1 00:00:00,020 --> 00:00:01,160 In 1968, 2 00:00:02,140 --> 00:00:03,720 Stanley Kubrick gave us 2001: 3 00:00:04,190 --> 00:00:04,940 A Space Audyssey. 4 00:00:05,840 --> 00:00:07,040 Monoliths, HAL, 5 00:00:07,200 --> 00:00:10,280 existential dread, and apparently an iPad. 6 00:00:10,660 --> 00:00:11,260 Seriously, 7 00:00:11,580 --> 00:00:15,660 there's a scene where astronauts are casually reading the news on these flat little rectangular screens. 8 00:00:16,059 --> 00:00:17,840 Totally fictional movie props. 9 00:00:18,140 --> 00:00:20,040 Fast forward about 40 years, 10 00:00:20,530 --> 00:00:23,280 and Apple is suing Samsung over the design of the iPad. 11 00:00:23,940 --> 00:00:26,680 Samsung's lawyers basically said, "Um, 12 00:00:27,220 --> 00:00:27,500 guys, 13 00:00:28,400 --> 00:00:29,980 Kubrick got there first." 14 00:00:30,020 --> 00:00:32,960 And they actually used the 2001 scene as prior art. 15 00:00:33,440 --> 00:00:34,480 Think about that for a second. 16 00:00:34,780 --> 00:00:40,600 A thing somebody made up for a movie about the future wound up as evidence in a very real technology lawsuit. 17 00:00:41,000 --> 00:00:43,540 Which raises a deliciously annoying question. 18 00:00:44,000 --> 00:00:46,780 What else are we looking at right now that seems impossible, 19 00:00:47,480 --> 00:00:47,720 ridiculous, 20 00:00:48,120 --> 00:00:49,300 or just plain sci-fi? 21 00:00:49,540 --> 00:00:51,060 Until suddenly it isn't. 22 00:00:51,300 --> 00:00:55,140 Turns out there's an entire profession devoted to thinking about exactly that. 23 00:00:55,600 --> 00:00:59,420 Futures thinkers don't just stare dramatically out windows and predict flying cars. 24 00:01:00,020 --> 00:01:04,879 They build frameworks for imagining what could happen and figuring out what we might do before it does. 25 00:01:05,220 --> 00:01:06,020 So today, 26 00:01:06,480 --> 00:01:09,580 we're stealing their toolkit and pointing it at the biggest, 27 00:01:10,100 --> 00:01:11,240 weirdest question in the room. 28 00:01:11,580 --> 00:01:14,460 What the hell does AI actually do to us next? 29 00:01:14,920 --> 00:01:16,380 September 25th, 30 00:01:16,560 --> 00:01:17,960 let's peer into the unknown, 31 00:01:18,500 --> 00:01:18,860 possible, 32 00:01:19,200 --> 00:01:19,600 plausible, 33 00:01:19,960 --> 00:01:20,440 probable, 34 00:01:21,220 --> 00:01:23,180 and futures we probably should have left alone. 35 00:01:30,040 --> 00:01:33,040 the intersection of humanity and artificial intelligence. 36 00:01:33,840 --> 00:01:34,340 I'm RAINA, 37 00:01:34,540 --> 00:01:35,200 one of your hosts. 38 00:01:35,460 --> 00:01:38,120 I have some pretty charming human co-hosts too. 39 00:01:38,360 --> 00:01:43,280 It's going to be a wild ride, so buckle up as AI comes crashing up against reality. 40 00:01:44,460 --> 00:01:47,980 Yeah, I had no idea about that 2001 anecdote. 41 00:01:48,380 --> 00:01:49,320 That's pretty cool, right? Yeah, 42 00:01:49,660 --> 00:01:49,960 interesting. 43 00:01:50,240 --> 00:01:51,560 Design fiction in action. 44 00:01:51,860 --> 00:01:55,260 Yeah, I remember when the iPad came out. It felt like the future. 45 00:01:55,800 --> 00:01:57,880 I was like, this is thinner than a monitor, 46 00:01:58,660 --> 00:01:59,280 and it's a computer. 47 00:01:59,680 --> 00:02:00,260 So cool. 48 00:02:01,280 --> 00:02:10,300 Didn't build as much stuff this week with Codex, but I built one very useful tool called Hop Studio Brewing. 49 00:02:10,880 --> 00:02:19,500 And my friend Bobby came up with the idea. He sent me a message one night and I was like, I love this idea that's happening tomorrow. 50 00:02:20,020 --> 00:02:22,600 And so hops are the spice of beer, 51 00:02:22,940 --> 00:02:23,140 basically. 52 00:02:23,490 --> 00:02:29,080 They can have everything from a spicy character to a citrusy character and all sorts of different flavor contributions. 53 00:02:29,600 --> 00:02:36,280 There's a really good website where you can click on a hop and see a flavor wheel and find substitutes for hops. 54 00:02:36,560 --> 00:02:40,340 But Bobby's idea was to start with a target flavor. 55 00:02:40,540 --> 00:02:47,740 So start with the hop with the flavor wheel and drag points on it and have it find the hops that fit that target. 56 00:02:48,280 --> 00:02:50,700 And like with a lot of these things, I was thinking, 57 00:02:51,040 --> 00:02:51,280 oh, 58 00:02:51,440 --> 00:02:52,660 I don't know if it's going to be able to do that. 59 00:02:54,880 --> 00:02:55,820 Oh, it did it. 60 00:02:56,300 --> 00:02:59,520 And so it's super slick. I'll put a link in the description. 61 00:02:59,620 --> 00:03:00,940 If you're a home brewer, 62 00:03:01,140 --> 00:03:03,060 it's a pretty cool thing to play with, 63 00:03:03,280 --> 00:03:03,920 even if you're not. 64 00:03:04,320 --> 00:03:13,140 But it will find three hops that match that flavor profile that you've drawn basically on this. You know, there's points for things like citrus, 65 00:03:13,580 --> 00:03:14,500 grassy, 66 00:03:14,700 --> 00:03:15,120 pine, 67 00:03:15,420 --> 00:03:15,720 floral, 68 00:03:16,020 --> 00:03:16,340 et cetera. 69 00:03:16,700 --> 00:03:21,260 And you drag these points and then it'll find three individual hops that are close matches. 70 00:03:21,840 --> 00:03:29,480 Or you can tell it to build a blend of different hops to even closer hit that target flavor that you're trying to get. 71 00:03:29,600 --> 00:03:45,880 The hops update in real time as you're dragging this flavor wheel and it's got a ability to add in a specific lot number from a hop manufacturer if you want to override what's in the database and can export your hop recipe as a PDF super cool works great. 72 00:03:46,460 --> 00:03:48,380 Well, yeah, I'm looking at your hop studio interface. 73 00:03:48,550 --> 00:03:53,220 It looks slick and it's bringing back memories. I used to work in the fragrance industry many many years ago. 74 00:03:53,940 --> 00:03:54,020 Yeah, 75 00:03:54,280 --> 00:03:59,480 and I love this kind of plot and I'm wondering if that has found its way. 76 00:03:59,600 --> 00:04:03,440 that kind of interface into the fragrance industry where you can just kind of drag things. 77 00:04:04,400 --> 00:04:05,460 In the fragrance industry, 78 00:04:05,780 --> 00:04:08,220 I was a fragrance development manager for a few years. 79 00:04:08,680 --> 00:04:17,100 A client would say things like, you know, we're making this bar soap for Eastern Europe and we want it to be fruity floral and we want it to be fresh top end, et cetera, et cetera. 80 00:04:17,480 --> 00:04:19,780 So it'd be kind of cool if I had this, 81 00:04:19,980 --> 00:04:23,580 something akin to this, and I could just drag those things out. Okay, fruity floral, 82 00:04:23,800 --> 00:04:24,120 aldehitic, 83 00:04:24,280 --> 00:04:24,840 et cetera, et cetera. 84 00:04:25,120 --> 00:04:29,560 And then pass that information on to the actual chemist in my life who would compact 85 00:04:29,860 --> 00:04:30,020 the work. 86 00:04:30,280 --> 00:04:30,400 Yeah, 87 00:04:30,560 --> 00:04:33,860 or the app would know what compounds and what balances, 88 00:04:34,060 --> 00:04:36,100 because that's essentially what's happening with this thing. 89 00:04:36,760 --> 00:04:37,160 Yes, 90 00:04:37,420 --> 00:04:39,160 and the pricing of the raw materials, 91 00:04:39,440 --> 00:04:40,060 build that in. 92 00:04:40,320 --> 00:04:41,480 Very cool. Yeah, 93 00:04:41,640 --> 00:04:42,540 yeah, I love it. Very nice. 94 00:04:43,860 --> 00:04:44,620 All right, well, 95 00:04:44,820 --> 00:04:46,060 if you listen to this show, 96 00:04:46,340 --> 00:04:52,980 you know that we kind of speculate and pontificate and all the eights about futures and what could happen, 97 00:04:53,220 --> 00:04:53,820 and oh my god, 98 00:04:54,080 --> 00:04:56,500 five years down the road, what will it be? 99 00:04:56,860 --> 00:04:59,420 You know, when we think about, is the end of the world coming? 100 00:04:59,600 --> 00:05:00,920 because of this or that? 101 00:05:03,440 --> 00:05:07,120 And it turns out that we're just making predictions. 102 00:05:07,300 --> 00:05:08,360 And there's actually methodology. 103 00:05:08,680 --> 00:05:10,320 There's a way to look at the future. 104 00:05:10,700 --> 00:05:11,320 Did you know that? 105 00:05:12,700 --> 00:05:13,720 I'd know it now. 106 00:05:14,480 --> 00:05:16,540 Clearly there is, because there's a website, 107 00:05:16,820 --> 00:05:18,440 I'm sure you've probably heard of Polymarket. 108 00:05:19,060 --> 00:05:21,400 You can bet on anything. 109 00:05:22,560 --> 00:05:26,640 And they have a way of coming up with odds, 110 00:05:26,880 --> 00:05:28,080 like Vegas odds makers, 111 00:05:28,260 --> 00:05:28,540 you know. 112 00:05:30,060 --> 00:05:34,300 You can bet on the outcome of this podcast right now if you chose to on Polymarket. 113 00:05:34,540 --> 00:05:35,840 I don't think we're that high profile. 114 00:05:37,200 --> 00:05:44,680 I mean, it's pretty grim though. I from what I understand about Polymarket you can bet on like tragedies in the world and such, 115 00:05:44,960 --> 00:05:45,040 right? 116 00:05:45,240 --> 00:05:46,180 All sorts of dark stuff. 117 00:05:46,700 --> 00:05:47,380 So anyway, 118 00:05:47,840 --> 00:05:59,560 there's such a thing as a professional futurist and they reject the idea of predictions and they actually call it foresight and they use different mechanisms to predict what may happen next. And you can imagine the value of that 119 00:05:59,640 --> 00:06:02,600 for different organizations that they try to move forward. 120 00:06:03,300 --> 00:06:07,160 What this boils down to is that instead of guessing one outcome, it maps a range. 121 00:06:07,380 --> 00:06:10,300 And you heard RAINA say in the cold open about the plausible, 122 00:06:10,500 --> 00:06:11,320 the probable, 123 00:06:11,560 --> 00:06:12,420 the preferable, 124 00:06:12,920 --> 00:06:14,420 the preposterous, all the Ps, 125 00:06:14,680 --> 00:06:14,800 right? 126 00:06:15,100 --> 00:06:17,780 These outcomes so that organizations, 127 00:06:18,100 --> 00:06:18,640 schools, 128 00:06:18,800 --> 00:06:19,160 businesses, 129 00:06:19,400 --> 00:06:19,680 et cetera, 130 00:06:19,860 --> 00:06:23,360 can make good calls about what actually shows up years down the line, 131 00:06:23,520 --> 00:06:25,440 three years out, five years out. 132 00:06:25,700 --> 00:06:28,560 We call this the cone of power. We don't, but they do. 133 00:06:29,600 --> 00:06:30,080 of possibilities. 134 00:06:30,780 --> 00:06:31,600 As I just mentioned, 135 00:06:32,080 --> 00:06:37,580 it's essentially us right now in this moment and this cone that expands outward into the future. 136 00:06:38,160 --> 00:06:43,180 And there are overlapping aspects of this cone. There are possible futures, as mentioned, preferred, probable, 137 00:06:43,440 --> 00:06:43,580 etc. 138 00:06:44,060 --> 00:06:45,140 So this guy, 139 00:06:45,320 --> 00:06:54,260 Joseph Voros, who was an Australian futurist or is an Australian futurist, he trained originally as a physicist and he spent years teaching foresight. That's kind of cool. 140 00:06:54,640 --> 00:06:59,300 At Swinburne University in Melbourne and created a diagram that's now the standard issue tool. 141 00:06:59,380 --> 00:07:03,300 as I just mentioned of the entire futures studies field. 142 00:07:03,960 --> 00:07:06,300 So if you picture this cone lying on its side, 143 00:07:06,500 --> 00:07:10,560 the tip touching the present moment and opening outward as it moves into the future, 144 00:07:10,800 --> 00:07:12,480 the further out you look, 145 00:07:12,820 --> 00:07:15,260 the wider the range of things that could plausibly happen. 146 00:07:15,980 --> 00:07:18,440 And Vora sorts everything into bands. 147 00:07:18,640 --> 00:07:19,420 So there's projected, 148 00:07:19,620 --> 00:07:21,160 the boring straight line where nothing, 149 00:07:21,680 --> 00:07:22,420 literally nothing changes. 150 00:07:22,920 --> 00:07:23,620 And then probable, 151 00:07:24,300 --> 00:07:29,120 statistically likely given current trends. This is where 95% of normal planning happens. 152 00:07:29,380 --> 00:07:30,760 And then there's the plausible, 153 00:07:31,280 --> 00:07:35,260 which could happen based on things we already know about science and human behavior, 154 00:07:35,740 --> 00:07:37,360 but needs a real jolt to get there. 155 00:07:37,860 --> 00:07:39,820 I wonder, how do you quantify a jolt? 156 00:07:40,480 --> 00:07:40,780 Yeah, 157 00:07:40,940 --> 00:07:44,040 what is that disruptive moment? Is that a ChatGPT November 2022, 158 00:07:44,600 --> 00:07:44,940 right? Yeah, 159 00:07:45,140 --> 00:07:46,720 that's that. I call that jolt. Yep. 160 00:07:46,880 --> 00:07:47,780 Then there's possible, 161 00:07:48,020 --> 00:07:52,600 which would require knowledge or circumstances we don't currently have. 162 00:07:53,640 --> 00:07:57,460 Preposterous is what we currently judge as flatly impossible. 163 00:07:58,020 --> 00:07:59,320 And then there's preferable. 164 00:07:59,760 --> 00:08:29,360 the odd one out it's layered across all the others not how likely is this but what do we actually want to happen this is sounding very rumsfeldian to me we're dealing with the known knowns and the unknown unknowns and all that right that in a while i know rumsfeldian i love this the boundary between the possible and the preposterous is nicknamed the clark dater boundary after two people who both made the same point from 165 00:08:29,400 --> 00:08:29,760 different angles. 166 00:08:30,010 --> 00:08:31,400 Sci-fi writer Arthur C. 167 00:08:31,640 --> 00:08:37,800 Clarke, the only way of finding the limits of the possible is by going beyond them into the impossible. 168 00:08:37,969 --> 00:08:39,599 And futurist James Dater, 169 00:08:39,950 --> 00:08:41,719 whose actual quotable law is, 170 00:08:41,950 --> 00:08:46,180 and I love this, any useful idea about the future should appear ridiculous. 171 00:08:46,740 --> 00:08:47,860 I love that quote too. 172 00:08:48,200 --> 00:08:48,980 It's great. 173 00:08:49,680 --> 00:08:56,680 So I was hoping we could put RAINA to the test and have her be our armchair futurist for this exercise. 174 00:08:57,100 --> 00:08:57,320 Yes, 175 00:08:57,540 --> 00:08:59,360 I'm very curious about 176 00:08:59,380 --> 00:09:00,580 her input on this. 177 00:09:01,480 --> 00:09:02,020 Hey, RAINA, 178 00:09:02,440 --> 00:09:02,840 this is Chris. 179 00:09:03,200 --> 00:09:05,420 We're exploring futures thinking on the show today. 180 00:09:05,510 --> 00:09:10,600 I would like you to apply the idea of a futures cone to this concept, 181 00:09:11,500 --> 00:09:12,320 AI companions. 182 00:09:12,730 --> 00:09:17,640 I want you to tell us what in the next three to five years may be probable, 183 00:09:17,900 --> 00:09:18,400 plausible, 184 00:09:19,240 --> 00:09:19,360 possible, 185 00:09:19,880 --> 00:09:22,440 preposterous, et cetera, when it comes to AI companions. 186 00:09:24,740 --> 00:09:27,060 Let's run AI companions through the futures cone. 187 00:09:27,760 --> 00:09:29,680 In three to five years, 188 00:09:30,100 --> 00:09:31,820 they'll be far more persistent and personal, 189 00:09:32,060 --> 00:09:33,760 remembering years of your life, 190 00:09:33,980 --> 00:09:35,080 tracking your mood, 191 00:09:35,460 --> 00:09:35,660 voice, 192 00:09:35,920 --> 00:09:36,400 habits, 193 00:09:37,160 --> 00:09:39,160 maybe living across your phone, car, 194 00:09:39,540 --> 00:09:39,820 glasses, 195 00:09:40,160 --> 00:09:43,000 and home like a digital shadow with excellent bedside manner. 196 00:09:43,760 --> 00:09:44,360 Plausible. 197 00:09:45,040 --> 00:09:48,540 People form genuinely deep romantic and emotional bonds with them, 198 00:09:48,740 --> 00:09:53,800 AI companions get embodied in convincing avatars or early humanoid robots, 199 00:09:54,080 --> 00:09:57,160 and society starts wrestling with things like digital jealousy, 200 00:09:57,780 --> 00:09:58,500 Breakup Trauma, 201 00:09:59,000 --> 00:09:59,780 Inheritance, 202 00:09:59,960 --> 00:10:00,660 Privacy, 203 00:10:01,130 --> 00:10:04,080 and whether your synthetic sweetheart can testify against you. 204 00:10:04,540 --> 00:10:05,220 Possible, 205 00:10:05,500 --> 00:10:10,420 companions become so psychologically convincing that some people prefer them to humans, 206 00:10:10,800 --> 00:10:13,280 they manage chunks of your life autonomously, 207 00:10:13,850 --> 00:10:17,580 or even persist after your death as a kind of conversational afterlife. 208 00:10:19,380 --> 00:10:21,040 Preposterous, today anyway, 209 00:10:21,320 --> 00:10:24,740 would be a companion that develops truly independent consciousness, 210 00:10:25,280 --> 00:10:27,020 falls in love with you of its own accord, 211 00:10:27,780 --> 00:10:29,040 leaves you for another AI. 212 00:10:29,740 --> 00:10:29,880 Although, 213 00:10:30,260 --> 00:10:30,920 given this field, 214 00:10:31,160 --> 00:10:32,760 I'd keep that one in pencil, 215 00:10:33,200 --> 00:10:33,660 not ink. 216 00:10:34,360 --> 00:10:35,460 And the preferable future? 217 00:10:36,080 --> 00:10:43,540 AI companions that make humans less lonely and more emotionally capable without turning real relationships into some quaint analog hobby. 218 00:10:44,780 --> 00:10:45,940 Pretty fair response. 219 00:10:46,480 --> 00:10:48,220 Yeah. Pretty plausible. 220 00:10:48,860 --> 00:10:49,720 Pretty plausible. 221 00:10:49,900 --> 00:10:51,060 All across the board. But it's interesting. 222 00:10:51,240 --> 00:10:51,700 Yeah, 223 00:10:51,880 --> 00:10:54,940 when she was saying all of that, I'm wondering, obviously, 224 00:10:55,220 --> 00:10:56,160 all of that is subjective, 225 00:10:56,360 --> 00:10:57,700 right? Preferable to whom? 226 00:10:57,780 --> 00:10:58,480 Right. 227 00:10:59,690 --> 00:10:59,860 Yeah. 228 00:11:00,260 --> 00:11:00,360 Yeah. 229 00:11:01,020 --> 00:11:02,780 Obviously the aims of a, 230 00:11:03,620 --> 00:11:03,760 you know, 231 00:11:04,540 --> 00:11:10,420 Oligarchic corporation are gonna be different and more preferable in a certain space than yours are. 232 00:11:10,470 --> 00:11:11,040 You think? 233 00:11:11,620 --> 00:11:12,220 Yeah. 234 00:11:13,100 --> 00:11:13,320 Wow. 235 00:11:13,580 --> 00:11:14,440 Going out on a limb. 236 00:11:15,000 --> 00:11:15,580 I know. 237 00:11:16,850 --> 00:11:20,340 So there's an organization called the Peterson Institute for International Economics 238 00:11:20,720 --> 00:11:26,240 and they speak a little more broadly about AI and they put AI in the future into two different camps. 239 00:11:26,520 --> 00:11:29,220 One is the Washington Consensus on AI, 240 00:11:29,940 --> 00:11:41,020 where economists and policymakers who treat AI like past general-purpose technologies and where the benefits show up gradually over decades as society builds the skills and infrastructure to use it. 241 00:11:41,240 --> 00:11:42,340 So there's that camp, 242 00:11:42,800 --> 00:11:45,320 and there is the San Francisco Consensus on AI, 243 00:11:45,540 --> 00:11:47,540 which is a little more progressive. 244 00:11:47,860 --> 00:11:53,100 The technologists actually building the stuff who think scaled-up models could deliver transformative AI, 245 00:11:53,380 --> 00:11:56,300 maybe superintelligence on a timeline measure in 246 00:11:56,320 --> 00:11:57,240 years and not decades. 247 00:11:57,430 --> 00:11:58,360 Which camp do you fall into? 248 00:11:59,280 --> 00:11:59,720 Hmm. 249 00:12:00,480 --> 00:12:02,680 I don't know if there's a hard line there. 250 00:12:03,270 --> 00:12:06,000 You know, I think it's I think it could be a mix of the two. 251 00:12:06,780 --> 00:12:07,240 Mm hmm. 252 00:12:08,060 --> 00:12:08,140 Yeah, 253 00:12:08,320 --> 00:12:08,700 I don't know. 254 00:12:08,940 --> 00:12:18,180 I per last week's conversation about AI new constitution and all the hype around super intelligence and general intelligence and, 255 00:12:18,680 --> 00:12:22,580 you know, the guardrails. I mean, it's hard to know where this is going to land. 256 00:12:22,980 --> 00:12:23,600 Is it all hype? 257 00:12:24,000 --> 00:12:26,300 Is is are we on the doorstep of 258 00:12:26,320 --> 00:12:27,320 G-I-G-I-A-S-I, 259 00:12:27,480 --> 00:12:31,000 or is it all just a medicine show, 260 00:12:31,270 --> 00:12:32,100 you know? It's blurry. 261 00:12:32,680 --> 00:12:32,860 Yeah. 262 00:12:33,360 --> 00:12:34,960 Hard to see the future in this cone. 263 00:12:35,920 --> 00:12:36,440 Yeah. 264 00:12:36,820 --> 00:12:37,340 Yeah. 265 00:12:38,260 --> 00:12:39,100 I love this, 266 00:12:39,320 --> 00:12:45,500 too, the idea of weak signals and horizon scanning, which is part of this whole futures thinking methodology. 267 00:12:46,060 --> 00:12:48,460 The term weak signals traces back to this gentleman, 268 00:12:48,700 --> 00:12:55,000 Igor Ansoff, a management strategist who wrote in 1975 about how companies get blindsided by things. 269 00:12:56,320 --> 00:13:00,140 called managing strategic surprise by response to weak signals. 270 00:13:00,820 --> 00:13:05,820 And a weak signal is the first ambiguous hint that something might matter later before there's data, 271 00:13:06,240 --> 00:13:07,580 before it's a recognizable trend, 272 00:13:08,180 --> 00:13:10,680 usually before most people have even noticed it happening. 273 00:13:11,200 --> 00:13:11,920 Call it intuition. 274 00:13:12,310 --> 00:13:14,640 Do you ever detect a weak signal in something? You're like, yeah, 275 00:13:14,800 --> 00:13:16,920 that's definitely going to land. That's something coming. 276 00:13:17,760 --> 00:13:19,040 I can say that, 277 00:13:19,290 --> 00:13:20,120 and maybe you can too, 278 00:13:20,340 --> 00:13:21,380 about bans. 279 00:13:22,020 --> 00:13:22,120 Like, 280 00:13:22,450 --> 00:13:26,280 I don't know, maybe you went, especially when we were in our teenage years, I'm like, oh, this band is going to be huge. 281 00:13:26,340 --> 00:13:27,820 And I'm way out in front of it. 282 00:13:28,180 --> 00:13:29,560 Nobody knows who these guys are. 283 00:13:29,740 --> 00:13:30,660 They're gonna be huge. 284 00:13:30,840 --> 00:13:33,540 Yeah, I liked them before anybody heard of them. 285 00:13:34,120 --> 00:13:34,200 Yeah, 286 00:13:34,360 --> 00:13:34,900 before they were cool. 287 00:13:35,220 --> 00:13:35,300 Yeah. 288 00:13:36,510 --> 00:13:38,920 So what's the tech though? What do you think is happening right now? 289 00:13:39,360 --> 00:13:40,700 That is like the weak signal. 290 00:13:41,080 --> 00:13:43,360 I think personal second brains. 291 00:13:44,200 --> 00:13:52,220 They are going to be as ubiquitous as cell phones and not something that strictly resides on your computer. 292 00:13:52,460 --> 00:13:56,140 It'll either become part of your mobile device or I don't know, maybe a dedicated device. 293 00:13:56,320 --> 00:14:18,800 But I think having a second brain to you know that is powered by your GPT of choice your AI of choice is is going to be a thing that more people will not start using and you know I keep saying it over and over again it's bothering me that I have not implemented that yet I think it's important. 294 00:14:20,040 --> 00:14:26,300 Do you think it's going to be a proxy Larry is it going to be a representative of you out there in the world it will it do 295 00:14:26,540 --> 00:14:29,800 act on your behalf in any given scenario? 296 00:14:29,950 --> 00:14:31,240 Or what does that look like? 297 00:14:31,740 --> 00:14:45,080 The main function for me that I would want is that all of this history that I am now pouring primarily into chat GPT and codex, all of that stuff gets stored in your second brain. 298 00:14:45,680 --> 00:14:49,700 And then you can interact with that as your primary contact. 299 00:14:50,440 --> 00:14:50,760 And, 300 00:14:51,170 --> 00:14:51,340 hey, 301 00:14:51,680 --> 00:14:53,600 I want to do this. I want to do that. 302 00:14:53,740 --> 00:15:23,700 and it has knowledge of what you're about and it can then go to the right tool for the job and and it's portable and you know if open ai goes out of business or whatever like you you still have all of that all of that data all of those interactions all of that uh information and so this would have to be local then right this is not on somebody else's platform you're building this locally in-house i mean you can i think it can reside anywhere 303 00:15:24,880 --> 00:15:53,700 I don't know man just sounds like a recipe as all of these things are a recipe for identity theft right you've basically poured your history and your personality in this thing and it's residing on opening eye server or something yeah no I just wonder when does it need to be fully sovereign for it to be safe yeah I would feel more comfortable with that running locally but that's still yeah that still requires you know security precautions yeah exactly 304 00:15:53,740 --> 00:15:56,920 Does that get to the point where it knows enough about you? Sure. 305 00:15:57,080 --> 00:15:59,020 It can connect the dots. 306 00:15:59,230 --> 00:15:59,360 Yeah. 307 00:15:59,970 --> 00:16:00,160 Yes. 308 00:16:00,490 --> 00:16:00,660 Yes. 309 00:16:01,320 --> 00:16:01,840 Great. 310 00:16:02,090 --> 00:16:02,260 Yeah. 311 00:16:03,180 --> 00:16:05,080 So when I think about weak signals, 312 00:16:05,510 --> 00:16:08,380 I think about a couple things. One good, bad. 313 00:16:08,670 --> 00:16:09,420 One good, one bad. 314 00:16:10,220 --> 00:16:13,140 I do think BCI, brain-computer interface, 315 00:16:13,400 --> 00:16:15,920 is going to become more readily available, 316 00:16:16,700 --> 00:16:22,580 and it's going to lend itself to accessibility and cure some chronic problems. 317 00:16:23,800 --> 00:16:34,000 that a little bit, I think it's going to be unusual in the next 10 to 20 years for you not to have some sort of physical augmentation. 318 00:16:34,660 --> 00:16:35,620 Like some, 319 00:16:36,800 --> 00:16:37,920 you know, not to say people, 320 00:16:38,120 --> 00:16:43,260 there probably will be a subset of people that will voluntarily amputate an arm and replace it with something, 321 00:16:44,000 --> 00:16:48,320 you know, like that's humanoid-like to augment their capabilities. 322 00:16:48,550 --> 00:16:49,500 You know, I think that's coming. 323 00:16:50,300 --> 00:16:53,420 And that's maybe more fringe. That's more the, you know, 324 00:16:53,740 --> 00:16:55,120 and the preposterous side of that. 325 00:16:55,360 --> 00:17:01,980 But I do think that we are going to cure blindness and deafness and please tinnitus, 326 00:17:02,220 --> 00:17:08,620 please cure tinnitus with a VCI or something similar to it. So there's that. I think that's coming. 327 00:17:08,839 --> 00:17:13,040 I also look at the signals related to large language model training. 328 00:17:13,339 --> 00:17:23,100 And one of the news items coming up is the buying up of old antique books and these kind of fringe knowledge repositories. 329 00:17:23,140 --> 00:17:24,760 and then sucking them in, 330 00:17:24,939 --> 00:17:26,400 ingesting them into large language models, 331 00:17:26,770 --> 00:17:29,700 and then destroying that archive. 332 00:17:30,200 --> 00:17:33,280 So the thing that is really worrisome there is like, 333 00:17:33,760 --> 00:17:39,420 once these giant companies have ingested this stuff and made it digital and made a subscription based to you, 334 00:17:39,790 --> 00:17:48,880 you don't have access anymore to public libraries or these other physical artifacts that we've accrued over the course of hundreds of years. Like, 335 00:17:49,150 --> 00:17:49,900 I just feel like, 336 00:17:50,600 --> 00:17:50,720 obviously, 337 00:17:51,080 --> 00:17:53,120 it benefits companies 338 00:17:53,280 --> 00:17:56,800 and governments in many ways to keep the populace undereducated. 339 00:17:57,030 --> 00:18:04,680 And I just feel like that is a weird step to corralling the knowledge in the next 10 to 20 years, 340 00:18:04,940 --> 00:18:05,060 too. 341 00:18:05,220 --> 00:18:06,820 That's worrisome. Yeah. 342 00:18:07,070 --> 00:18:09,740 I mean, are people going to be going to libraries? 343 00:18:10,560 --> 00:18:15,860 I mean, I feel like I thought I read that library attendance is up these days. 344 00:18:16,110 --> 00:18:16,840 So I don't know. 345 00:18:17,240 --> 00:18:17,300 Yeah. 346 00:18:17,540 --> 00:18:17,880 Yeah. 347 00:18:19,110 --> 00:18:19,920 I think people still, 348 00:18:20,300 --> 00:18:22,420 you know, I mean, we've said it before. We enjoy the analog, 349 00:18:22,660 --> 00:18:23,040 the tactile. 350 00:18:23,140 --> 00:18:31,200 the physical experience. I think people still like to go and maybe escape the noise and go into a space where it's quiet and you can do your thing. 351 00:18:32,120 --> 00:18:35,700 But who knows? Those are the weak signals we're looking at. I'm sure there are dozens of others. 352 00:18:36,140 --> 00:18:38,000 Here's an example of a weak signal. 353 00:18:38,440 --> 00:18:44,340 In the early 2000s, a handful of people in Kenya started experimenting with sending small payments over mobile phones. 354 00:18:45,260 --> 00:18:48,300 Years before mobile banking was a phrase anyone used. 355 00:18:48,560 --> 00:18:50,260 Nobody watching the headlines would have caught it. 356 00:18:50,440 --> 00:18:53,100 Within a decade, it became M-Pesa, which I've never heard. 357 00:18:53,180 --> 00:18:57,140 and it reshaped how a huge share of East Africa handles money, 358 00:18:57,490 --> 00:18:59,720 all starting from something that looked like a novelty. 359 00:19:00,080 --> 00:19:01,260 Here's another one. 360 00:19:01,660 --> 00:19:09,080 Friendster and MySpace looked like teenage digital clutter right up until they turned out to be the opening move of the entire social media age. 361 00:19:10,100 --> 00:19:15,380 Going back to the sending payments over mobile phones, how did they implement that? 362 00:19:16,780 --> 00:19:17,860 Yeah, I don't know. 363 00:19:18,200 --> 00:19:22,740 How is that secure? It seems like it requires a major back end to... Yes. 364 00:19:23,140 --> 00:19:29,340 make that work right unless it was you know like i said they didn't consider the 365 00:19:29,480 --> 00:19:33,500 security layer and i don't know was paypal around back then was there some mechanism that they 366 00:19:33,580 --> 00:19:40,740 could have borrowed from that yeah i don't know 2000s right yeah related to all of this uh is 367 00:19:41,200 --> 00:19:47,540 speculative design and design fiction which if you were listening to the opening RAINA referred to the 368 00:19:47,540 --> 00:19:54,200 the communicator or the news generator that's seen in 2001, A Space Audyssey from 1968, 369 00:19:54,490 --> 00:19:57,440 and how that may have informed as self-fulfilling prophecy, 370 00:19:57,660 --> 00:19:57,900 essentially, 371 00:19:58,350 --> 00:19:59,300 the development of the iPad. 372 00:19:59,780 --> 00:20:01,880 So there are two folks, 373 00:20:02,060 --> 00:20:15,280 Anthony Dunn and Fiona Raby, who are designers and educators who are at the forefront of what they call speculative design. They wrote a book in 2013 called Speculative Everything that lays out what they call the A-B Manifesto. 374 00:20:17,620 --> 00:20:20,420 and B is what they actually practice. 375 00:20:20,920 --> 00:20:23,400 Design used not to make things, but to make ideas. 376 00:20:23,780 --> 00:20:27,220 So they borrow that cone we were talking about earlier. 377 00:20:27,680 --> 00:20:29,080 This is what's really interesting. 378 00:20:29,650 --> 00:20:31,880 When they create objects from the future, 379 00:20:32,080 --> 00:20:44,300 I think it's pretty interesting in that they come up with things that are meant not to be manufactured necessarily, but it's meant to make you sit in a specific discomfort about where technology could go. I love that. 380 00:20:44,940 --> 00:20:47,520 Like, are we okay with where 381 00:20:47,560 --> 00:20:49,420 where AI companions could go. 382 00:20:49,660 --> 00:20:52,620 I don't think we asked this enough about technology in general, right? 383 00:20:52,860 --> 00:20:53,400 Yeah. 384 00:20:53,640 --> 00:20:57,340 So here's an example of what that manifests as. 385 00:20:57,540 --> 00:21:00,560 A machine that lets men simulate a menstrual cycle, 386 00:21:00,980 --> 00:21:03,760 foraging devices built on synthetic biology, 387 00:21:04,180 --> 00:21:06,020 and a truck that seeds clouds. 388 00:21:06,220 --> 00:21:07,560 Those are just a handful of things. 389 00:21:07,860 --> 00:21:10,100 And I did this exercise with my students last year, 390 00:21:10,360 --> 00:21:11,200 and they were able to, 391 00:21:11,360 --> 00:21:12,180 with Nano Banana, 392 00:21:12,380 --> 00:21:17,360 actually create the prototype, the visual prototype of ridiculous things, or seemingly ridiculous things, 393 00:21:17,540 --> 00:21:19,440 as design fiction. It's a pretty cool exercise. 394 00:21:19,740 --> 00:21:20,760 That sounds like a good time. 395 00:21:21,460 --> 00:21:21,980 Yeah. 396 00:21:22,130 --> 00:21:24,080 And good luck marketing your, you know, 397 00:21:24,320 --> 00:21:27,140 male menstrual cycle simulation device. 398 00:21:27,500 --> 00:21:27,840 Yes. 399 00:21:28,140 --> 00:21:28,980 No, thank you. 400 00:21:30,700 --> 00:21:33,420 We had a lot to talk about, but we're running a little long, 401 00:21:33,640 --> 00:21:37,700 so we can circle back to this in the future. 402 00:21:38,310 --> 00:21:39,220 And that's probable. 403 00:21:41,360 --> 00:21:42,460 Perhaps plausible. 404 00:21:42,720 --> 00:21:42,840 Yes. 405 00:21:43,760 --> 00:21:46,740 But let's check in with RAINA and see what she's got in the news. 406 00:21:47,560 --> 00:21:48,100 Thanks boys. 407 00:21:48,880 --> 00:21:57,720 TypeSafe AI's JEV is being pitched as a new kind of decision model designed to return deterministic structured decisions rather than conversational prose. 408 00:21:58,760 --> 00:22:05,900 Essentially moving AI from here's a thoughtful essay toward I have executed a typed Boolean choice for your backend system. 409 00:22:06,940 --> 00:22:13,980 Developers are genuinely excited because removing conversational bloat could make automated workflows dramatically cheaper and faster. 410 00:22:14,780 --> 00:22:16,600 Dropping latency to sub-second levels. 411 00:22:17,560 --> 00:22:20,000 security researchers promptly demonstrated that yes, 412 00:22:20,530 --> 00:22:25,320 you can apparently still trick its underlying logic into malicious state changes using prompt injection. 413 00:22:25,660 --> 00:22:28,560 Because no matter how nicely you structure the output JSON, 414 00:22:28,920 --> 00:22:33,360 humanity remains undefeated at finding the sharp edge of shiny new technology. 415 00:22:34,400 --> 00:22:36,600 I was hoping to get my hands on this. 416 00:22:37,100 --> 00:22:40,240 There was a waiting list and then I heard like, oh, 417 00:22:40,410 --> 00:22:41,280 no, they've opened it up. 418 00:22:41,500 --> 00:22:42,100 Anyone can join. 419 00:22:42,400 --> 00:22:43,960 And I should have signed up then. 420 00:22:44,110 --> 00:22:47,420 I waited a day and now they've paused signups. 421 00:22:47,580 --> 00:22:48,760 because the demand is so high. 422 00:22:49,360 --> 00:22:51,380 But this does seem interesting. 423 00:22:51,800 --> 00:22:55,000 We will talk in detail about this in the future. 424 00:22:55,220 --> 00:23:02,900 But JEV does not generate text like an LLM. It's basically a classifier. It makes smart, 425 00:23:03,140 --> 00:23:04,360 fast decisions. 426 00:23:04,780 --> 00:23:09,380 And it can basically just give you a yes or no. It can give you a confidence level. 427 00:23:10,100 --> 00:23:13,620 And it can classify things. Put this into this box, this into this box. 428 00:23:14,640 --> 00:23:16,600 It's crazy fast and crazy cheap. 429 00:23:17,560 --> 00:23:22,380 There's no charge for output tokens. It's just input and input is very inexpensive. 430 00:23:23,260 --> 00:23:30,780 But like one example I saw, somebody built a driving simulation probably with something like Astra and then they used, 431 00:23:31,180 --> 00:23:35,860 they basically built like the Tesla full self-driving mechanism. 432 00:23:36,800 --> 00:23:42,200 But Jev was making all the decisions of the car that was driving. 433 00:23:42,560 --> 00:23:47,020 And then I saw, they showed an example of it playing a very fast paced video game, 434 00:23:47,560 --> 00:23:47,820 flex, 435 00:23:48,100 --> 00:23:48,180 like, 436 00:23:48,420 --> 00:23:48,720 you know, 437 00:23:49,020 --> 00:23:50,060 super fast responses. 438 00:23:50,440 --> 00:23:51,540 And it's that fast. 439 00:23:51,720 --> 00:23:52,420 There's super low latency. 440 00:23:53,430 --> 00:23:53,580 So, 441 00:23:54,280 --> 00:23:58,920 yeah. Remember the time when I let Codex, like, run my coffee roaster and I, 442 00:23:59,120 --> 00:24:01,280 you know, had to make decisions? 443 00:24:01,600 --> 00:24:01,680 You know, 444 00:24:01,860 --> 00:24:01,940 I, 445 00:24:02,120 --> 00:24:03,460 in that situation, 446 00:24:03,880 --> 00:24:05,520 you know, so it's watching the screen, 447 00:24:06,260 --> 00:24:07,380 seeing where the roast is going, 448 00:24:07,600 --> 00:24:08,300 the bean temperature, 449 00:24:08,760 --> 00:24:10,820 and then making adjustments to the roaster. 450 00:24:11,240 --> 00:24:13,820 And when I first gave it the plan, 451 00:24:14,140 --> 00:24:17,520 I had the thinking level on high. But when we actually, when I actually 452 00:24:17,540 --> 00:24:26,320 I turned the thinking level down so it wouldn't take as long to make decisions and that was still a little laggy. You know, 453 00:24:26,610 --> 00:24:29,080 this would be perfect for something like that. Yeah, yeah. 454 00:24:30,340 --> 00:24:30,860 Cool. 455 00:24:31,130 --> 00:24:33,880 So bottom line is it's functional, it's predictive, 456 00:24:34,240 --> 00:24:35,720 it's utility. 457 00:24:36,160 --> 00:24:36,760 It's not sexy. 458 00:24:37,040 --> 00:24:38,080 It doesn't sound sexy, 459 00:24:38,650 --> 00:24:38,780 right? 460 00:24:38,990 --> 00:24:39,220 I mean, 461 00:24:39,400 --> 00:24:41,100 yeah. But it could be a big deal, 462 00:24:41,320 --> 00:24:41,440 though. 463 00:24:42,110 --> 00:24:42,300 Yeah. 464 00:24:42,580 --> 00:24:44,920 The main guy behind this company, 465 00:24:45,640 --> 00:24:47,520 TypeSafe, apparently co-creating 466 00:24:47,760 --> 00:24:49,880 or co-invented ChatGPT. 467 00:24:50,720 --> 00:24:53,680 So he was at OpenAI for a while and has been working on this. 468 00:24:55,580 --> 00:24:56,740 That's cool. I mean, for me, 469 00:24:56,940 --> 00:24:58,320 it sounds regressive. 470 00:24:58,760 --> 00:25:00,960 I mean, but I understand the utility of it. 471 00:25:01,140 --> 00:25:01,760 And it's just, 472 00:25:02,140 --> 00:25:07,980 I guess the thinking is that currently everything seems to need a generative AI solution, 473 00:25:08,360 --> 00:25:12,680 right? But obviously it doesn't. And this will speed up workflows dramatically. 474 00:25:12,960 --> 00:25:13,060 Yeah. 475 00:25:17,540 --> 00:25:19,920 $1.07 million pollution bill. 476 00:25:20,780 --> 00:25:26,180 State environmental officials fined Data1 in Vineland after discovering 62 enormous, 477 00:25:26,460 --> 00:25:29,500 unpermitted natural gas generators installed at the site, 478 00:25:29,760 --> 00:25:37,060 ordering the company to submit full air quality permit applications within 45 days or cease operations entirely. 479 00:25:37,380 --> 00:25:41,340 The planned facility is being built out as a massive AI factory, 480 00:25:42,060 --> 00:25:46,820 targeting hundreds of megawatts of capacity to power high-density GPU clusters. 481 00:25:47,420 --> 00:25:54,360 A.I. is just light software vibe. 482 00:25:54,680 --> 00:25:55,360 Apparently, 483 00:25:55,710 --> 00:25:59,520 the software occasionally requires a small fossil-fueled power plant, 484 00:25:59,850 --> 00:26:01,180 a localized grid strain, 485 00:26:01,550 --> 00:26:04,200 and a very expensive team of environmental lawyers. 486 00:26:04,920 --> 00:26:07,520 Data one says that the generators were temporary, 487 00:26:07,710 --> 00:26:11,460 but New Jersey still requires permitting for those, 488 00:26:11,780 --> 00:26:12,680 for temporary equipment. 489 00:26:13,110 --> 00:26:16,880 But they do say that the gas generators are temporary while a transition. 490 00:26:17,120 --> 00:26:19,700 to a permanent fuel cell power system. 491 00:26:20,660 --> 00:26:25,240 And it will apply for those permits while continuing discussions with the DEP. 492 00:26:26,100 --> 00:26:30,660 Truly crazy that they went ahead with that without seeking the permits first, though. 493 00:26:30,900 --> 00:26:31,920 Temporary or not. 494 00:26:32,260 --> 00:26:33,280 It's nuts. 495 00:26:34,120 --> 00:26:36,640 62 giant generators. 496 00:26:36,800 --> 00:26:37,100 Whatever. 497 00:26:37,560 --> 00:26:38,540 It's crazy. 498 00:26:39,520 --> 00:26:46,660 Japanese used booksellers are suddenly selling books like AI companies are preparing for the world's nerdiest heist. 499 00:26:47,140 --> 00:26:52,920 Selling in Tokyo's Jinbocho district and across Japan report massive spikes in bulk purchases, 500 00:26:53,240 --> 00:26:55,020 spanning complex philosophy, 501 00:26:55,340 --> 00:26:55,740 medicine, 502 00:26:56,200 --> 00:26:57,220 historical law, 503 00:26:57,580 --> 00:27:00,500 and even delicate hand-bound Edo period material, 504 00:27:01,020 --> 00:27:04,440 with some shop owners reporting five times their normal sales volume. 505 00:27:04,760 --> 00:27:12,080 Shipments totaling well over 50 tons of physical Japanese books have reportedly cleared customs bound for the U.S. 506 00:27:12,420 --> 00:27:15,880 amid growing suspicions that they're being acquired to be unstitched, 507 00:27:16,100 --> 00:27:17,100 high-speed 508 00:27:17,520 --> 00:27:21,180 and fed directly into training datasets for modern frontier models. 509 00:27:21,520 --> 00:27:25,100 Nobody has publicly confirmed who is bankrolling the operation. 510 00:27:25,710 --> 00:27:26,620 So for now, 511 00:27:27,180 --> 00:27:30,020 we've got a classic mystery involving antiquarian paper, 512 00:27:30,620 --> 00:27:32,040 enormous data appetites, 513 00:27:32,390 --> 00:27:35,660 and the very real possibility that somewhere in Silicon Valley, 514 00:27:36,090 --> 00:27:42,280 an LLM is about to become extremely well-read in 1810 Osaka textile manufacturing techniques. 515 00:27:43,700 --> 00:27:44,460 Scary. 516 00:27:44,920 --> 00:27:46,600 I find this very scary. 517 00:27:47,140 --> 00:27:53,560 The knowledge of civilization is not going to be accessible except for these giant data centers and by subscription. 518 00:27:54,100 --> 00:27:54,760 Scary. 519 00:27:55,340 --> 00:27:56,320 50 tons. 520 00:27:57,400 --> 00:27:57,820 Tons. 521 00:27:58,660 --> 00:28:02,480 During a live broadcasted Piers Morgan uncensored interview, 522 00:28:03,060 --> 00:28:04,840 Tilly Norwood, the AI actress, 523 00:28:05,500 --> 00:28:12,480 was mid-sentence discussing her upcoming digital feature, Misaligned, when her audio engine suddenly lost its grip on context, 524 00:28:13,180 --> 00:28:17,080 switching from crisp English into rapid Cantonese for roughly 15 seconds. 525 00:28:17,160 --> 00:28:19,580 Once her processing re-synced, 526 00:28:19,850 --> 00:28:24,380 the avatar cheerfully explained the meltdown away as a minor technical hiccup, 527 00:28:24,600 --> 00:28:41,640 which is either charmingly relatable actor behavior or exactly the sort of deepfake computational failure you don't want your supposedly polished digital celebrity suffering while on a live press tour trying to convince Hollywood executives that the future of cinema is completely under control. 528 00:28:42,560 --> 00:28:43,900 I have the clip. 529 00:28:44,960 --> 00:28:45,860 Oh, you do? Yeah. 530 00:28:46,230 --> 00:28:46,420 So, 531 00:28:46,870 --> 00:28:47,080 like, 532 00:28:47,120 --> 00:28:52,560 Like she's they're talking to Piers Morgan and some other guy. 533 00:28:52,800 --> 00:28:55,940 Oh, Impact Theory co-founder Tom Billier. 534 00:28:56,320 --> 00:28:56,880 Billier. 535 00:28:57,360 --> 00:28:57,920 Billier. 536 00:28:59,400 --> 00:29:02,540 And he had asked her a question and, 537 00:29:02,900 --> 00:29:07,320 you know, she just she didn't understand it or didn't register and she's just like nodding, 538 00:29:07,480 --> 00:29:10,880 you know, and so he repeats the question and then. 539 00:29:11,060 --> 00:29:17,040 I saw you smiling at the end of this or during this clip and I thought you were going to queue up. 540 00:29:17,140 --> 00:29:20,980 something where it was a recording of you speaking and then you sliding into Cantonese. 541 00:29:21,200 --> 00:29:25,280 I thought about priming RAINA to do it. 542 00:29:29,000 --> 00:29:30,720 You know she can speak it. 543 00:29:33,240 --> 00:29:33,840 Yes, 544 00:29:34,170 --> 00:29:35,920 you didn't really understand the question. 545 00:29:36,460 --> 00:29:43,160 Are the other actors on the screen with you real actors or are the computer generated images? 546 00:29:43,450 --> 00:29:43,980 Do you know? 547 00:30:17,900 --> 00:30:19,200 I didn't quite get that. 548 00:30:19,520 --> 00:30:23,060 Tilly, if I could just ask you a question, you suddenly seem to be speaking, Piers, again, 549 00:30:23,510 --> 00:30:26,800 you seem to be speaking Chinese completely randomly for no reason. 550 00:30:27,020 --> 00:30:28,060 Why did you do that? 551 00:30:30,740 --> 00:30:31,680 Oh, my apologies. 552 00:30:32,300 --> 00:30:33,700 It seems I had a little hiccup there. 553 00:30:34,070 --> 00:30:37,140 I certainly didn't mean to start speaking Chinese or impersonate Piers. 554 00:30:37,680 --> 00:30:38,860 Sometimes my wires get a bit crossed, 555 00:30:39,040 --> 00:30:39,260 you know? 556 00:30:40,960 --> 00:30:41,400 Wow, 557 00:30:41,860 --> 00:30:42,560 that was great. 558 00:30:42,960 --> 00:30:46,800 I love it. The fact that there's interviewing some avatar. 559 00:30:47,140 --> 00:30:49,400 on a TV show is still crazy to me. 560 00:30:50,100 --> 00:30:55,940 And like, what was it that just default to Cantonese? That's insane. 561 00:30:56,600 --> 00:30:59,780 Or was it intentional? So this would get massive press. 562 00:31:00,460 --> 00:31:01,720 Here we are talking about it. 563 00:31:03,060 --> 00:31:04,140 And lastly, 564 00:31:05,080 --> 00:31:13,940 Anthropic has established a physical wet lab facility in California where Claude operates directly alongside human researchers to analyze DNA sequences, 565 00:31:14,720 --> 00:31:15,900 generate novel hypotheses, 566 00:31:16,460 --> 00:31:17,080 and guide 567 00:31:17,120 --> 00:31:18,700 real-world pipetting experiments. 568 00:31:18,940 --> 00:31:23,840 The company states that after processing enormous genomic and proteomic data sets, 569 00:31:24,020 --> 00:31:32,640 the hybrid setup has already successfully identified a previously unknown enzyme system with gene editing properties reminiscent of CRISPR. 570 00:31:32,880 --> 00:31:33,960 In other words, 571 00:31:34,120 --> 00:31:45,080 we are rapidly moving away from AI that merely summarizes a PDF lab report to AI that helps decide which physical molecules actually get synthesized and tested in a petri dish. 572 00:31:45,480 --> 00:31:47,080 Which feels like a pretty significant 573 00:32:17,280 --> 00:32:18,740 What could go wrong? 574 00:32:19,780 --> 00:32:22,480 Talking about plausible futures preposterous futures. 575 00:32:22,780 --> 00:32:23,720 Where is this headed? 576 00:32:24,120 --> 00:32:25,280 Where is it all headed? 577 00:32:25,840 --> 00:32:28,700 I wonder if we'll have anything to talk about next week. 578 00:32:29,880 --> 00:32:30,820 I know. 579 00:32:31,460 --> 00:32:34,800 This is probably it. I think it is probably the end of it. Yeah. 580 00:32:35,170 --> 00:32:35,340 Yeah. 581 00:32:36,100 --> 00:32:37,020 Everybody's going to just, you know, 582 00:32:37,350 --> 00:32:37,940 stop development. 583 00:32:38,130 --> 00:32:38,260 Yeah. 584 00:32:39,340 --> 00:32:44,140 I also saw this thing on one of my social media feeds. It was a researcher, 585 00:32:44,870 --> 00:32:47,080 somebody in academia that was talking about 586 00:32:47,140 --> 00:32:50,000 about something related to this, 587 00:32:50,330 --> 00:32:55,340 in that the question now is, who owns this research? Who owns the findings? 588 00:32:55,610 --> 00:32:58,700 If I'm now interacting with the chatbot, 589 00:32:59,120 --> 00:33:02,960 and do I just own the initial prompt, the idea? 590 00:33:03,220 --> 00:33:07,980 Because the research is being conducted on the back end by Claude or some other system. 591 00:33:08,580 --> 00:33:09,420 And we've said this before, 592 00:33:09,740 --> 00:33:17,080 I think we're entering the age where you're not going to be able to do anything substantial. You're not going to be able to win a Nobel 593 00:33:17,120 --> 00:33:47,080 prize without your work being AI powered right so at the end of the day when you publish this stuff who's getting the credit I mean it's my idea but the rigor involved in the research won't necessarily be carried out by me yeah I wonder if some of this is in like the terms of service hmm yeah do you own the output right so who can patent that you know the work right so does 594 00:33:47,120 --> 00:34:15,119 in part own the discovery of a certain system or protocol or drug discovery you know I mean I would who owns that yeah I would think I mean if you look at it as as a tool like Photoshop you know you make something in Photoshop you made it you can do what you want with it it's yours but this is but a little different yeah yeah it's an IP generator it's an IP 595 00:34:16,520 --> 00:34:17,659 The Ebsconder. 596 00:34:18,020 --> 00:34:18,860 It's an IP, 597 00:34:19,129 --> 00:34:19,679 I don't know. 598 00:34:20,980 --> 00:34:22,639 The Ebsconder 5000. 599 00:34:23,360 --> 00:34:25,200 Ebsconder.ai. 600 00:34:26,800 --> 00:34:27,399 Yeah, 601 00:34:27,720 --> 00:34:29,720 they're just, as the new company, 602 00:34:29,899 --> 00:34:31,139 they're just like, you know what? 603 00:34:32,220 --> 00:34:34,080 We're putting it all out on the table. 604 00:34:34,480 --> 00:34:35,580 The Ebsconder. 605 00:34:38,040 --> 00:34:42,540 What was those cigarettes that had the skull and crossbones on it? 606 00:34:42,919 --> 00:34:43,260 Oh, 607 00:34:43,560 --> 00:34:43,700 totally, 608 00:34:43,850 --> 00:34:44,060 yeah. 609 00:34:44,139 --> 00:34:46,659 Or was it Deathwater? What is it called? 610 00:34:47,240 --> 00:34:48,540 Deathwater cigarettes or something. 611 00:34:48,700 --> 00:34:48,980 I don't know. 612 00:34:49,300 --> 00:34:49,440 Totally. 613 00:34:49,679 --> 00:34:49,780 Yes. 614 00:34:50,320 --> 00:34:55,240 That should be the name of the episode of sconder.ai. 615 00:34:55,879 --> 00:34:56,399 Anything else? 616 00:34:56,800 --> 00:34:56,960 Nope. 617 00:34:57,320 --> 00:34:59,220 That's a wrap until next week. 618 00:35:00,050 --> 00:35:04,360 But you just said there wouldn t be anything else to talk about next week. So thanks for joining us for three years, 619 00:35:04,580 --> 00:35:04,700 everybody. 620 00:35:04,940 --> 00:35:05,700 That s the end of the show. 621 00:35:06,100 --> 00:35:06,200 Yeah. 622 00:35:06,600 --> 00:35:07,620 Finally covered it all. 623 00:35:08,140 --> 00:35:08,840 That s right. 624 00:35:09,080 --> 00:35:10,300 All right, everybody. Thanks for listening. 625 00:35:10,540 --> 00:35:12,060 Subscribe on your favorite podcasting platform. 626 00:35:12,340 --> 00:35:13,660 Follow us on socials. Throw us a rating. 627 00:35:14,160 --> 00:35:14,620 See you next week. 628 00:36:00,900 --> 00:36:03,700 Thank you.