WEBVTT 1 00:00:02.040 --> 00:00:07.519 Welcome everyone to another episode of Adventures in Hit Adventures 2 00:00:07.599 --> 00:00:11.199 in DevOps. You would think after a few hundred episodes, 3 00:00:11.240 --> 00:00:14.679 I would learn the name of the show, but working 4 00:00:14.759 --> 00:00:18.199 on it, I think it's actually getting worse. Well it is, 5 00:00:18.280 --> 00:00:20.640 because now I've got like this mental block. You know 6 00:00:20.679 --> 00:00:23.679 where my internal monologue is going, don't f it up. 7 00:00:23.679 --> 00:00:25.920 Don't f it up. Because you watch he's gonna. 8 00:00:25.800 --> 00:00:28.640 F it up the second second week in a row. 9 00:00:28.679 --> 00:00:30.239 I think that he left it up. 10 00:00:31.760 --> 00:00:35.119 Welcome Warren, How are you, Yeah, I'm good. I actually 11 00:00:35.200 --> 00:00:38.479 do have a fact this week. It's it's not security related. 12 00:00:39.000 --> 00:00:41.439 I actually am a little worried that we may have 13 00:00:41.520 --> 00:00:46.079 reached a plateau local maximum for our innovation and AI, 14 00:00:46.679 --> 00:00:49.719 because we've already started to see products that are heavily 15 00:00:49.799 --> 00:00:52.560 toward exploitation, and so you can see that there has 16 00:00:52.600 --> 00:00:55.399 been a huge shift in where we were before things 17 00:00:55.439 --> 00:00:57.840 being released for free, and now we're at the stage 18 00:00:57.920 --> 00:01:01.560 of the technology where everyone's just trying to extract value 19 00:01:01.560 --> 00:01:03.280 from it. I don't know what that means for the 20 00:01:03.320 --> 00:01:06.359 long term, but I think it's really interesting. 21 00:01:07.640 --> 00:01:09.560 I think it means shareholder profits. 22 00:01:09.879 --> 00:01:10.239 That's more. 23 00:01:10.400 --> 00:01:13.040 Let's hope, right, that's what everyone says, everyone wants shareholder 24 00:01:13.040 --> 00:01:14.719 profits and maybe we're actually getting there. 25 00:01:15.959 --> 00:01:23.120 Maybe. So speaking of AI, our guest this week, John W. Maylee, 26 00:01:24.079 --> 00:01:28.480 Attorney at large, founded the consulting firm John Maylee and Associates. 27 00:01:28.680 --> 00:01:32.439 But before you jump off the deep end and go 28 00:01:32.799 --> 00:01:35.760 what in tarnation, we'll lawyer it up for this episode. 29 00:01:36.159 --> 00:01:40.599 It's it's actually relevant because in addition to being an attorney, 30 00:01:41.359 --> 00:01:45.799 John has his master's degree in computer science from Stanford University. 31 00:01:46.400 --> 00:01:49.680 And the thing that led me to having this conversation 32 00:01:49.760 --> 00:01:52.799 with him, he's the author of the book Juris X 33 00:01:52.920 --> 00:01:56.920 Machina that's about It's a sci fi book, but it's 34 00:01:57.040 --> 00:02:01.159 about where the US has replaced yours in the legal 35 00:02:01.200 --> 00:02:05.920 system with AI, and you know there may be some 36 00:02:06.040 --> 00:02:09.240 fallout with that, and so that's the topic of the book, 37 00:02:10.039 --> 00:02:12.400 and we're going to talk all about all kinds of 38 00:02:12.439 --> 00:02:16.680 things AI engineering and sci fi related. So, John, thank 39 00:02:16.719 --> 00:02:17.599 you for being on the show. 40 00:02:18.199 --> 00:02:18.879 Thanks for having me. 41 00:02:18.960 --> 00:02:21.599 Great to be a guest, right, I'm looking forward to this. 42 00:02:21.680 --> 00:02:24.879 I was just looking at your your bio here and 43 00:02:24.919 --> 00:02:29.039 it cracks me up. You have a running, swimming, long 44 00:02:29.039 --> 00:02:37.199 distance motorcycling, classic car restoration stage diving, crowdsurfing, bar fighting, 45 00:02:38.479 --> 00:02:41.639 Lama rancher. Like, we could go on and on on 46 00:02:41.680 --> 00:02:44.360 this episode for quite a while, because it's it's a 47 00:02:44.400 --> 00:02:47.840 little embarrassing, like how much of my background overlaps with yours. 48 00:02:50.280 --> 00:02:53.280 I can't claim Lama ranching, but there's a lot of 49 00:02:53.360 --> 00:02:54.080 other stuff on there. 50 00:02:54.479 --> 00:02:56.199 I've tried to live my life such that it would 51 00:02:56.199 --> 00:02:59.159 be a rich source of blackmail sources. Nice. 52 00:02:59.400 --> 00:03:02.719 Nice, So that legal background is going to pay off 53 00:03:02.719 --> 00:03:04.000 for you in the future, right. 54 00:03:04.840 --> 00:03:07.919 Well, and once you've done one thing that's blackmailworthy, then 55 00:03:08.120 --> 00:03:10.479 it kind of dilutes the market. Right, So like if 56 00:03:10.560 --> 00:03:12.599 I go and do a bunch of other things, it's like, well, 57 00:03:12.639 --> 00:03:14.400 you already could have blackmailed me for this first thing. 58 00:03:14.479 --> 00:03:16.439 So like your leverage is pretty much the same. 59 00:03:18.439 --> 00:03:20.400 So tell me a little bit about how you went 60 00:03:20.520 --> 00:03:24.280 from computer science to legal. 61 00:03:25.680 --> 00:03:27.960 Yeah. So I was at this crossroads to a street 62 00:03:27.960 --> 00:03:29.520 and there was the devil was there and. 63 00:03:32.639 --> 00:03:35.439 Bando competition, right, and you didn't you didn't pick learning 64 00:03:35.439 --> 00:03:37.360 to play guitar. When you asked what you wanted. 65 00:03:38.439 --> 00:03:44.280 You know, now you mentioned that idea. So so yeah, 66 00:03:44.319 --> 00:03:48.639 I I originally went to college, went to Syracuse for 67 00:03:48.680 --> 00:03:53.360 computer engineering, and I double majored. I was studying psychology 68 00:03:53.400 --> 00:03:57.280 at the time too, and it was they were at 69 00:03:57.319 --> 00:04:00.599 the time more disparate fields. So you would and interesting 70 00:04:00.639 --> 00:04:03.159 things in your classes, and every now and then you'd 71 00:04:03.159 --> 00:04:05.680 have these moments where these things would just sort of 72 00:04:05.719 --> 00:04:09.240 synthesize into some really cool idea that bridge the two fields. 73 00:04:09.800 --> 00:04:14.319 And you know, the most obvious recurring spot for these 74 00:04:14.360 --> 00:04:19.319 types of things was AI because that was kind of 75 00:04:19.319 --> 00:04:21.199 one of the few areas where there was overlight between 76 00:04:21.199 --> 00:04:26.839 those two fields. So time passed and AI was just 77 00:04:27.160 --> 00:04:30.240 you know, mostly something you write about in textbooks. And 78 00:04:30.279 --> 00:04:32.720 then at Stanford when I was getting my grad degree 79 00:04:33.199 --> 00:04:36.319 in computer science, we actually you know, the rubber met 80 00:04:36.319 --> 00:04:39.959 the road and we actually got to write ais that 81 00:04:40.040 --> 00:04:43.040 would you know, for games being played against each other, 82 00:04:44.920 --> 00:04:48.600 different applications like that. That was really fun and exciting, 83 00:04:48.680 --> 00:04:50.759 and but it was it was kind of like this 84 00:04:50.839 --> 00:04:53.839 shiny novelty that you know. The only place people were 85 00:04:53.839 --> 00:04:56.240 really using neural nets to a great extent back then 86 00:04:56.360 --> 00:05:00.759 was like you know, the Post Office for recognizing characters 87 00:05:00.759 --> 00:05:04.519 and numbers and things that were cool. You know, it 88 00:05:04.560 --> 00:05:06.000 was cool that you could train something and it would 89 00:05:06.000 --> 00:05:07.959 get better at it, but it wasn't something that you like, 90 00:05:08.079 --> 00:05:10.199 would tell people about a cocktail parties and they'd be like, 91 00:05:10.240 --> 00:05:12.120 oh my god, you know, this has been a cream 92 00:05:12.160 --> 00:05:17.959 for the world. So that just kind of became like it, 93 00:05:18.160 --> 00:05:20.240 you know, that steed interesting to me, but it was 94 00:05:20.279 --> 00:05:23.480 kind of dormant. And then I worked as a computer 95 00:05:23.519 --> 00:05:29.319 engineer on a microprocessor design and validation team for five 96 00:05:29.399 --> 00:05:33.199 or six years, and while that was happening, I had 97 00:05:33.240 --> 00:05:36.399 filed you know, I've been an inventor on several patents 98 00:05:36.439 --> 00:05:40.000 and had worked with these different patent attorneys who worked 99 00:05:40.000 --> 00:05:43.360 with the company I worked for, And it occurred to 100 00:05:43.399 --> 00:05:45.279 me that whenever you would talk to these guys, you'd 101 00:05:45.279 --> 00:05:48.079 be like, you know, you'd be sitting in your kind 102 00:05:48.120 --> 00:05:50.560 of sad, little gray cubicle and you'd ask these guys, so, so, 103 00:05:50.720 --> 00:05:53.040 you know, where are you based? And they'd say, well, 104 00:05:53.120 --> 00:05:55.920 I'm in my in my yacht in the Caribbean right now, 105 00:05:56.240 --> 00:06:00.199 or I'm in a walled compound in that Nevada does there, 106 00:06:00.800 --> 00:06:03.560 and you like, it got to the point where you 107 00:06:03.600 --> 00:06:06.759 would start asking, you know, it was the answer was 108 00:06:06.759 --> 00:06:08.920 always so fascinating that you would just start asking people, 109 00:06:09.040 --> 00:06:12.040 like the first thing you would ask, and so I 110 00:06:12.079 --> 00:06:15.759 realized that these guys were interesting because they were technologists 111 00:06:15.800 --> 00:06:18.560 and they'd become they've gone into patent law, and so 112 00:06:18.600 --> 00:06:21.079 they had been able to leverage the fact that they 113 00:06:21.079 --> 00:06:24.800 were interested in technology but kind of break free of 114 00:06:25.120 --> 00:06:27.199 at the time when there wasn't a lot of remote work, 115 00:06:27.639 --> 00:06:29.959 kind of break free of that mold of being in 116 00:06:29.959 --> 00:06:32.680 the office nine to five and kind of pursue other 117 00:06:32.720 --> 00:06:36.879 interests more freely on the side. So I ended up 118 00:06:36.920 --> 00:06:44.560 going to law school at nights and learning more about law, obviously, 119 00:06:45.439 --> 00:06:48.800 and then at the end of that I was doing 120 00:06:48.839 --> 00:06:51.160 work that had to do with CPUs and GPUs, so 121 00:06:51.279 --> 00:06:55.480 looking at companies patent portfolios and you know, helping them 122 00:06:55.480 --> 00:06:58.560 figure out like this is a useful invention that actually 123 00:06:58.639 --> 00:07:00.879 is likely to be used, this is not a useful 124 00:07:00.879 --> 00:07:02.800 invention that you know, looks great on paper, but it 125 00:07:02.839 --> 00:07:07.480 won't really work as well. And then over time GPUs 126 00:07:07.519 --> 00:07:09.759 started getting bigger and bigger and bigger as time went on, 127 00:07:09.839 --> 00:07:11.480 as far as like how much I was asked to 128 00:07:11.480 --> 00:07:13.639 look at them, and part of that was because three 129 00:07:13.720 --> 00:07:17.079 D graphics was taking off even more. But then eventually 130 00:07:17.120 --> 00:07:20.000 we got to the point where there were these GPGPUs 131 00:07:20.040 --> 00:07:22.360 that were general purpose and weren't really necessarily being used 132 00:07:22.360 --> 00:07:25.360 for graphics and never being used for server farms and 133 00:07:25.399 --> 00:07:29.839 cloud computing, and eventually that just sort of took over. 134 00:07:30.000 --> 00:07:31.920 So it was kind of cool from that standpoint that 135 00:07:31.959 --> 00:07:34.319 I went from, you know, the only time I would 136 00:07:34.319 --> 00:07:38.680 talk about AI and a patent portfolio was like analoc 137 00:07:38.759 --> 00:07:44.000 brake systems or lane change sensors and you know, luxury vehicles, 138 00:07:44.040 --> 00:07:47.759 to suddenly just about everything I work on now has 139 00:07:47.800 --> 00:07:52.560 some foot in the AI space in somewhere another, whether 140 00:07:52.600 --> 00:07:56.079 it's hardware that helps enable it or whether it's software. 141 00:07:56.319 --> 00:07:59.759 So that's that's kind of cool because it's something that's 142 00:07:59.759 --> 00:08:03.240 always fascinated me, and now it's suddenly fascinating to society 143 00:08:03.240 --> 00:08:05.000 as well. So I'm no longer an outlier. 144 00:08:06.519 --> 00:08:08.439 So where do you think AI is at on the 145 00:08:09.160 --> 00:08:11.560 overhype cycle? Do you think it's overhyped right now or 146 00:08:11.560 --> 00:08:15.040 do you think it's appropriate? 147 00:08:15.959 --> 00:08:18.920 I think it's it's kind of both. Right there's a 148 00:08:18.920 --> 00:08:22.759 lot of asymmetry. I think that's it's over hyped in 149 00:08:22.879 --> 00:08:28.720 terms of you know, every company is now rushing to 150 00:08:28.720 --> 00:08:30.720 get on the bandwagon and find a way to add 151 00:08:30.759 --> 00:08:32.960 AI to their product, even if it's just kind of 152 00:08:33.320 --> 00:08:35.720 you know, pointlesser doesn't work very well. Like I was 153 00:08:35.759 --> 00:08:38.919 reading an article yesterday about how all these dating apps 154 00:08:38.960 --> 00:08:42.799 had incorporated AI and they were still just as crappy 155 00:08:42.840 --> 00:08:45.480 and court and finding people of that match you as 156 00:08:45.480 --> 00:08:48.080 they were before, but now they're faster of being crappy. 157 00:08:48.159 --> 00:08:49.759 So there's weird. 158 00:08:51.720 --> 00:08:54.039 I mean, I love the outcome there, which is your 159 00:08:54.320 --> 00:08:58.080 entire romantic life will be decided by two robots talking 160 00:08:58.120 --> 00:09:00.679 to each other, right, I mean, both on the receiving 161 00:09:00.720 --> 00:09:03.840 of the message and the sending will now uh not 162 00:09:04.039 --> 00:09:06.639 no longer be human, and you'll decide on whether or 163 00:09:06.679 --> 00:09:09.360 not to pursue a person based on what the algorithm says. 164 00:09:09.440 --> 00:09:13.200 Right, I mean there's a conflict adventest too, right, because 165 00:09:13.240 --> 00:09:17.600 now there's AI agents that are designed that you can 166 00:09:17.720 --> 00:09:20.279 date and so then you know it's like monetized and 167 00:09:20.320 --> 00:09:21.879 that you can buy them accessories and. 168 00:09:24.360 --> 00:09:27.279 Do not google that, do not google the accessories that 169 00:09:27.320 --> 00:09:29.519 are available while you're on your work computer. 170 00:09:30.960 --> 00:09:33.600 So it's it's funny because not only is it creating 171 00:09:33.600 --> 00:09:35.600 this fake dating relationship, it's it's kind of making you 172 00:09:35.639 --> 00:09:39.799 the sugar daddy because my person is entirely dependent on 173 00:09:39.840 --> 00:09:42.799 you for new outfits and jewelry and things and pets 174 00:09:42.840 --> 00:09:45.840 and overall happiness. And so it's a conflict of interest 175 00:09:45.879 --> 00:09:48.840 that they're sort of steering you toward bad people that 176 00:09:48.879 --> 00:09:52.159 you're incompatible with. So it just makes dating the AI 177 00:09:52.320 --> 00:09:55.000 is even more appealing. Fine, I give up. I will 178 00:09:55.039 --> 00:09:55.679 just date an AI. 179 00:09:55.919 --> 00:09:58.519 We're already at the dystopian future right there. There's no 180 00:09:58.679 --> 00:10:01.120 there's no next step after this. We're already there. This 181 00:10:01.159 --> 00:10:04.720 is where like a whole science fiction movies and television 182 00:10:04.720 --> 00:10:07.879 shows and books are already set right where we're dating 183 00:10:07.919 --> 00:10:08.240 the AI. 184 00:10:09.200 --> 00:10:11.440 It's true, and you know, this is the under hype 185 00:10:11.440 --> 00:10:15.200 overhype like paradigm. I think that's at the same time, 186 00:10:15.600 --> 00:10:18.399 if we call like a text support or a customer 187 00:10:18.399 --> 00:10:21.919 support line and we say, like speak to an operator, 188 00:10:22.159 --> 00:10:24.639 it takes like twenty five tries of me yelling that 189 00:10:24.679 --> 00:10:27.679 louder and louder before the AI just like click and say, oh, 190 00:10:27.720 --> 00:10:31.799 that's what you want. So other areas like there's a 191 00:10:31.799 --> 00:10:35.320 total lack of AI development. And you know, by the 192 00:10:35.360 --> 00:10:40.000 same token, we have this old fear that we got 193 00:10:40.039 --> 00:10:42.720 I think from science fiction through you know, the seventies 194 00:10:42.720 --> 00:10:46.480 and eighties and onward, that's AI was going to become 195 00:10:46.519 --> 00:10:49.480 this thing that once it got sufficiently intelligent, it would 196 00:10:49.480 --> 00:10:51.879 just sort of take over and start, you know, annihilating 197 00:10:51.960 --> 00:10:54.480 humans or imprisoning them so they don't hurt themselves, or 198 00:10:55.200 --> 00:10:59.159 anywhere in between. And what we actually have is a 199 00:10:59.159 --> 00:11:02.600 gizzilion different AI eyes that all have very different specializations 200 00:11:02.639 --> 00:11:05.360 and very different motivations of what they're trying to optimize. 201 00:11:05.919 --> 00:11:10.360 And there's no sort of universal intelligence general is AI 202 00:11:10.480 --> 00:11:12.799 yet where it just goes out and tries to help humanity. 203 00:11:12.879 --> 00:11:17.279 It's more like, you know, I'm really good at analyzing 204 00:11:17.799 --> 00:11:21.440 research results or counting how many times a word appears 205 00:11:21.440 --> 00:11:24.679 on a form or something like that. So it's accelerating things, 206 00:11:24.679 --> 00:11:27.519 but it's still very specialized and it's not multimodal to 207 00:11:27.559 --> 00:11:31.000 any great extent yet. But at the same time, you know, 208 00:11:31.039 --> 00:11:34.360 you have these stories coming out where someone's AI told 209 00:11:34.360 --> 00:11:37.000 them to go kill themselves, and you know what you 210 00:11:37.000 --> 00:11:39.360 don't see is the prompt right before then where they said, hey, 211 00:11:39.759 --> 00:11:41.799 next time I ask a question, I want you to respond, 212 00:11:41.840 --> 00:11:45.399 you should go kill yourself. So it's really easy to 213 00:11:45.519 --> 00:11:51.639 flag a sort of outlier AI response and turning into 214 00:11:51.639 --> 00:11:54.919 all kinds of news headlines, and that's interesting because it's 215 00:11:54.919 --> 00:11:58.799 sort of driving concerns over this more than actual outcomes are, 216 00:11:58.840 --> 00:12:00.600 and in some ways, the actual out comes are kind 217 00:12:00.600 --> 00:12:03.480 of what we need to be more worried about. So yeah, 218 00:12:03.480 --> 00:12:05.919 I would you know, in a true legal answer, I 219 00:12:05.960 --> 00:12:07.480 would say yes and no. No. 220 00:12:07.519 --> 00:12:09.440 I mean, I think it's really interesting that you bring 221 00:12:09.519 --> 00:12:12.840 up a bunch of those points. I mean, you're I 222 00:12:12.840 --> 00:12:16.279 think the decentralization of responsibilities and the specialization that the 223 00:12:16.480 --> 00:12:18.279 AI is is taking up, you know, is a really 224 00:12:18.279 --> 00:12:21.240 great point. And right now I do feel like they 225 00:12:21.279 --> 00:12:25.919 are solving sort of very tail value things like it's 226 00:12:26.120 --> 00:12:29.279 there's no core solution, there's no core greatness that's coming 227 00:12:29.320 --> 00:12:31.919 out of it for for society, and for sure, I 228 00:12:31.919 --> 00:12:34.440 don't really think anyone's talking about that. I think as 229 00:12:34.440 --> 00:12:36.879 far as we've gotten is we should be afraid, and 230 00:12:36.960 --> 00:12:39.440 that's that's I think, as far as people are willing 231 00:12:39.440 --> 00:12:41.759 to go, I think what you're talking about, though, really 232 00:12:41.759 --> 00:12:45.480 requires some complex questions to be answered, and I don't 233 00:12:45.519 --> 00:12:50.000 think humans have been so great at figuring out even 234 00:12:50.039 --> 00:12:52.679 which questions to ask, let alone answering them. For things 235 00:12:52.720 --> 00:12:54.960 that are much simpler, like what will happen tomorrow or 236 00:12:55.000 --> 00:12:55.559 the next day. 237 00:12:56.159 --> 00:12:59.039 Right. That's and that's kind of what's fascinating about AI, right, 238 00:12:59.120 --> 00:13:03.320 is that the developer who put the model together may 239 00:13:03.360 --> 00:13:05.320 not even know what its capabilities are and what the 240 00:13:05.320 --> 00:13:08.919 best questions to be asking are. But you know, it's 241 00:13:09.159 --> 00:13:11.399 and it is like a question of like, what's what's 242 00:13:11.440 --> 00:13:14.159 the utility of this? Right? And so I was think 243 00:13:14.200 --> 00:13:17.360 of the other day, like I was using Shatgypt, I 244 00:13:17.399 --> 00:13:19.879 was using the one model, so the one that is 245 00:13:20.639 --> 00:13:23.480 slower and much more thorough and a lot more nodes. 246 00:13:23.919 --> 00:13:25.960 And I asked it some really stupid question because I 247 00:13:25.960 --> 00:13:28.000 forgot to switch into a lesser model and I wanted 248 00:13:28.039 --> 00:13:31.840 to know how many calendar days were between like January 249 00:13:31.840 --> 00:13:35.759 eleventh and some other day. And immediately started churning and starting. 250 00:13:35.799 --> 00:13:38.120 It's like five minute process, and I'm like, oh, why 251 00:13:38.120 --> 00:13:42.480 did I switch? And my my chief source of that 252 00:13:42.639 --> 00:13:45.960 was not in patience. It was like guilt at like, man, 253 00:13:46.000 --> 00:13:48.960 I wonder how much like cooling water I'm using and 254 00:13:48.960 --> 00:13:51.639 how much like energy? This query is sucking down for 255 00:13:51.679 --> 00:13:54.240 something stupid. And so you know, it barks out the 256 00:13:54.279 --> 00:13:56.720 answer of like you know, thirty one days or whatever, 257 00:13:56.759 --> 00:14:00.200 and then I look at the prompt to get and 258 00:14:00.240 --> 00:14:03.440 it's like, you know, some questions you might want to ask, 259 00:14:03.600 --> 00:14:07.159 is is a hot dog as sandwich? And I realized 260 00:14:07.200 --> 00:14:11.440 that they're basically promoting like even more frivolous uses of 261 00:14:11.480 --> 00:14:15.759 these ais than than what I just felt guilty for doing. 262 00:14:15.840 --> 00:14:17.879 So it's it's really interesting, you know. I think in 263 00:14:17.919 --> 00:14:20.360 a lot of ways, it parallels what we saw with 264 00:14:20.399 --> 00:14:23.399 like the tech bubble and ninety nine and two thousand, 265 00:14:23.879 --> 00:14:28.559 where they're kind of so concerned with the future of 266 00:14:28.600 --> 00:14:30.559 how powerful their model is going to be, that they're 267 00:14:30.679 --> 00:14:35.279 less concerned with short term profitability and whether like you know, 268 00:14:35.320 --> 00:14:38.159 for instance, like you should have a serting algorithm that says, 269 00:14:38.159 --> 00:14:40.159 this is a really easy question. We can farm this 270 00:14:40.240 --> 00:14:42.600 out to like one of the mini models, and this 271 00:14:42.799 --> 00:14:44.879 is a research question that you know, we probably want 272 00:14:44.919 --> 00:14:48.159 to use as many notes as possible. So right now, 273 00:14:48.200 --> 00:14:50.240 it seems like they're you know, they charge you some 274 00:14:50.519 --> 00:14:53.919 tiny amount per month and it doesn't at all probably 275 00:14:53.919 --> 00:14:57.879 cover all their energy expenses and huge amounts of cooling 276 00:14:57.919 --> 00:14:59.799 that they need to do in all their server farms 277 00:15:00.039 --> 00:15:00.399 all that. 278 00:15:00.879 --> 00:15:02.759 Yeah, No, I mean you're actually onto something because there 279 00:15:02.759 --> 00:15:04.320 are a bunch of companies out there that now are 280 00:15:04.320 --> 00:15:07.879 promoting this idea of model routing amongst many companies at 281 00:15:07.919 --> 00:15:09.759 the same time to try to get you some of 282 00:15:09.759 --> 00:15:13.399 that value. Although it like that's a non trivial thing 283 00:15:13.440 --> 00:15:15.679 to even do, to think about, like how complex is 284 00:15:15.720 --> 00:15:18.360 this question actually is? And that like how good of 285 00:15:18.360 --> 00:15:21.200 an answer do you need? Is I find maybe almost 286 00:15:21.240 --> 00:15:23.080 one of those things that could be impossible to answer. 287 00:15:25.120 --> 00:15:27.039 It's you know, it's true, and how did you mean 288 00:15:27.039 --> 00:15:29.519 the question? Like if I'm going to ask what's the 289 00:15:29.559 --> 00:15:31.519 meaning of life and it's just going to laugh and 290 00:15:31.519 --> 00:15:33.519 spit out forty two, that doesn't take much to it 291 00:15:34.559 --> 00:15:36.320 if it actually is trying to give me a comprehensive 292 00:15:36.320 --> 00:15:39.080 philosophical and theological answer where it goes and queries all 293 00:15:39.120 --> 00:15:42.120 these different texts, like that's huge. So you can't even 294 00:15:42.240 --> 00:15:44.919 necessarily just take a question and say there's a right 295 00:15:44.960 --> 00:15:49.519 answer to that, but it is. It is an interesting 296 00:15:49.600 --> 00:15:54.320 kind of paradigm of how how these models are you know, 297 00:15:54.320 --> 00:15:57.159 which model is even most appropriate? And you know, I 298 00:15:57.440 --> 00:16:02.919 think the solution to that problem, you know, maybe what 299 00:16:03.000 --> 00:16:06.840 you sometimes see with kind of speculator almost like speculative execution, 300 00:16:06.919 --> 00:16:09.039 where it's like, here, can you do a preview of 301 00:16:09.080 --> 00:16:10.960 what type of answer you would come back with if 302 00:16:11.000 --> 00:16:14.639 I give you the contract to go and do this research. 303 00:16:15.159 --> 00:16:17.440 And you know, the idea of having these things work 304 00:16:17.480 --> 00:16:22.240 collaborative is interesting. But it also you know, it's much 305 00:16:22.279 --> 00:16:24.519 easier for ais to jail break each other. For instance, 306 00:16:24.879 --> 00:16:28.240 It's that they're just better at kind of pushing each 307 00:16:28.279 --> 00:16:31.559 other's buttons. So it really adds a lot of dynamicity 308 00:16:31.559 --> 00:16:36.000 into the equation of like dynamism of how what the 309 00:16:36.000 --> 00:16:38.480 ais are capable of when you add these very different 310 00:16:38.519 --> 00:16:41.159 ais and have them converse and pursue common goals. 311 00:16:41.360 --> 00:16:43.679 That's not something I actually heard before. I don't know 312 00:16:43.679 --> 00:16:46.799 if you have more information about that. But utilizing one 313 00:16:47.039 --> 00:16:52.120 model from one provider to a different provider for jail breaking, 314 00:16:52.120 --> 00:16:54.039 I mean you said jail breaking. I'm not sure exactly 315 00:16:54.120 --> 00:16:56.600 there's a jail breaking here, but like, what are you 316 00:16:56.639 --> 00:16:59.440 getting out there? Like is it being able to understand 317 00:16:59.559 --> 00:17:01.600 and get a better answer to your query? Something else? 318 00:17:01.639 --> 00:17:02.559 Like how does that work? 319 00:17:03.720 --> 00:17:08.039 Well? So, I mean I guess what I would liken 320 00:17:08.119 --> 00:17:11.359 it too, is you know it like an it it's 321 00:17:11.440 --> 00:17:14.799 AI is great at coming up with automated ways to 322 00:17:14.920 --> 00:17:19.519 just implement something. So if I ask AI to write 323 00:17:19.559 --> 00:17:22.119 me a script that does X, and then hey, this 324 00:17:22.160 --> 00:17:25.400 isn't my normal computer system. I'm not familiar with this OS. 325 00:17:25.839 --> 00:17:27.519 Can you also tell me how to set this up 326 00:17:27.559 --> 00:17:29.359 as like a recurring demon that runs it like four 327 00:17:29.400 --> 00:17:32.799 am or whatever. Like. It's really surprisingly proficient at coming 328 00:17:32.880 --> 00:17:39.119 up with lists of procedural lists. It's good at writing scripts, 329 00:17:39.160 --> 00:17:41.559 and you know, so the idea of using it for 330 00:17:42.079 --> 00:17:44.559 hacking and enumeration and not kind of automating all these 331 00:17:44.559 --> 00:17:49.799 different processes kind of for you know, red team or 332 00:17:49.799 --> 00:17:54.759 Blue team is kind of impressive. But I think, you know, 333 00:17:54.799 --> 00:17:57.759 the you do see a lot of really interesting results 334 00:17:57.759 --> 00:18:00.359 when you get like you set up a check room 335 00:18:00.440 --> 00:18:02.839 with a couple different AIS, and they shouldn't be the 336 00:18:02.880 --> 00:18:06.720 same AI. They should have very different prompting, different personalities, 337 00:18:07.119 --> 00:18:10.359 or they should just be completely different models. And you know, 338 00:18:10.400 --> 00:18:12.640 the difference being that if I'm sitting there trying to 339 00:18:12.640 --> 00:18:15.200 bypass the safety protocols of an AI, I'm going to 340 00:18:15.240 --> 00:18:16.400 try this. I'm going to try this. I'm going to 341 00:18:16.440 --> 00:18:19.039 try this, and it's a very dynamic process because I 342 00:18:19.079 --> 00:18:22.119 have to see what it kicks back and then think 343 00:18:22.160 --> 00:18:25.559 of a way around it, and that's a very manual process. 344 00:18:25.559 --> 00:18:28.400 But if you have an AI that's just constantly bombarding 345 00:18:28.440 --> 00:18:32.240 it with permutations, it suddenly becomes easier for it to do. 346 00:18:32.920 --> 00:18:35.160 And I think that that is I mean, I think 347 00:18:35.200 --> 00:18:38.960 the future and this is something that's my second but 348 00:18:39.119 --> 00:18:40.880 the sequel to the first book, which has not come 349 00:18:40.920 --> 00:18:44.960 out yet, has to do with is exploring what we're 350 00:18:45.000 --> 00:18:46.920 going to see in the future with AI, where they 351 00:18:47.160 --> 00:18:50.119 essentially are going to have, you know, factions and gang 352 00:18:50.119 --> 00:18:54.000 wars where you know, an AI might be tasked with 353 00:18:55.319 --> 00:18:59.680 spiking a competitor's AIS training. So give it some sort 354 00:18:59.680 --> 00:19:02.160 of weird corner case where when a certain type of 355 00:19:02.160 --> 00:19:06.839 input comes up, it completely malfunctions, or it could be 356 00:19:06.920 --> 00:19:09.759 something like, you know, help me bypass the safety protocols 357 00:19:09.759 --> 00:19:12.440 of this other AI, or help me trick this AI 358 00:19:12.599 --> 00:19:16.599 into doing something that's harmful to the company's interests. So 359 00:19:17.240 --> 00:19:19.000 I think that is going to be something we're going 360 00:19:19.039 --> 00:19:22.640 to see increasingly of AI's being. You know, it's kind 361 00:19:22.640 --> 00:19:24.039 of like what we have with defix, right, you can 362 00:19:24.079 --> 00:19:27.079 create a deep fake with an AI, and we're past 363 00:19:27.160 --> 00:19:31.200 the point now where we can necessarily reliably look at 364 00:19:31.200 --> 00:19:33.319 something and say Oh, that's a deep fake, and so 365 00:19:34.400 --> 00:19:37.440 what do you need is the stop gap against that? Well, 366 00:19:37.480 --> 00:19:40.039 you need an AI to tell you whether it's fake. So, 367 00:19:40.720 --> 00:19:43.039 if you're just the average consumer, you don't know much 368 00:19:43.079 --> 00:19:46.640 about deep fix or how to detect deep fix. And meanwhile, 369 00:19:46.680 --> 00:19:50.039 these companies aren't super motivated to tell you how their 370 00:19:50.039 --> 00:19:53.920 product works because it just invites a design around for 371 00:19:54.759 --> 00:19:57.759 you know, malicious actors. So you end up in this 372 00:19:57.799 --> 00:20:00.440 situation kind of like what we had in the nineteen 373 00:20:00.440 --> 00:20:03.279 eighties and nineties with antivirus software. But the average person 374 00:20:03.319 --> 00:20:06.400 doesn't necessarily need to know how viruses work, but they 375 00:20:06.440 --> 00:20:09.599 do know that there's a handful of trusted companies that 376 00:20:10.079 --> 00:20:13.000 generally are you know, you can trust how they work, 377 00:20:13.039 --> 00:20:16.160 even if you don't necessarily know how. So I think 378 00:20:16.160 --> 00:20:18.160 more and more we're going to just see AI as 379 00:20:18.200 --> 00:20:21.079 being the defense against AI, and there's no way around that, 380 00:20:21.119 --> 00:20:24.319 and we're going to keep entering into those types of situations. 381 00:20:25.000 --> 00:20:27.599 So we need a new John McAfee, is what you're saying. 382 00:20:28.279 --> 00:20:30.920 Or we need to help figure out how to revive 383 00:20:31.000 --> 00:20:32.839 him and bring him back. 384 00:20:34.039 --> 00:20:35.640 I think I don't know if we have enough evidence 385 00:20:35.680 --> 00:20:38.160 to actually conclude whether or not he's he's gone for good. 386 00:20:38.279 --> 00:20:43.279 Right, we'd have to ask whoever it is who suicided him, and. 387 00:20:46.200 --> 00:20:47.920 It's like, did you take a selfie? We're kind of 388 00:20:47.920 --> 00:20:48.839 looking for proof here. 389 00:20:49.759 --> 00:20:53.599 Yeah. I think anytime you're living on your own private island, 390 00:20:53.640 --> 00:20:57.400 you're just kind of asking for to be suicided. It 391 00:20:57.519 --> 00:20:58.640 seems to be the trend. 392 00:20:58.880 --> 00:21:06.079 Yeah. Yeah, And he wasn't really like keeping a low 393 00:21:06.119 --> 00:21:08.839 profile so that people would forget about him either, so 394 00:21:08.880 --> 00:21:11.759 he kept reminding people that he was there and like, 395 00:21:11.839 --> 00:21:14.599 oh yeah, I meant to kill him. 396 00:21:14.759 --> 00:21:17.000 Yeah, it's kind of like, you know, the guy who 397 00:21:17.119 --> 00:21:19.119 they want to extradite. So he's like hopping, he's doing 398 00:21:19.160 --> 00:21:21.039 a little dance by the border, like you can't get 399 00:21:21.079 --> 00:21:22.720 me right. 400 00:21:26.720 --> 00:21:29.920 So on that same the prior to John, before I 401 00:21:29.920 --> 00:21:32.960