Yeah, you're going to need 506 00:36:20,679 --> 00:36:23,360 to be retrained. Yeah, but I also believe in the growth model here. 507 00:36:23,679 --> 00:36:28,599 When we brought the web into travel, we eliminated a lot of travel 508 00:36:28,639 --> 00:36:32,400 agents, but we actually grew the overall travel travel business so much there's many 509 00:36:32,480 --> 00:36:37,440 more jobs than the ones that were eliminated. I just don't think that getting 510 00:36:37,519 --> 00:36:40,800 chatchyp Hall to find some code for you to use is that different from you 511 00:36:40,880 --> 00:36:45,320 googling it or stackle, stack overflow. It's the same gig. It's just 512 00:36:45,360 --> 00:36:47,599 like now you're doing any different context, and I think it are I think 513 00:36:47,639 --> 00:36:52,440 it shifts the low end of the business. It shifts the entry level business, 514 00:36:52,480 --> 00:36:57,880 and maybe that's the larger problem those It shifts an entry point for the 515 00:36:57,920 --> 00:37:01,360 beginning programmer. How to I enter as a brand new programmer? How do 516 00:37:01,440 --> 00:37:07,760 I get my name recognition out there on the different freelance marketplaces. How do 517 00:37:07,840 --> 00:37:10,960 I start, and as in a small country, to make a few bocks. 518 00:37:12,280 --> 00:37:15,719 If GPT is going to do this for everyone, you know, how 519 00:37:15,760 --> 00:37:19,239 do I make a little bit of extra money on the side. I don't 520 00:37:19,239 --> 00:37:21,920 know. There are some people who are going to be cut out and retraining 521 00:37:21,920 --> 00:37:24,239 isn't going to solve it for them. Well, and folks have talked about 522 00:37:24,280 --> 00:37:30,599 you'll use chap EGPT and your ability to craft requests of that engine. It 523 00:37:30,000 --> 00:37:32,719 turns out to be your actual skill, and in the end you have to 524 00:37:32,719 --> 00:37:37,400 decide whether or not what they gave you is good or not. Back in 525 00:37:37,440 --> 00:37:42,400 the early days of programming, I sort of glibly said, you know, 526 00:37:42,440 --> 00:37:45,360 maybe the only job left in the world someday will be programmers. Amber Scott 527 00:37:45,400 --> 00:37:51,559 Stanfield said, yeah, and battery makers. Programmers and battery makers. That's 528 00:37:51,599 --> 00:37:52,880 those of you, the only jobs left. Now I'm thinking, you know, 529 00:37:53,159 --> 00:37:58,880 proofreaders might be the only job left someday in the literary world. Oh 530 00:37:58,960 --> 00:38:01,440 dude, they're being cut in the newspaper business, so I wouldn't even publishing 531 00:38:01,440 --> 00:38:05,239 business. I wouldn't even count on that. Really. Oh yeah, yeah, 532 00:38:05,280 --> 00:38:08,039 it's it's they're one of the first jobs to be cut in publishing. 533 00:38:08,320 --> 00:38:12,480 Yeah. Well it shows, um, it does, it really does. 534 00:38:12,719 --> 00:38:15,039 I can't tell you how many errors I see. And my wife, Kelly, 535 00:38:15,119 --> 00:38:21,519 Kelly, my wife is very um attentive when it comes to details, 536 00:38:21,679 --> 00:38:24,239 right, and so she can spot, she says. She also says she 537 00:38:24,239 --> 00:38:29,280 should be a continuity person on a movie set. M H, same thing, 538 00:38:29,400 --> 00:38:32,239 details details. Your coffee cup is in your left hand, that last 539 00:38:32,239 --> 00:38:35,719 shot. You moved it to your right hand, putting your arms up not 540 00:38:35,880 --> 00:38:38,840 down. Yeah. Right. But I wonder if we're on a I wonder 541 00:38:39,159 --> 00:38:43,559 if we are, have been, and are now accelerating down a path of 542 00:38:43,800 --> 00:38:50,719 like a version of yellow journalism, this idea of the quality eventually the you 543 00:38:50,719 --> 00:38:55,039 know, yellow journalism was that the time when folks were sensationalizing news back in 544 00:38:55,079 --> 00:38:59,920 the in the broadsheet era, and eventually people stopped reading it because it was 545 00:39:00,039 --> 00:39:04,000 garbage and the industry had to change. And I wonder if we're not on 546 00:39:04,000 --> 00:39:07,480 that exact path where it's like, hey, quality matters, and eventually people 547 00:39:07,480 --> 00:39:13,079 will ignore it because it's garbage. Well, I've sat in on a lot 548 00:39:13,239 --> 00:39:16,079 of the future of news panels. There's nothing news people love to talk about 549 00:39:16,119 --> 00:39:21,159 more than the future of their business. And this is a this has been 550 00:39:21,199 --> 00:39:22,840 a topic for as long as I've been involved in the news business, which 551 00:39:22,880 --> 00:39:30,039 is like thirty years exactly this question. And generally the consensus is, and 552 00:39:30,079 --> 00:39:35,519 it has always been, there's not enough division between opinion and news, and 553 00:39:35,679 --> 00:39:38,960 there never will be because there's too much money and opinion and the public doesn't 554 00:39:38,960 --> 00:39:43,079 know how to separate the two either. When you put opinion at the top 555 00:39:43,119 --> 00:39:45,519 of the page or on the chiron at the bottom of the screen, people 556 00:39:45,519 --> 00:39:50,719 still think it's news. But we're talking about an inflection point. This may 557 00:39:50,760 --> 00:39:53,679 be part of that inflection point is to say that that we do need to 558 00:39:53,719 --> 00:39:59,119 discriminate on these things. It has happened before. I cite the yellow journalism 559 00:39:59,159 --> 00:40:01,800 case as a what are the few cases where it did happen. Yeah, 560 00:40:01,800 --> 00:40:06,159 but what I'm saying. What I'm saying is in the last thirty years of 561 00:40:06,679 --> 00:40:08,920 all we've seen is the rise of the twenty four hour news channel, which 562 00:40:08,960 --> 00:40:15,320 is what eighty percent opinion and twenty percent hard news, eighty percent advertising, 563 00:40:15,480 --> 00:40:21,119 ten percent opinion and ten percent news. Yeah, well, and lots of 564 00:40:21,159 --> 00:40:23,760 repeating over and over and over again. It's expensive. What do you think 565 00:40:23,760 --> 00:40:30,119 of Twitter? Well, God help me. Yeah, we kind of feeling 566 00:40:30,159 --> 00:40:34,000 that dread too. I mean, it was such a great language research tool 567 00:40:34,119 --> 00:40:37,360 until these new fees. I guess they're not underway quite yet, but these 568 00:40:37,400 --> 00:40:43,559 new fees are now threatening a great language research tool. Yeah, where the 569 00:40:43,960 --> 00:40:52,039 geocoding on tweets meant that you could actually do really sophisticated time based, region 570 00:40:52,119 --> 00:40:58,760 based database searches. And say this is a term is more common here and 571 00:40:58,840 --> 00:41:02,039 less common there um and has been over the last ten years or however long 572 00:41:02,079 --> 00:41:08,000 Twitter has been a thirteen years? How long his Twitter fifteen maybe fifteen m 573 00:41:08,480 --> 00:41:14,480 And so now we have longitudinal data and it's fantastic for that, And now 574 00:41:14,519 --> 00:41:16,599 that's going to be just thrown in the crapper. Yeah, I'm not. 575 00:41:16,639 --> 00:41:20,960 I mean, I'm not convinced. I don't think Twitter is all that important 576 00:41:20,960 --> 00:41:22,320 to Elon. Actually, I think he's mostly just trying to get out of 577 00:41:22,320 --> 00:41:25,039 the situation he's in, Like, I wonder if we're not going to get 578 00:41:25,039 --> 00:41:29,400 back to something, but he blew up the revenue model, so he's just 579 00:41:29,400 --> 00:41:32,679 trying to scrounge around for revenue. Dingdong. Yeah, I agree, what 580 00:41:32,800 --> 00:41:37,639 a terrible, terrible waste of time. Yes, it was a good product. 581 00:41:37,360 --> 00:41:39,480 Maybe it'll get back there. I don't Yeah, maybe you're right, 582 00:41:39,519 --> 00:41:43,880 Maybe it we'll get back there, because ultimately it doesn't bat at the same 583 00:41:43,960 --> 00:41:46,760 level as getting people to Mars are inventing the electric car, Like, why 584 00:41:46,760 --> 00:41:50,360 are you wasting time out? Yes, I just wish he would spend more 585 00:41:50,360 --> 00:41:52,400 time with his family. Yeah, yeah, that that dude needs a friend. 586 00:41:52,440 --> 00:41:54,920 Like in a worse way. I'm just saying, like out of the 587 00:41:54,920 --> 00:41:59,800 public eye, he just he needs attention, and he's got all those people 588 00:42:00,079 --> 00:42:02,239 who could get him all the attention he ever wanted. Yeah, I think 589 00:42:02,239 --> 00:42:07,480 he went off the rails when Grimes left. Like the Elon that was kind 590 00:42:07,480 --> 00:42:09,079 of fun. You're the guy who said let's fly my sports car into space 591 00:42:09,159 --> 00:42:13,480 like that. Yeah, that was a great elon. And that's when he 592 00:42:13,559 --> 00:42:15,800 had he had a significant other around him. That was, you know, 593 00:42:15,880 --> 00:42:20,039 at least something you reflect against this and she's been gone. He's just gotten 594 00:42:20,039 --> 00:42:22,440 progressively crazier. Yeah. I don't I don't know enough about his personal life. 595 00:42:22,480 --> 00:42:25,199 I just I just know how you know. I'm just trying to focused 596 00:42:25,360 --> 00:42:30,360 on the utility of the tool rather than the nature of the man. And 597 00:42:30,400 --> 00:42:35,639 the utility of the tool is is decline, especially for a guy like you, 598 00:42:35,920 --> 00:42:38,840 like he were saying, a great research tool for language usage. It 599 00:42:38,920 --> 00:42:43,559 really is. Look at the work by Jack Grieve if you just type in 600 00:42:43,639 --> 00:42:46,719 Jack Grieves g R I e vee just type in Jack Grieve linguist. He's 601 00:42:46,760 --> 00:42:54,639 written some phenomenal papers using Twitter and showing African American English online in British English 602 00:42:54,639 --> 00:43:00,920 and just different ways you can track regionality using Twitter. It's brilliant stuff, 603 00:43:00,079 --> 00:43:06,639 just brilliant stuff. Well, and that only works if Twitter is serving as 604 00:43:06,679 --> 00:43:12,079 that sort of Piazza model where you have a significant portion of the population's routinely 605 00:43:12,119 --> 00:43:21,280 communicating, right, a broad spectrum across all ethnic backgrounds and economic backgrounds and 606 00:43:21,360 --> 00:43:23,559 political backgrounds. Right. But if it's going to shift very far right, 607 00:43:23,599 --> 00:43:28,519 we're going to lose that. Are there any other tools out there that can 608 00:43:29,000 --> 00:43:31,400 that have potential in your mind to be used in the same way someday? 609 00:43:31,599 --> 00:43:36,239 No, No, no, nothing else geocodes like that. It's not online. 610 00:43:36,239 --> 00:43:38,039 As far as I know, I've not thought of Twitter as all that 611 00:43:38,159 --> 00:43:42,679 special until I hear what you do with Yeah, I'm like, Okay, 612 00:43:42,960 --> 00:43:45,360 it's because it's not the product of Twitter is. It's a byproduct, but 613 00:43:45,400 --> 00:43:49,199 it's an important byproduct. Yeah, that's right. That A lot of the 614 00:43:49,239 --> 00:43:53,719 work that we do is linguist and mexicographers is based on byproduct. For example, 615 00:43:53,920 --> 00:44:00,079 I learned years ago to use Google alerts to track new words because writers, 616 00:44:00,199 --> 00:44:05,079 authors, journalists will often tag words that are new to the vocabulary with 617 00:44:05,119 --> 00:44:10,280 phrases like also known as are known to police as are referred to by referred 618 00:44:10,320 --> 00:44:15,719 to by um chemists as. So that I have like three or one hundred 619 00:44:15,840 --> 00:44:20,039 or so expressions, and then following those expressions, is a word is a 620 00:44:20,039 --> 00:44:22,880 new word? Yeah? Possibly a new word? Yeah. I was gonna 621 00:44:22,880 --> 00:44:24,159 say, how do you how do you filter for new words? Yeah, 622 00:44:24,280 --> 00:44:28,800 but no, it's it's when people are trying to relate that contact for the 623 00:44:28,880 --> 00:44:31,880 relation phrases, yeah, and then they'll sometimes define them right, and so 624 00:44:32,280 --> 00:44:37,000 you know, it's not very successful. I mean, I could probably automate 625 00:44:37,039 --> 00:44:40,159 that if I wanted to hire somebody, But you know, for a while 626 00:44:40,159 --> 00:44:44,079 there, I was just reading those alerts and marking down the words that were 627 00:44:44,079 --> 00:44:46,719 new to me also, and it was pretty successful. I found thousands and 628 00:44:46,760 --> 00:44:51,480 thousands of words that way. Yeah. I think you also used Google books 629 00:44:51,719 --> 00:44:55,159 right to find the in other tools that you have at your disposal, to 630 00:44:55,239 --> 00:45:00,480 find the first time word has been scene or a phrase has been seen in 631 00:45:00,559 --> 00:45:04,840 print. Yeah, you know, it's always like, as far as we 632 00:45:04,880 --> 00:45:07,440 know, you always have to add that shad that caveat as far as we 633 00:45:07,480 --> 00:45:09,599 know. Yep. These days, I more often use archive dot org. 634 00:45:10,079 --> 00:45:15,119 Yeah, because they have done they have a grant great new search engine by 635 00:45:15,159 --> 00:45:20,039 the way that it allows date range searches. And although the OCR is kind 636 00:45:20,079 --> 00:45:23,880 of iffy because it uses I think a modified version of tess iact, which 637 00:45:23,960 --> 00:45:28,400 is way better than it used to be, but it's still not perfect and 638 00:45:28,440 --> 00:45:31,920 it often depends upon script recognition, which isn't that great, so sometimes it 639 00:45:32,000 --> 00:45:37,320 will misrecognize, for example, of Cirillac as another language. But it's pretty 640 00:45:37,320 --> 00:45:42,159 good. I just found in an old um website that I used to run 641 00:45:42,159 --> 00:45:45,639 and started in nineteen ninety four, Carling Carey's VB homepage. It's where I 642 00:45:45,679 --> 00:45:49,440 first ran across you back in the day. Yeah, but what I'm talking 643 00:45:49,440 --> 00:45:52,760 about is the books portion of archive dot org, and that's where the gold 644 00:45:52,920 --> 00:45:57,360 is. The wet portion, the books portion of Archive, That's where the 645 00:45:57,360 --> 00:46:00,239 gold is. Yeah. If you're going to donate money on Internet Resource stone 646 00:46:00,239 --> 00:46:02,760 it to archive dot org. Yeah, I do. I give. It's 647 00:46:02,800 --> 00:46:05,840 not a lot, but I give them like twenty five a month. Wow. 648 00:46:06,000 --> 00:46:07,679 That's nice because I make such a huge use of this side. And 649 00:46:07,719 --> 00:46:12,920 then there's a few software projects like since I use OCR and I PDF pretty 650 00:46:12,920 --> 00:46:15,920 regularly, so sometimes I'll download something that I don't think was OCR very well 651 00:46:15,960 --> 00:46:21,400 there, and then I'll run OCR my PDF on it, which is great 652 00:46:21,480 --> 00:46:25,800 script just wonderful. It's got tests iact on the back end. It's Python 653 00:46:25,880 --> 00:46:30,639 based. You throw a PDF at it, tell it which languages the it 654 00:46:30,719 --> 00:46:35,960 should OCR with, and it just wonderful things like de skewing and some a 655 00:46:36,000 --> 00:46:42,400 little bit of contrast correction. Just fantastic, Just kind of a nice front 656 00:46:42,480 --> 00:46:45,119 end for tests direct to handle some of the real common cases. So it 657 00:46:45,119 --> 00:46:51,239 converts some PDF to PDF. Just fantastic, beautiful. Yeah. Why do 658 00:46:51,360 --> 00:46:52,519 new words emerge? I mean, I can think of a few reasons, 659 00:46:52,519 --> 00:46:55,440 but I like to hear your version of that. Well, in your business. 660 00:46:55,480 --> 00:46:58,639 You know why they emerge? In the tech business, You get a 661 00:46:58,639 --> 00:47:01,239 new product. It doesn't is a new problem, and that new problem has 662 00:47:01,239 --> 00:47:06,320 a name, right, It's often a silly name like zing boing or something. 663 00:47:06,480 --> 00:47:09,760 Well, what I'm talking about For example, if we oh, speaking 664 00:47:09,760 --> 00:47:14,519 of Elon Musk, what did he create upstairs in the sky? He created 665 00:47:14,760 --> 00:47:20,760 mum starlink starlink? What is starlink? When you see all those satellites together, 666 00:47:20,920 --> 00:47:23,400 it's a mega constellation. Right, So that's a new word because he 667 00:47:23,440 --> 00:47:29,360 created something that didn't exist before a mega constellation, and so you're you're creating 668 00:47:29,400 --> 00:47:32,599 new terms to describe new things. Yeah, often it is, and it's 669 00:47:32,599 --> 00:47:37,559 easier with nouns, and it's easier when you're not creating new senses for verbs, 670 00:47:37,559 --> 00:47:42,119 those are much harder, or new senses for words, those are much 671 00:47:42,119 --> 00:47:46,320 harder to come across and to notice. But but generally, it's about the 672 00:47:46,360 --> 00:47:52,119 advancement of ideas, culture, people. It's about things changing, and that's 673 00:47:52,119 --> 00:47:54,400 when the language has to change too. I learned so many great terms from 674 00:47:54,400 --> 00:47:59,880 listening to yours, like your show, like voluntold is a great one, 675 00:48:00,400 --> 00:48:02,880 has been volun told. Anybody who's worked in the military knows that way. 676 00:48:04,719 --> 00:48:08,480 That's a great word and it doesn't require explanation. You know. So many 677 00:48:08,519 --> 00:48:15,760 of our three letter acronyms TLAs are just so baffling to people who don't who 678 00:48:15,800 --> 00:48:21,559 aren't in the business, that we have to constantly stop and define what what 679 00:48:21,679 --> 00:48:23,480 we're talking about. Yeah, we put a show in the can this week 680 00:48:23,800 --> 00:48:29,000 that you'll hear in a couple weeks. Martha and I talked with a caller 681 00:48:29,039 --> 00:48:32,440 who was telling us about working with Aussie's people from Australia, and he was 682 00:48:32,480 --> 00:48:38,880 talking about how mysterious their language was at first. And so sometimes the language 683 00:48:38,960 --> 00:48:45,360 changes just because you have this natural friction with other groups of people, so 684 00:48:45,440 --> 00:48:50,039 you you pick up how they speak, you know, and it becomes part 685 00:48:50,039 --> 00:48:57,840 of your speech. So language change is about osmosis, you know, growth 686 00:48:57,920 --> 00:49:01,639 between different groups of people having to accommodate each other. Accommodation is a real 687 00:49:02,079 --> 00:49:07,880 big part of new language, an adoption of other people's speech or adoption of 688 00:49:07,920 --> 00:49:12,920 their ways of living. So, for example, on the border between Brazil 689 00:49:13,039 --> 00:49:16,480 and all the Spanish speaking countries around it, there are varieties of language no 690 00:49:16,559 --> 00:49:22,039 as Portugnol, which is a mix of Portuguese and Spanish, but they're different 691 00:49:22,119 --> 00:49:28,639 depending on which border you're at. They're each one of them has kind of 692 00:49:28,840 --> 00:49:31,760 created a new crayle, you know, just like the varieties of Spanglish and 693 00:49:31,920 --> 00:49:37,280 spoken along the Mexico US border are a little different. So Miami Spanglish is 694 00:49:37,320 --> 00:49:45,320 different from Texas Spanglish. Yeah, just because the English brushing up different kinds 695 00:49:45,320 --> 00:49:49,599 of Spanish. They've just they're varied when you're banging against different cultures. Like 696 00:49:49,639 --> 00:49:52,559 if you think about the northern countries in the northern side of Brazil, like 697 00:49:52,639 --> 00:49:57,880 Surinam and Guyana. They've got some strong French influences too, exactly. Yeah, 698 00:49:58,000 --> 00:50:00,760 it's the same reason that the cabe Quaff wrench is different from say, 699 00:50:01,000 --> 00:50:06,400 uh, I don't know Martinique French, and they're going to Caribbean for sure. 700 00:50:07,159 --> 00:50:08,880 The other thing I wanted to say is like in the tech world, 701 00:50:09,320 --> 00:50:15,559 so the jargon of say, people who work in cloud services versus people who 702 00:50:15,679 --> 00:50:20,840 do front frontline support. They're both in tech, but they're different flavors of 703 00:50:20,880 --> 00:50:22,760 tech, so they're going to have some overlap, but they're still gonna have 704 00:50:22,800 --> 00:50:25,800 different lingo. They might have different lingo for the same stuff. You know, 705 00:50:25,840 --> 00:50:29,880 how they talk about customers for example, or how they talk about service 706 00:50:29,920 --> 00:50:34,719 calls. Yeah. Yeah, I'm still not convinced that what's happening with the 707 00:50:34,800 --> 00:50:37,199 large language models is going well enough that it's going to continue. We may 708 00:50:37,599 --> 00:50:43,400 we may be thrown into a winter like where I think the whole Live with 709 00:50:43,559 --> 00:50:47,639 byng was a pr play. I think that the most important part of the 710 00:50:47,679 --> 00:50:51,880 case, that what stimulated all this was one hundred million users signing up for 711 00:50:51,960 --> 00:50:54,400 chat GBT. You think it's a money grab, Yeah, well, I 712 00:50:54,400 --> 00:50:57,760 think I always think it's a money grab. I think it's a pr play. 713 00:50:57,840 --> 00:51:01,079 I think byng was in the single digits of use as a search engine, 714 00:51:01,360 --> 00:51:06,679 and then chat gpt comes along and we now have evidence, you know, 715 00:51:06,719 --> 00:51:10,159 they've made posts of things talking about they had been tinkering with a language 716 00:51:10,199 --> 00:51:15,320 model attached to a search engine for a couple of years. They weren't happy 717 00:51:15,360 --> 00:51:17,719 with it. It wasn't good enough. And then chat gpt comes out, 718 00:51:17,840 --> 00:51:22,159 signs up one hundred million users in two months. They're like, that's more 719 00:51:22,199 --> 00:51:25,320 people that are using bing. Let's put it out there. And then Google 720 00:51:25,320 --> 00:51:29,000 freaks out and takes the bait. I wonder how long it's going to be 721 00:51:29,000 --> 00:51:36,679 before Amazon Alexa has that sort of contextual memory that it does not really currently 722 00:51:36,719 --> 00:51:39,039 have. Now, I don't know. Alexa has turned out to be a 723 00:51:39,119 --> 00:51:44,599 real money sync for Amazon. It's something like ten billion dollars put into it, 724 00:51:44,639 --> 00:51:46,800 and it looks like a failing portion of their business. Well all of 725 00:51:49,199 --> 00:51:52,920 those voice bots all the same time. I mean, it's coming into twenty 726 00:51:52,920 --> 00:51:55,199 twenty three where we have a threat of an economic downturn. I don't know 727 00:51:55,280 --> 00:51:59,119 we actually have a we have a threat of one. It's a very stylish 728 00:51:59,199 --> 00:52:02,440 or tech companies to lay people off and so forth. They're clearly scrutinizing their 729 00:52:02,480 --> 00:52:07,199 businesses and wanting and signally they're going to cut things that aren't working. And 730 00:52:07,360 --> 00:52:14,360 both Google Home and their voice device and Amazon Alexa both were supposed to make 731 00:52:14,440 --> 00:52:17,400 money and didn't to the tune of billions, and so they're saying they make 732 00:52:17,440 --> 00:52:21,679 it noisy. I don't know if they've actually done anything, but they've made 733 00:52:21,719 --> 00:52:24,360 noises about winding it down. Yeah, exactly what I've been reading, because 734 00:52:24,519 --> 00:52:29,800 from my perspective as somebody who hosts a radio show and podcasts, I'm always 735 00:52:29,840 --> 00:52:35,480 interested in how many listeners can tune into my show on those devices. And 736 00:52:35,519 --> 00:52:39,360 it's incredibly hard to make it happen, Yeah, because they're Unfortunately. My 737 00:52:39,440 --> 00:52:43,880 show is called Away with Words, and there are a zillion songs, and 738 00:52:44,039 --> 00:52:50,159 other podcasts either have episodes or portions of their programming called Away with Words. 739 00:52:50,320 --> 00:52:53,159 Yeah, and the wrong thing comes Dude, we have a podcast it starts 740 00:52:53,159 --> 00:53:00,519 with a period. Oh God. It's really hard that anyway, And it's 741 00:53:00,519 --> 00:53:05,639 surprised a number of people create new shows constantly called Away with Words without checking 742 00:53:05,800 --> 00:53:08,159 to see if there's another show already called that. People are just do not 743 00:53:08,280 --> 00:53:12,480 think yeah, yeah, it's not good for them either. I mean, 744 00:53:12,840 --> 00:53:15,199 so no, I find it. You have to be very explicit with the 745 00:53:15,239 --> 00:53:19,960 one who starts today. You have to say, play the latest episode of 746 00:53:19,960 --> 00:53:23,159 the podcast Away with Words with Grant Barrett and Martha Parnett, and then it 747 00:53:23,199 --> 00:53:27,840 will But yeah, people don't do No, they don't do that because speech 748 00:53:27,880 --> 00:53:30,719 is supposed to be easy. I would say it's supposed to be efficient. 749 00:53:30,840 --> 00:53:34,800 I'm supposed to be able to say very little and you know, and you 750 00:53:35,239 --> 00:53:37,559 or should understand what I'm talking about. How far do you think we are 751 00:53:37,719 --> 00:53:43,519 from the way it works in movies or Star Trek to speaking to the computer 752 00:53:43,559 --> 00:53:45,920 and having it all be understood. I think we can't even see our way 753 00:53:45,920 --> 00:53:52,960 there right now because we don't have real understanding coming from the machine of intent 754 00:53:52,159 --> 00:53:55,679 or any of those things. I think it's almost like we're in the same 755 00:53:55,719 --> 00:54:00,960 space that Tassel is with the autopilot. You've made something that feels like it 756 00:54:00,960 --> 00:54:05,079 should exist. It has science fiction context, so people want to believe exists, 757 00:54:05,159 --> 00:54:08,559 and so you ignore how very broken it actually is. Right, That's 758 00:54:08,559 --> 00:54:13,440 exactly you put it into. So the words that I was trying to say, 759 00:54:13,599 --> 00:54:15,920 that's exactly how I feel about it, And I wonder if it won't 760 00:54:15,920 --> 00:54:20,679 be thrown into a winter because it deserves to me. I like that ex 761 00:54:20,760 --> 00:54:23,400 question too. Thrown into a winter where it just everything chills for a while. 762 00:54:23,480 --> 00:54:25,880 Yeah, put it on ice. It's kind of like what happened with 763 00:54:25,920 --> 00:54:30,960 electric cars. Electric cars were big at the beginning of the automoile ears automobile 764 00:54:30,000 --> 00:54:34,559 era, and how long did they lay in the deepres? Yeah, for 765 00:54:35,079 --> 00:54:37,280 decades and then they you know, a little a little poke up in the 766 00:54:37,320 --> 00:54:40,880 eighties, in the seventies when with the oil crisis, and then they went 767 00:54:40,880 --> 00:54:44,800 away again, and a little poke up in the nineties, and then it 768 00:54:44,840 --> 00:54:47,679 went away again. People in the crazy billionaire showed up. People gonna yell 769 00:54:47,679 --> 00:54:51,079 at me for saying lay instead of lie. Just so you know, here 770 00:54:51,079 --> 00:54:57,320 you go. But I grant you know, it's contextual to us because of 771 00:54:57,360 --> 00:55:01,199 the artificial intelligence winters, right raise. Artificial intelligence first comes up the fifties 772 00:55:01,199 --> 00:55:06,760 with Martin Minsky working for the military. They get some stuff to work, 773 00:55:06,760 --> 00:55:10,599 but their promises are way out of scope and so winter. Then it reappears 774 00:55:10,639 --> 00:55:15,639 as the expert systems in the seventies and they build these thirty thousand dollar computers 775 00:55:15,639 --> 00:55:20,639 and sold several thousand of them. But that's about it. Winter then in 776 00:55:20,679 --> 00:55:22,840 the nineties with the decision tree systems and deep Blue and so forth, and 777 00:55:22,920 --> 00:55:28,519 playing chess, and it can't get further than that winter. And then it 778 00:55:28,559 --> 00:55:31,599 comes up again in the twenty eleven's with Jeffrey Hinton's model and an image net 779 00:55:31,599 --> 00:55:36,639 and vision recognition systems, and that's what we're the tail into this right now. 780 00:55:36,719 --> 00:55:39,679 Hinton's now saying, here comes another winter. But you know the big 781 00:55:39,760 --> 00:55:43,280 change now is we have the clouds. So with this huge amount of compute 782 00:55:43,320 --> 00:55:49,159 hanging around and companies highly motivated to find work for their cloud, clouds are 783 00:55:49,159 --> 00:55:52,639 full arranging the kicker to all this open AI stuff and jatchebetween all these things 784 00:55:52,800 --> 00:55:58,360 all consume cloud. I have a story for you, Grant. Do you 785 00:55:58,400 --> 00:56:04,719 remember a program back when we were all young called Eliza. Yes, I 786 00:56:04,760 --> 00:56:07,840 played with Eliza. I had the MAC version. Yeah, yeah, the 787 00:56:07,960 --> 00:56:13,320 Rogerian psychotherapist. You basically would ask it a question and it would answer with 788 00:56:13,400 --> 00:56:20,360 a question that contained parts or was triggered by your question. Yes. Yeah, 789 00:56:20,360 --> 00:56:22,480 you can still find it online. Yeah, I'll include a link to 790 00:56:22,519 --> 00:56:27,159 an online one, so you would say a sentence that included my brother, 791 00:56:27,480 --> 00:56:30,719 right, and so picked up on those keywords and said, tell me more 792 00:56:30,760 --> 00:56:34,519 about your family. And my first experience of this was, oh, my 793 00:56:34,639 --> 00:56:37,760 god, it's brilliant. It's a genius. But here's the thing. The 794 00:56:37,840 --> 00:56:43,960 guy who wrote that originally wrote it as a goof against the term artificial intelligence 795 00:56:43,960 --> 00:56:47,719 to prove that it wasn't intelligent. It was simply doing a lookup of keywords 796 00:56:47,800 --> 00:56:54,239 and responding with responses appropriate for those keywords. And everybody, kind of including 797 00:56:54,280 --> 00:56:59,119 me as a kid, I was blown away by this, and I actually 798 00:56:59,159 --> 00:57:02,360 rewrote it in visual Basic a long time ago, and I think I think 799 00:57:02,400 --> 00:57:05,840 I might have even done an online version of it. But but I was 800 00:57:05,880 --> 00:57:08,519 just fascinated by it. You know, from a consumer standpoint, it looked 801 00:57:08,559 --> 00:57:13,079 like magic to me. Yeah, people want to believe these things exist, 802 00:57:13,239 --> 00:57:22,880 right between peridolia and anthropomorphization, Like we're looking for faces and we cast humanity 803 00:57:22,920 --> 00:57:25,360 on things that aren't you. I'm not the first to say this, but 804 00:57:25,400 --> 00:57:30,440 we are terribly bad at pattern matching, and we just think we're fantastic. 805 00:57:30,519 --> 00:57:37,119 Yeah right, the pattern we matches. I'm a genius, but I am, 806 00:57:37,239 --> 00:57:39,400 Yeah, but in your case, we're making an exception. But no, 807 00:57:39,440 --> 00:57:43,880 I'm not. I always say that like I'm somewhere in the middle, 808 00:57:43,960 --> 00:57:46,840 just a little bit above average. So the well beg Yeah, and and 809 00:57:46,840 --> 00:57:51,559 the other thing we know about humans is when they don't understand something or there's 810 00:57:51,639 --> 00:57:54,639 ambiguity, we make stuff up. Yeah, and we're scared and we're scared. 811 00:57:54,840 --> 00:57:59,000 Yeah, yeah, we make stuff up and it always blames someone but 812 00:57:59,199 --> 00:58:02,320 us. Right, So go read a book or a bunch of books. 813 00:58:02,480 --> 00:58:06,800 Yeah, let me ask you guys a question while I have here. What 814 00:58:06,880 --> 00:58:07,960 do you what are you working on? What do you what do you do 815 00:58:08,079 --> 00:58:12,000 besides this podcast? Well I could tell you if well, first of all, 816 00:58:12,079 --> 00:58:15,800 Richard's working on a book about the history of dot net. Okay, 817 00:58:15,000 --> 00:58:19,159 Microsoft's you probably know what dot net is. Yeah, yeah, yeah, 818 00:58:19,239 --> 00:58:22,920 Well I only have the like open source version of it installed. But yeah, 819 00:58:22,079 --> 00:58:25,360 right, so it goes back to being a Windows only product back in 820 00:58:25,440 --> 00:58:31,199 the two thousands and went all the way through tablets and phones and open source. 821 00:58:32,360 --> 00:58:35,800 And he can tell you more about that. But I'm basically, um 822 00:58:36,119 --> 00:58:42,920 a web developer, and I've got a couple of YouTube videos series one on 823 00:58:43,280 --> 00:58:49,280 Blazer, which is there the state of the art web UI technology, and 824 00:58:49,400 --> 00:58:54,239 another one on Maui, which is their mobile technology. So i'm my job 825 00:58:54,400 --> 00:59:00,000 is essentially to come up with good examples of how to use this stuff every 826 00:58:59,840 --> 00:59:01,760 week, do some research, write some code, and then do a video 827 00:59:01,760 --> 00:59:06,360 about it. So i'm this is my wheelhouse. I love it. Oh 828 00:59:06,400 --> 00:59:09,119 that's fantastic. I'm working on a book too. It's on the on the 829 00:59:09,159 --> 00:59:15,320 words suck, s u c K, suck these there a whole book there. 830 00:59:15,639 --> 00:59:17,000 Yeah. Yeah, well it'll be a short one thirty thousand words, 831 00:59:17,079 --> 00:59:21,159 but yeah, a mirror and it will be on it'll be like on the 832 00:59:21,199 --> 00:59:22,920 figurative uses of it, like you suck or that sucks. No, no, 833 00:59:23,000 --> 00:59:28,039 we have a we have a version of it. That's um a noun. 834 00:59:28,880 --> 00:59:30,360 Right, So we went the big suck, right, Well, we 835 00:59:30,400 --> 00:59:35,079 turned down the suck and we turn up the awesome in our production, in 836 00:59:35,079 --> 00:59:37,320 our production that's yeah, yeah, exactly, Yeah, Yeah, we have 837 00:59:37,360 --> 00:59:39,800 the suck now. Yeah, I will be addressing that now, Yes, 838 00:59:39,920 --> 00:59:45,960 turn it down. I appreciate that. Grant. It's been an absolute pleasure 839 00:59:45,960 --> 00:59:50,480 to talk to you and keep in touch, will you Yeah, absolutely, 840 00:59:50,480 --> 00:59:52,000 thanks for having me on. This was a great conversation. It's nice to 841 00:59:52,559 --> 00:59:57,400 flex the tech brain a little bit. Yeah, and again, you know, 842 00:59:57,559 --> 01:00:00,599 it was a really good intersection of all these things and at the right 843 01:00:00,599 --> 01:00:02,719 time, I think. So thanks again and we'll see you on the way 844 01:00:02,760 --> 01:00:06,559 with words. All right, Thank you, take happy well you guys. 845 01:00:06,599 --> 01:00:08,400 All right, byebye. We'll talk to you, dear listener next time on 846 01:00:08,599 --> 01:00:34,719 dot net rocks. Dot net Rocks is brought to you by Franklin's Net and 847 01:00:34,880 --> 01:00:39,199 produced by Pop Studios, a full service audio, video and post production facility 848 01:00:39,480 --> 01:00:45,440 located physically in New London, Connecticut, and of course in the cloud online 849 01:00:45,440 --> 01:00:51,559 at pwop dot com. Visit our website at dt n et r ocks dot 850 01:00:51,599 --> 01:00:55,960 com for RSS feeds, downloads, mobile apps, comments, and access to 851 01:00:57,000 --> 01:01:00,679 the full archives going back to show number one, recorded in September two thousand 852 01:01:00,719 --> 01:01:05,159 and two. And make sure you check out our sponsors. They keep us 853 01:01:05,159 --> 01:01:08,599 in business. Now go write some code. See you next time. You 854 01:01:08,679 --> 01:01:20,920 got a check. Mile Band