1 00:00:00,000 --> 00:00:03,720 Ever wonder how AI might impact not just your health, 2 00:00:03,720 --> 00:00:04,920 but like your wallet too. 3 00:00:04,920 --> 00:00:06,240 Oh yeah, for sure. 4 00:00:06,240 --> 00:00:08,880 We're diving deep today into this paper. 5 00:00:08,880 --> 00:00:13,520 The fiscal frontier, projecting AI's long-term impact 6 00:00:13,520 --> 00:00:15,600 on the US fiscal outlook. 7 00:00:15,600 --> 00:00:17,860 It's from the Brookings Institution. 8 00:00:17,860 --> 00:00:22,040 And this paper tackles the pretty fascinating question 9 00:00:22,040 --> 00:00:27,040 of how AI in healthcare could like totally revolutionize 10 00:00:28,060 --> 00:00:29,320 the US federal budget. 11 00:00:29,320 --> 00:00:31,040 Yeah, it's interesting because we usually hear about 12 00:00:31,040 --> 00:00:33,600 like AI doing things like, you know, 13 00:00:33,600 --> 00:00:38,160 beating chess grandmasters or powering self-driving cars. 14 00:00:38,160 --> 00:00:41,480 But this paper focuses on something way more tangible. 15 00:00:41,480 --> 00:00:44,080 Like how AI could actually change healthcare 16 00:00:44,080 --> 00:00:45,360 and the government's bottom line. 17 00:00:45,360 --> 00:00:48,240 So okay, so less about robots taking over the world. 18 00:00:48,240 --> 00:00:49,800 More about robots taking our temperatures. 19 00:00:49,800 --> 00:00:52,800 Exactly, this paper walks us through how AI 20 00:00:52,800 --> 00:00:54,920 is already changing healthcare. 21 00:00:54,920 --> 00:00:57,320 Written by a group of economic experts by the way. 22 00:00:57,320 --> 00:00:58,160 Oh cool. 23 00:00:58,160 --> 00:01:00,160 That lays out some pretty thought-provoking scenarios 24 00:01:00,160 --> 00:01:01,000 for the future. 25 00:01:01,000 --> 00:01:02,520 So before we get to the sci-fi stuff, 26 00:01:02,520 --> 00:01:04,120 let's talk about what's happening right now. 27 00:01:04,120 --> 00:01:04,960 Okay, sure. 28 00:01:04,960 --> 00:01:07,480 What are some examples of AI 29 00:01:07,480 --> 00:01:10,160 already making waves in healthcare? 30 00:01:10,160 --> 00:01:12,280 Well, one example is AlphaFold 2. 31 00:01:12,280 --> 00:01:13,120 Okay. 32 00:01:13,120 --> 00:01:15,640 This AI system is revolutionizing 33 00:01:15,640 --> 00:01:18,280 how we predict protein structures. 34 00:01:18,280 --> 00:01:19,120 Wow. 35 00:01:19,120 --> 00:01:23,180 So imagine like being able to design new drugs way faster 36 00:01:23,180 --> 00:01:26,840 because an AI can figure out how proteins fold. 37 00:01:26,840 --> 00:01:28,480 That's like a game changer 38 00:01:28,480 --> 00:01:30,360 for treating all sorts of diseases. 39 00:01:30,360 --> 00:01:31,200 That's incredible. 40 00:01:31,200 --> 00:01:32,040 It sounds like something straight 41 00:01:32,040 --> 00:01:33,080 out of a science fiction movie. 42 00:01:33,080 --> 00:01:34,360 It really does. 43 00:01:34,360 --> 00:01:39,080 And then there's AME, Google's large language model. 44 00:01:39,080 --> 00:01:42,120 This AI is essentially a super smart chat bot 45 00:01:42,120 --> 00:01:42,960 for healthcare. 46 00:01:42,960 --> 00:01:43,800 Oh wow. 47 00:01:43,800 --> 00:01:47,600 And get this, it's actually outperforming human doctors. 48 00:01:47,600 --> 00:01:48,440 What? 49 00:01:48,440 --> 00:01:50,560 In diagnosing some complex medical cases. 50 00:01:50,560 --> 00:01:51,400 Wait, hold on. 51 00:01:51,400 --> 00:01:52,240 Yeah. 52 00:01:52,240 --> 00:01:54,200 AI that's better at diagnosing than human doctors. 53 00:01:54,200 --> 00:01:55,040 Yeah. 54 00:01:55,040 --> 00:01:56,160 That's a bit unsettling, isn't it? 55 00:01:56,160 --> 00:01:58,360 It definitely raises some questions. 56 00:01:58,360 --> 00:02:01,080 But it also highlights the potential for AI 57 00:02:01,080 --> 00:02:03,520 to improve diagnostic accuracy. 58 00:02:03,520 --> 00:02:05,280 Potentially reducing errors, 59 00:02:05,280 --> 00:02:08,120 which as you know is a huge issue in healthcare. 60 00:02:08,120 --> 00:02:08,960 Right. 61 00:02:08,960 --> 00:02:12,000 Fewer misdiagnoses would mean like better outcomes 62 00:02:12,000 --> 00:02:12,840 for patients. 63 00:02:12,840 --> 00:02:13,680 Exactly. 64 00:02:13,680 --> 00:02:14,520 And lower costs. 65 00:02:14,520 --> 00:02:15,360 Yeah. 66 00:02:15,360 --> 00:02:16,520 Are there any other areas where AI 67 00:02:16,520 --> 00:02:17,880 is already making a difference? 68 00:02:17,880 --> 00:02:18,720 Absolutely. 69 00:02:18,720 --> 00:02:22,520 AI algorithms are already being used to detect diseases. 70 00:02:22,520 --> 00:02:23,360 Okay. 71 00:02:23,360 --> 00:02:25,680 Like diabetic retinopathy earlier. 72 00:02:25,680 --> 00:02:28,200 And more accurately than traditional methods. 73 00:02:28,200 --> 00:02:29,040 Okay. 74 00:02:29,040 --> 00:02:30,520 And early detection is key. 75 00:02:30,520 --> 00:02:31,360 Yeah. 76 00:02:31,360 --> 00:02:33,640 For effective treatment and preventing long term 77 00:02:33,640 --> 00:02:34,760 complications. 78 00:02:34,760 --> 00:02:37,440 So AI is like a super powered early warning system 79 00:02:37,440 --> 00:02:38,280 for our health. 80 00:02:38,280 --> 00:02:39,120 Exactly. 81 00:02:39,120 --> 00:02:41,160 And think about the potential for AI 82 00:02:41,160 --> 00:02:43,080 to help prevent future pandemics. 83 00:02:43,080 --> 00:02:43,920 Oh wow. 84 00:02:43,920 --> 00:02:47,320 By analyzing data and identifying patterns, 85 00:02:47,320 --> 00:02:50,480 AI could alert us to potential outbreaks sooner. 86 00:02:50,480 --> 00:02:51,320 Okay. 87 00:02:51,320 --> 00:02:52,880 And help us allocate resources more effectively. 88 00:02:52,880 --> 00:02:54,240 That's pretty amazing. 89 00:02:54,240 --> 00:02:56,160 It sounds like AI could really be a game changer 90 00:02:56,160 --> 00:02:57,000 in public health. 91 00:02:57,000 --> 00:02:57,840 Yeah. 92 00:02:57,840 --> 00:02:59,680 But this paper isn't just about the cool things 93 00:02:59,680 --> 00:03:00,640 AI can do right. 94 00:03:00,640 --> 00:03:01,480 Right. 95 00:03:01,480 --> 00:03:02,680 It's also about the economic impact. 96 00:03:02,680 --> 00:03:03,520 Right. 97 00:03:03,520 --> 00:03:06,480 The paper then shifts gears to look at how all these AI 98 00:03:06,480 --> 00:03:09,440 advancements could impact the US budget. 99 00:03:09,440 --> 00:03:11,800 They focus on four key areas. 100 00:03:11,800 --> 00:03:12,640 Right. 101 00:03:12,640 --> 00:03:13,480 Mortality rates. 102 00:03:13,480 --> 00:03:14,320 Mm-hmm. 103 00:03:14,320 --> 00:03:15,160 Healthcare demand. 104 00:03:15,160 --> 00:03:16,000 Okay. 105 00:03:16,000 --> 00:03:16,840 Healthcare prices. 106 00:03:16,840 --> 00:03:17,680 Mm-hmm. 107 00:03:17,680 --> 00:03:18,520 And overall productivity. 108 00:03:18,520 --> 00:03:19,360 Okay. 109 00:03:19,360 --> 00:03:20,200 Let's break those down first. 110 00:03:20,200 --> 00:03:24,160 Mortality rates is AI gonna make us all live forever. 111 00:03:24,160 --> 00:03:26,200 Well, maybe not forever. 112 00:03:26,200 --> 00:03:27,040 Okay. 113 00:03:27,040 --> 00:03:29,840 But it could significantly extend lifespans. 114 00:03:29,840 --> 00:03:30,680 Okay. 115 00:03:30,680 --> 00:03:32,760 And that has huge implications for programs 116 00:03:32,760 --> 00:03:34,800 like social security and Medicare. 117 00:03:34,800 --> 00:03:35,640 Right. 118 00:03:35,640 --> 00:03:37,000 If people are living longer, 119 00:03:37,000 --> 00:03:39,600 they'll be collecting benefits for a longer period of time. 120 00:03:39,600 --> 00:03:40,440 Exactly. 121 00:03:40,440 --> 00:03:41,400 Which could put a strain on the budget. 122 00:03:41,400 --> 00:03:42,240 Exactly. 123 00:03:42,240 --> 00:03:44,840 Then there's a question of healthcare demand and utilization. 124 00:03:44,840 --> 00:03:45,680 Okay. 125 00:03:45,680 --> 00:03:49,120 Will AI lead to more efficient healthcare, 126 00:03:49,120 --> 00:03:51,840 reducing the need for expensive interventions? 127 00:03:51,840 --> 00:03:52,680 Okay. 128 00:03:52,680 --> 00:03:56,280 Or will its increased efficacy actually drive up demand 129 00:03:56,280 --> 00:03:57,320 for services? 130 00:03:57,320 --> 00:03:58,160 Oh, that's interesting. 131 00:03:58,160 --> 00:03:59,280 It's a complex equation. 132 00:03:59,280 --> 00:04:01,600 So it's possible that AI could make healthcare 133 00:04:01,600 --> 00:04:02,720 more affordable. 134 00:04:02,720 --> 00:04:03,560 Right. 135 00:04:03,560 --> 00:04:05,880 But it could also make it so good that everyone wants it. 136 00:04:05,880 --> 00:04:06,720 Right. 137 00:04:06,720 --> 00:04:08,360 Which could end up costing more. 138 00:04:08,360 --> 00:04:09,200 Exactly. 139 00:04:09,200 --> 00:04:10,200 That's a tricky balance. 140 00:04:10,200 --> 00:04:11,040 Yeah. 141 00:04:11,040 --> 00:04:11,880 It is. 142 00:04:11,880 --> 00:04:13,560 And then there's the question of healthcare prices themselves. 143 00:04:13,560 --> 00:04:14,400 Okay. 144 00:04:14,400 --> 00:04:18,320 Will AI-driven efficiencies translate into lower costs 145 00:04:18,320 --> 00:04:20,080 for individuals and the government? 146 00:04:20,080 --> 00:04:20,920 Yeah. 147 00:04:20,920 --> 00:04:23,240 Or will other factors offset those savings? 148 00:04:23,240 --> 00:04:24,080 Right. 149 00:04:24,080 --> 00:04:25,280 Like the cost of developing and implementing 150 00:04:25,280 --> 00:04:26,400 all these AI technologies. 151 00:04:26,400 --> 00:04:27,240 Exactly. 152 00:04:27,240 --> 00:04:29,280 And finally, they consider the impact of AI 153 00:04:29,280 --> 00:04:30,640 on overall productivity. 154 00:04:30,640 --> 00:04:31,480 Okay. 155 00:04:31,480 --> 00:04:33,840 Will it boost the economy leading to higher tax revenues? 156 00:04:33,840 --> 00:04:34,680 Yeah. 157 00:04:34,680 --> 00:04:36,400 Or will there be job displacement 158 00:04:36,400 --> 00:04:38,320 and other economic consequences? 159 00:04:38,320 --> 00:04:39,160 Right. 160 00:04:39,160 --> 00:04:40,520 It's a multifaceted issue. 161 00:04:40,520 --> 00:04:41,360 Mm-hmm. 162 00:04:41,360 --> 00:04:44,120 With both potential positive and negative impacts. 163 00:04:44,120 --> 00:04:48,360 So it's not just about how AI will affect healthcare costs. 164 00:04:48,360 --> 00:04:49,200 Right. 165 00:04:49,200 --> 00:04:50,640 But how it will affect the entire economy. 166 00:04:50,640 --> 00:04:51,480 Yeah. 167 00:04:51,480 --> 00:04:52,320 That's a big picture view. 168 00:04:52,320 --> 00:04:53,160 It is. 169 00:04:53,160 --> 00:04:55,000 And to illustrate these potential impacts. 170 00:04:55,000 --> 00:04:55,840 Okay. 171 00:04:55,840 --> 00:04:58,320 The authors have created four distinct scenarios. 172 00:04:58,320 --> 00:04:59,160 Okay. 173 00:04:59,160 --> 00:05:01,440 Each with different assumptions about how AI evolves 174 00:05:01,440 --> 00:05:02,640 and is implemented. 175 00:05:02,640 --> 00:05:05,800 So they're basically saying here are four possible futures. 176 00:05:05,800 --> 00:05:06,640 Uh-huh. 177 00:05:06,640 --> 00:05:08,280 Depending on how things play out with AI. 178 00:05:08,280 --> 00:05:09,120 Exactly. 179 00:05:09,120 --> 00:05:10,840 And by comparing these scenarios, 180 00:05:10,840 --> 00:05:14,680 we can get a better sense of the potential range of outcomes. 181 00:05:14,680 --> 00:05:15,680 That's really helpful. 182 00:05:15,680 --> 00:05:18,040 Okay, so let's dive into these scenarios. 183 00:05:18,040 --> 00:05:18,880 Okay. 184 00:05:18,880 --> 00:05:21,120 What does the first scenario look like? 185 00:05:21,120 --> 00:05:24,080 The first scenario focuses on a modest reduction 186 00:05:24,080 --> 00:05:25,240 in mortality rates. 187 00:05:25,240 --> 00:05:26,080 Okay. 188 00:05:26,080 --> 00:05:28,680 Without any major changes to other factors. 189 00:05:28,680 --> 00:05:29,520 Right. 190 00:05:29,520 --> 00:05:30,560 It's sort of a baseline scenario 191 00:05:30,560 --> 00:05:32,320 to compare the others against. 192 00:05:32,320 --> 00:05:34,280 Okay, so a slightly longer lifespan. 193 00:05:34,280 --> 00:05:35,120 Yeah. 194 00:05:35,120 --> 00:05:37,360 But no major disruptions to the healthcare system. 195 00:05:37,360 --> 00:05:38,200 Right. 196 00:05:38,200 --> 00:05:39,040 Or the economy. 197 00:05:39,040 --> 00:05:39,880 Exactly. 198 00:05:39,880 --> 00:05:40,720 What about scenario two? 199 00:05:40,720 --> 00:05:43,680 Scenario two assumes modest efficiency gains 200 00:05:43,680 --> 00:05:44,880 in healthcare delivery. 201 00:05:44,880 --> 00:05:45,720 Okay. 202 00:05:45,720 --> 00:05:49,080 To slightly slower growth in utilization and prices. 203 00:05:49,080 --> 00:05:49,920 Mm-hmm. 204 00:05:49,920 --> 00:05:53,440 Along with higher tax revenues from productivity gains. 205 00:05:53,440 --> 00:05:54,560 So in this scenario, 206 00:05:54,560 --> 00:05:57,720 AI starts to make healthcare a bit more efficient. 207 00:05:57,720 --> 00:05:58,560 Right. 208 00:05:58,560 --> 00:06:00,800 Which helps control costs and boost the economy. 209 00:06:00,800 --> 00:06:01,640 Right. 210 00:06:01,640 --> 00:06:03,560 Scenario three takes those efficiency gains 211 00:06:03,560 --> 00:06:04,400 a step further. 212 00:06:04,400 --> 00:06:05,240 Okay. 213 00:06:05,240 --> 00:06:06,640 Imagining major improvements in care delivery. 214 00:06:06,640 --> 00:06:07,480 Mm-hmm. 215 00:06:07,480 --> 00:06:08,960 Even slower growth in utilization 216 00:06:08,960 --> 00:06:10,120 and healthcare inflation. 217 00:06:10,120 --> 00:06:10,960 Okay. 218 00:06:10,960 --> 00:06:12,120 And even higher tax revenue. 219 00:06:12,120 --> 00:06:14,040 So this is the optimistic scenario 220 00:06:14,040 --> 00:06:16,920 where AI really starts to deliver on its promises. 221 00:06:16,920 --> 00:06:17,760 Exactly. 222 00:06:17,760 --> 00:06:19,600 But then there's scenario four, 223 00:06:19,600 --> 00:06:21,840 which is where things get a little more complicated. 224 00:06:21,840 --> 00:06:23,760 Uh-oh, what happens in scenario four? 225 00:06:23,760 --> 00:06:26,680 It also assumes major improvements in care delivery. 226 00:06:26,680 --> 00:06:27,520 Okay. 227 00:06:27,520 --> 00:06:29,880 But it predicts that increased utilization. 228 00:06:29,880 --> 00:06:30,720 Okay. 229 00:06:30,720 --> 00:06:32,400 Meaning more people seeking care. 230 00:06:32,400 --> 00:06:33,240 Mm-hmm. 231 00:06:33,240 --> 00:06:35,120 Due to its increased effectiveness 232 00:06:35,120 --> 00:06:37,800 will actually offset those price decreases. 233 00:06:37,800 --> 00:06:38,640 Oh, wow. 234 00:06:38,640 --> 00:06:42,080 This scenario still factors in those higher tax revenues. 235 00:06:42,080 --> 00:06:42,920 Okay. 236 00:06:42,920 --> 00:06:44,120 The increased demand for care 237 00:06:44,120 --> 00:06:45,680 throws a wrench in the equation. 238 00:06:45,680 --> 00:06:48,880 So even if AI makes healthcare more efficient and effective. 239 00:06:48,880 --> 00:06:49,720 Right. 240 00:06:49,720 --> 00:06:51,760 It might also make it more desirable. 241 00:06:51,760 --> 00:06:52,600 Exactly. 242 00:06:52,600 --> 00:06:54,200 Which could lead to higher overall costs. 243 00:06:54,200 --> 00:06:55,040 Exactly. 244 00:06:55,040 --> 00:06:56,280 It highlights the delicate balance 245 00:06:56,280 --> 00:06:58,120 between increased efficacy. 246 00:06:58,120 --> 00:06:58,960 Yeah. 247 00:06:58,960 --> 00:07:00,280 And potentially higher costs. 248 00:07:00,280 --> 00:07:01,120 Right. 249 00:07:01,120 --> 00:07:03,480 And it shows that the impact of AI on the US budget 250 00:07:03,480 --> 00:07:05,160 is not a simple equation. 251 00:07:05,160 --> 00:07:06,000 No. 252 00:07:06,000 --> 00:07:06,840 A lot of moving parts. 253 00:07:06,840 --> 00:07:07,680 Absolutely. 254 00:07:07,680 --> 00:07:10,080 And that's why the authors of this paper 255 00:07:10,080 --> 00:07:12,640 go to great lengths to explain their methodology. 256 00:07:12,640 --> 00:07:14,360 And the assumptions they've made, 257 00:07:14,360 --> 00:07:16,760 they want to be transparent about the uncertainties. 258 00:07:16,760 --> 00:07:17,600 Yeah. 259 00:07:17,600 --> 00:07:19,560 Involved in predicting AI's impact. 260 00:07:19,560 --> 00:07:20,400 That makes sense. 261 00:07:20,400 --> 00:07:22,960 They're not claiming to have all the answers. 262 00:07:22,960 --> 00:07:23,800 Right. 263 00:07:23,800 --> 00:07:25,240 But they're providing a framework 264 00:07:25,240 --> 00:07:27,160 for thinking about these complex issues. 265 00:07:27,160 --> 00:07:28,000 Exactly. 266 00:07:28,000 --> 00:07:30,920 You're saying here's how we approach this problem. 267 00:07:30,920 --> 00:07:33,200 Here are our assumptions. 268 00:07:33,200 --> 00:07:35,600 And here are some possible outcomes. 269 00:07:35,600 --> 00:07:36,440 Okay. 270 00:07:36,440 --> 00:07:37,520 So we've got these four scenarios. 271 00:07:37,520 --> 00:07:38,360 Yeah. 272 00:07:38,360 --> 00:07:41,600 Each with different assumptions about how AI will affect 273 00:07:41,600 --> 00:07:43,000 healthcare and the economy. 274 00:07:43,000 --> 00:07:43,840 Right. 275 00:07:43,840 --> 00:07:46,760 But how do they actually measure the impact 276 00:07:46,760 --> 00:07:47,880 on the US budget? 277 00:07:47,880 --> 00:07:48,720 Yeah. 278 00:07:48,720 --> 00:07:49,560 What are they looking at? 279 00:07:49,560 --> 00:07:51,040 They're looking at a bunch of factors. 280 00:07:51,040 --> 00:07:51,880 Okay. 281 00:07:51,880 --> 00:07:53,560 But one of the most important is Medicare spending. 282 00:07:53,560 --> 00:07:54,400 Okay. 283 00:07:54,400 --> 00:07:56,240 It's a huge part of the federal budget. 284 00:07:56,240 --> 00:08:00,160 And it's likely to be like significantly impacted by AI. 285 00:08:00,160 --> 00:08:01,000 That makes sense. 286 00:08:01,000 --> 00:08:05,160 So how do they go about projecting how AI will affect 287 00:08:05,160 --> 00:08:06,400 Medicare spending? 288 00:08:06,400 --> 00:08:07,240 Right. 289 00:08:07,240 --> 00:08:09,280 It seems like a pretty tricky thing to predict. 290 00:08:09,280 --> 00:08:10,120 It is. 291 00:08:10,120 --> 00:08:11,960 They're very upfront about the challenges. 292 00:08:11,960 --> 00:08:15,280 But they've developed a pretty clever approach. 293 00:08:15,280 --> 00:08:17,800 They start by breaking down Medicare spending 294 00:08:17,800 --> 00:08:21,080 into changes in price and utilization. 295 00:08:21,080 --> 00:08:21,920 Okay. 296 00:08:21,920 --> 00:08:23,640 So they're trying to separate out how much of the change 297 00:08:23,640 --> 00:08:27,360 in spending is due to lower prices for healthcare services. 298 00:08:27,360 --> 00:08:28,200 Exactly. 299 00:08:28,200 --> 00:08:30,600 And how much is due to fewer people needing those services. 300 00:08:30,600 --> 00:08:31,440 Exactly. 301 00:08:31,440 --> 00:08:33,440 And to get a handle on these factors. 302 00:08:33,440 --> 00:08:34,280 Okay. 303 00:08:34,280 --> 00:08:37,600 They look back at historical trends and events 304 00:08:37,600 --> 00:08:40,760 to help estimate potential future impacts. 305 00:08:40,760 --> 00:08:43,600 So they're using the past to try to predict the future. 306 00:08:43,600 --> 00:08:44,440 Right. 307 00:08:44,440 --> 00:08:46,320 What kind of historical events are they looking at? 308 00:08:46,320 --> 00:08:49,440 For example, they look at Japan's dramatic decline 309 00:08:49,440 --> 00:08:50,440 in death rates. 310 00:08:50,440 --> 00:08:51,280 Okay. 311 00:08:51,280 --> 00:08:52,600 Post World War II. 312 00:08:52,600 --> 00:08:56,720 To help model how AI might impact longevity. 313 00:08:56,720 --> 00:08:57,560 Okay. 314 00:08:57,560 --> 00:08:59,480 It's a clever way to ground their projections 315 00:08:59,480 --> 00:09:00,920 in real world examples. 316 00:09:00,920 --> 00:09:04,000 So instead of just speculating about how AI might extend 317 00:09:04,000 --> 00:09:04,840 life spans. 318 00:09:04,840 --> 00:09:05,680 Right. 319 00:09:05,680 --> 00:09:07,560 They're looking at a case where something similar happened 320 00:09:07,560 --> 00:09:08,400 in the past. 321 00:09:08,400 --> 00:09:09,240 Exactly. 322 00:09:09,240 --> 00:09:10,080 That's a smart approach. 323 00:09:10,080 --> 00:09:13,080 It is, and they use a similar approach for other factors. 324 00:09:13,080 --> 00:09:13,920 Okay. 325 00:09:13,920 --> 00:09:17,560 For instance, to estimate how AI might impact healthcare prices. 326 00:09:17,560 --> 00:09:20,320 They look at events like the Affordable Care Act. 327 00:09:20,320 --> 00:09:21,160 Okay. 328 00:09:21,160 --> 00:09:23,400 Which had a significant impact on healthcare costs. 329 00:09:23,400 --> 00:09:25,880 So they're trying to identify historical parallels. 330 00:09:25,880 --> 00:09:26,720 Right. 331 00:09:26,720 --> 00:09:27,800 To give their projections more weight. 332 00:09:27,800 --> 00:09:28,640 Exactly. 333 00:09:28,640 --> 00:09:30,400 So they're saying, okay, we can't predict the future 334 00:09:30,400 --> 00:09:31,240 perfectly. 335 00:09:31,240 --> 00:09:32,080 Right. 336 00:09:32,080 --> 00:09:34,760 But we can learn from the past and use that knowledge 337 00:09:34,760 --> 00:09:36,560 to make some educated guesses. 338 00:09:36,560 --> 00:09:37,400 Exactly. 339 00:09:37,400 --> 00:09:40,160 And they're also transparent about the limitations 340 00:09:40,160 --> 00:09:41,200 of their model. 341 00:09:41,200 --> 00:09:42,040 Okay. 342 00:09:42,040 --> 00:09:45,800 They emphasize that these are exploratory simulations 343 00:09:45,800 --> 00:09:49,120 designed to illustrate a range of plausible outcomes. 344 00:09:49,120 --> 00:09:49,960 Right. 345 00:09:49,960 --> 00:09:51,600 Not to provide definitive answers. 346 00:09:51,600 --> 00:09:54,560 Which is important because AI is such a rapidly evolving 347 00:09:54,560 --> 00:09:55,400 field. 348 00:09:55,400 --> 00:09:56,240 Right. 349 00:09:56,240 --> 00:09:57,800 Even the most sophisticated models can't capture 350 00:09:57,800 --> 00:09:58,840 every nuance. 351 00:09:58,840 --> 00:09:59,680 Right. 352 00:09:59,680 --> 00:10:01,360 And then there's the question of productivity. 353 00:10:01,360 --> 00:10:02,200 Okay. 354 00:10:02,200 --> 00:10:04,520 How do they factor in the potential impact of AI? 355 00:10:04,520 --> 00:10:05,360 Yeah. 356 00:10:05,360 --> 00:10:07,040 On overall economic output. 357 00:10:07,040 --> 00:10:07,880 Yeah. 358 00:10:07,880 --> 00:10:09,600 That seems like a crucial piece of the puzzle. 359 00:10:09,600 --> 00:10:10,440 Right. 360 00:10:10,440 --> 00:10:14,040 If AI can boost productivity and lead to higher tax revenues. 361 00:10:14,040 --> 00:10:16,320 That could offset some of the potential increases 362 00:10:16,320 --> 00:10:17,720 in healthcare spending. 363 00:10:17,720 --> 00:10:18,560 Exactly. 364 00:10:18,560 --> 00:10:22,240 And they use historical data from the 1990s productivity 365 00:10:22,240 --> 00:10:23,080 boom. 366 00:10:23,080 --> 00:10:23,920 Okay. 367 00:10:23,920 --> 00:10:26,480 Which was fueled by the adoption of the internet. 368 00:10:26,480 --> 00:10:31,480 To estimate how AI might similarly boost economic output. 369 00:10:31,760 --> 00:10:35,000 So they're looking at a time when technology led to a surge 370 00:10:35,000 --> 00:10:35,960 in productivity. 371 00:10:35,960 --> 00:10:36,800 Right. 372 00:10:36,800 --> 00:10:39,520 And using that as a model for how AI might do the same. 373 00:10:39,520 --> 00:10:40,360 Precisely. 374 00:10:40,360 --> 00:10:41,400 They're essentially asking. 375 00:10:41,400 --> 00:10:42,240 Okay. 376 00:10:42,240 --> 00:10:44,840 What if AI has a similar impact on the economy 377 00:10:44,840 --> 00:10:45,960 as the internet did? 378 00:10:45,960 --> 00:10:47,360 That's a fascinating question. 379 00:10:47,360 --> 00:10:48,200 Yeah. 380 00:10:48,200 --> 00:10:50,640 And it seems like the answer could have a huge impact 381 00:10:50,640 --> 00:10:51,920 on the US budget outlook. 382 00:10:51,920 --> 00:10:53,000 It absolutely could. 383 00:10:53,000 --> 00:10:55,480 And to explore those potential impacts. 384 00:10:55,480 --> 00:10:57,160 They combine these variables. 385 00:10:57,160 --> 00:10:58,480 Mortality rates. 386 00:10:58,480 --> 00:10:59,960 Healthcare prices. 387 00:10:59,960 --> 00:11:01,760 Healthcare utilization. 388 00:11:01,760 --> 00:11:03,520 And overall productivity. 389 00:11:03,520 --> 00:11:04,360 Right. 390 00:11:04,360 --> 00:11:05,880 Into those four distinct scenarios we talked about 391 00:11:05,880 --> 00:11:06,720 earlier. 392 00:11:06,720 --> 00:11:09,400 Four possible futures depending on how things play out 393 00:11:09,400 --> 00:11:10,240 with AI. 394 00:11:10,240 --> 00:11:11,080 Exactly. 395 00:11:11,080 --> 00:11:13,640 And the results of their simulations are well, 396 00:11:13,640 --> 00:11:14,720 pretty eye opening. 397 00:11:14,720 --> 00:11:15,560 Okay. 398 00:11:15,560 --> 00:11:16,400 I'm on the edge of my seat. 399 00:11:16,400 --> 00:11:17,240 What do they find? 400 00:11:17,240 --> 00:11:19,720 They found that depending on which scenario plays out, 401 00:11:19,720 --> 00:11:23,920 the impact on the US fiscal outlook could vary dramatically. 402 00:11:23,920 --> 00:11:24,760 Okay. 403 00:11:24,760 --> 00:11:25,600 Over the next two decades. 404 00:11:25,600 --> 00:11:26,960 Give me some specifics. 405 00:11:26,960 --> 00:11:28,480 What's the best case scenario? 406 00:11:28,480 --> 00:11:29,320 Okay. 407 00:11:29,320 --> 00:11:30,880 And what's the worst case scenario? 408 00:11:30,880 --> 00:11:33,080 Well, in the most optimistic scenario. 409 00:11:33,080 --> 00:11:33,920 Okay. 410 00:11:33,920 --> 00:11:34,760 Scenario three. 411 00:11:34,760 --> 00:11:37,680 Project major improvements in care delivery. 412 00:11:37,680 --> 00:11:38,520 Okay. 413 00:11:38,520 --> 00:11:41,800 And slower growth in utilization and prices. 414 00:11:41,800 --> 00:11:46,040 They estimate that annual budget deficits could be reduced. 415 00:11:46,040 --> 00:11:46,880 Okay. 416 00:11:46,880 --> 00:11:50,880 By as much as 1.5% of GDP by 2044. 417 00:11:50,880 --> 00:11:52,440 Wow, that's huge. 418 00:11:52,440 --> 00:11:55,640 Imagine the US erasing a sizable chunk of its debt. 419 00:11:55,640 --> 00:11:56,480 Right. 420 00:11:56,480 --> 00:11:58,240 Thanks to AI driven healthcare savings. 421 00:11:58,240 --> 00:11:59,080 That's incredible. 422 00:11:59,080 --> 00:12:01,800 That's the powerful potential upside this paper highlights. 423 00:12:01,800 --> 00:12:02,640 Right. 424 00:12:02,640 --> 00:12:03,520 It's like finding a trillion dollar bill 425 00:12:03,520 --> 00:12:05,040 just from smarter healthcare. 426 00:12:05,040 --> 00:12:07,240 It is a remarkable possibility. 427 00:12:07,240 --> 00:12:08,960 But then in scenario four. 428 00:12:08,960 --> 00:12:09,800 Okay. 429 00:12:09,800 --> 00:12:11,280 Where increased healthcare utilization 430 00:12:11,280 --> 00:12:13,200 offsets those price decreases. 431 00:12:13,200 --> 00:12:14,040 Yeah. 432 00:12:14,040 --> 00:12:15,320 Outcome is completely different. 433 00:12:15,320 --> 00:12:16,160 Oh no. 434 00:12:16,160 --> 00:12:18,320 They actually project that annual budget deficits 435 00:12:18,320 --> 00:12:21,480 could be nearly 1% of GDP higher 436 00:12:21,480 --> 00:12:23,960 than the baseline by 2044. 437 00:12:23,960 --> 00:12:25,520 So in the worst case scenario. 438 00:12:25,520 --> 00:12:26,360 Yeah. 439 00:12:26,360 --> 00:12:29,360 AI could actually make the US budget deficit worse. 440 00:12:29,360 --> 00:12:30,280 Right. 441 00:12:30,280 --> 00:12:31,640 That's a sobering thought. 442 00:12:31,640 --> 00:12:33,880 It is and it underscores the importance of making 443 00:12:33,880 --> 00:12:37,480 thoughtful choices about how we develop and implement AI. 444 00:12:37,480 --> 00:12:38,320 Right. 445 00:12:38,320 --> 00:12:39,160 It's not a magic bullet. 446 00:12:39,160 --> 00:12:40,000 Right. 447 00:12:40,000 --> 00:12:42,360 It's a powerful tool that could be used for good or for ill. 448 00:12:42,360 --> 00:12:43,200 Exactly. 449 00:12:43,200 --> 00:12:44,040 Depending on how we wield it. 450 00:12:44,040 --> 00:12:44,880 Exactly. 451 00:12:44,880 --> 00:12:46,680 And this paper really highlights the potential 452 00:12:46,680 --> 00:12:48,800 for both positive and negative impacts. 453 00:12:48,800 --> 00:12:49,640 Yeah. 454 00:12:49,640 --> 00:12:51,920 It's a reminder that we need to be proactive 455 00:12:51,920 --> 00:12:54,360 in shaping the future of AI. 456 00:12:54,360 --> 00:12:56,880 Rather than just letting it happen to us. 457 00:12:56,880 --> 00:12:59,400 So what are some of the key takeaways from this paper? 458 00:12:59,400 --> 00:13:00,240 Yeah. 459 00:13:00,240 --> 00:13:02,560 We'll boil it down to a few key points. 460 00:13:02,560 --> 00:13:03,120 OK. 461 00:13:03,120 --> 00:13:04,040 What would they be? 462 00:13:04,040 --> 00:13:05,880 I think one of the biggest takeaways 463 00:13:05,880 --> 00:13:08,760 is that AI's impact on the US fiscal outlook 464 00:13:08,760 --> 00:13:10,200 is far from certain. 465 00:13:10,200 --> 00:13:12,560 It really depends on how this technology evolves 466 00:13:12,560 --> 00:13:14,640 and how we choose to implement it. 467 00:13:14,640 --> 00:13:16,240 So it's not a foregone conclusion 468 00:13:16,240 --> 00:13:19,080 that AI will either save or bankrupt the country. 469 00:13:19,080 --> 00:13:19,960 Exactly. 470 00:13:19,960 --> 00:13:21,560 There are a lot of factors at play 471 00:13:21,560 --> 00:13:23,880 and the outcome is very much in our hands. 472 00:13:23,880 --> 00:13:26,480 That's both empowering and a bit daunting. 473 00:13:26,480 --> 00:13:27,080 I agree. 474 00:13:27,080 --> 00:13:30,120 Another key takeaway is that simply boosting productivity 475 00:13:30,120 --> 00:13:32,800 while helpful might not be enough to offset 476 00:13:32,800 --> 00:13:34,920 the potential increase in health care spending, 477 00:13:34,920 --> 00:13:37,920 especially if AI leads to longer lifespans. 478 00:13:37,920 --> 00:13:41,080 So even if AI makes us all a bit richer, 479 00:13:41,080 --> 00:13:43,640 we can still end up with a bigger budget deficit 480 00:13:43,640 --> 00:13:46,560 if health care costs continue to rise. 481 00:13:46,560 --> 00:13:48,160 That's the concern the authors raise. 482 00:13:48,160 --> 00:13:50,560 It's like they're saying, hey, don't get complacent. 483 00:13:50,560 --> 00:13:52,440 Even with higher tax revenues, we 484 00:13:52,440 --> 00:13:55,000 need to be smart about how we manage health care costs, 485 00:13:55,000 --> 00:13:56,880 especially as the population ages. 486 00:13:56,880 --> 00:13:58,440 So what's the solution? 487 00:13:58,440 --> 00:14:01,480 How do we make sure AI leads to a more sustainable 488 00:14:01,480 --> 00:14:02,960 fiscal future? 489 00:14:02,960 --> 00:14:06,000 The authors argue that the key lies in AI's ability 490 00:14:06,000 --> 00:14:08,120 to bend the cost curve for health care. 491 00:14:08,120 --> 00:14:11,040 In other words, AI needs to make health care more efficient 492 00:14:11,040 --> 00:14:12,880 and effective, not just more expensive. 493 00:14:12,880 --> 00:14:15,080 I've heard that phrase before, bend the cost curve, 494 00:14:15,080 --> 00:14:16,560 but I'm not sure I fully understand it. 495 00:14:16,560 --> 00:14:19,480 It basically means slowing down the rate at which health care 496 00:14:19,480 --> 00:14:20,640 costs increase. 497 00:14:20,640 --> 00:14:22,800 If we can do that, it would have a huge impact 498 00:14:22,800 --> 00:14:23,680 on the budget outlook. 499 00:14:23,680 --> 00:14:25,120 OK, that makes sense. 500 00:14:25,120 --> 00:14:28,800 So how can AI help us bend the cost curve? 501 00:14:28,800 --> 00:14:30,960 We've already talked about things like early detection 502 00:14:30,960 --> 00:14:32,840 and more accurate diagnoses. 503 00:14:32,840 --> 00:14:34,560 But are there other ways AI could 504 00:14:34,560 --> 00:14:36,000 make health care more efficient? 505 00:14:36,000 --> 00:14:36,760 Absolutely. 506 00:14:36,760 --> 00:14:39,400 AI could help streamline administrative tasks, 507 00:14:39,400 --> 00:14:43,240 personalize treatment plans, optimize hospital operations, 508 00:14:43,240 --> 00:14:47,240 and even develop new, more cost effective drugs and therapies. 509 00:14:47,240 --> 00:14:49,840 The possibilities are really quite vast. 510 00:14:49,840 --> 00:14:52,640 So AI could help us save money at every stage of the health 511 00:14:52,640 --> 00:14:56,800 care process, from diagnosis to treatment to ongoing care. 512 00:14:56,800 --> 00:14:57,640 Exactly. 513 00:14:57,640 --> 00:14:58,320 But there's a catch. 514 00:14:58,320 --> 00:15:00,640 The authors also warn about the potential for AI 515 00:15:00,640 --> 00:15:02,640 to exacerbate health care inequality. 516 00:15:02,640 --> 00:15:05,360 What do you mean how could AI make health care inequality worse? 517 00:15:05,360 --> 00:15:08,440 Well, imagine if all these amazing AI powered tools 518 00:15:08,440 --> 00:15:10,480 and treatments are only available to the wealthy, 519 00:15:10,480 --> 00:15:12,040 that would create a two tiered system 520 00:15:12,040 --> 00:15:14,040 where the rich have access to the best care 521 00:15:14,040 --> 00:15:15,640 while everyone else is left behind. 522 00:15:15,640 --> 00:15:17,120 That would be a real tragedy. 523 00:15:17,120 --> 00:15:18,760 So how do we prevent that from happening? 524 00:15:18,760 --> 00:15:21,160 How do we make sure the benefits of AI and health care 525 00:15:21,160 --> 00:15:22,560 are shared by everyone? 526 00:15:22,560 --> 00:15:24,600 That's the million dollar question, isn't it? 527 00:15:24,600 --> 00:15:27,680 It requires a multifaceted approach. 528 00:15:27,680 --> 00:15:31,680 We need policies that promote equitable access 529 00:15:31,680 --> 00:15:34,000 to AI powered health care investments 530 00:15:34,000 --> 00:15:37,640 in training and education to ensure everyone has the skills 531 00:15:37,640 --> 00:15:40,840 to thrive in an AI driven economy and regulations 532 00:15:40,840 --> 00:15:43,480 that prevent AI from being used in discriminatory ways. 533 00:15:43,480 --> 00:15:45,680 So it's not just about developing the technology. 534 00:15:45,680 --> 00:15:48,800 It's about creating a system where everyone can benefit from it. 535 00:15:48,800 --> 00:15:49,400 Exactly. 536 00:15:49,400 --> 00:15:51,360 And that's why this paper is so important. 537 00:15:51,360 --> 00:15:53,520 It's not just a dry economic analysis. 538 00:15:53,520 --> 00:15:54,560 It's a call to action. 539 00:15:54,560 --> 00:15:56,360 It's a plea for us to think carefully 540 00:15:56,360 --> 00:15:58,400 about the future we want to create with AI 541 00:15:58,400 --> 00:16:00,560 and to take steps to ensure that it's a future that 542 00:16:00,560 --> 00:16:01,800 works for everyone. 543 00:16:01,800 --> 00:16:03,200 That's a powerful message. 544 00:16:03,200 --> 00:16:05,600 This has been a really thought provoking deep dive. 545 00:16:05,600 --> 00:16:07,640 What are some final thoughts for our listeners 546 00:16:07,640 --> 00:16:09,640 as they grapple with these complex issues? 547 00:16:09,640 --> 00:16:12,800 I think the most important thing is to stay informed and engaged. 548 00:16:12,800 --> 00:16:15,000 The future of AI is not predetermined. 549 00:16:15,000 --> 00:16:16,560 It's something we're actively shaping 550 00:16:16,560 --> 00:16:18,200 through our choices and actions. 551 00:16:18,200 --> 00:16:20,640 So stay curious, ask questions, and demand a future 552 00:16:20,640 --> 00:16:24,360 where AI is used to create a more just and equitable world. 553 00:16:24,360 --> 00:16:26,840 That's a great note to end on. 554 00:16:26,840 --> 00:16:27,960 This paper is business. 555 00:16:27,960 --> 00:16:29,560 It's given us a lot to think about. 556 00:16:29,560 --> 00:16:31,880 And I encourage everyone to check it out for themselves. 557 00:16:31,880 --> 00:16:33,640 It's a fascinating and important read. 558 00:16:33,640 --> 00:16:52,640 Until next time, keep diving deep and keep thinking critically.