1 00:00:00,000 --> 00:00:02,220 Welcome back to the deep dive. 2 00:00:02,220 --> 00:00:06,380 Today we're taking a closer look at, well, AI in healthcare. 3 00:00:06,380 --> 00:00:07,560 Sounds pretty timely, right? 4 00:00:07,560 --> 00:00:10,820 Absolutely, it seems like every day there's some new headline 5 00:00:10,820 --> 00:00:14,960 about how AI is gonna revolutionize, well, everything. 6 00:00:14,960 --> 00:00:16,080 It really does. 7 00:00:16,080 --> 00:00:17,920 So we thought it'd be fascinating to kind of 8 00:00:17,920 --> 00:00:20,920 step back from the hype a little bit. 9 00:00:20,920 --> 00:00:21,760 Right, right. 10 00:00:21,760 --> 00:00:24,360 And actually examine how AI is being used 11 00:00:24,360 --> 00:00:25,800 on the ground in healthcare. 12 00:00:25,800 --> 00:00:27,880 Absolutely, get into the nitty gritty. 13 00:00:27,880 --> 00:00:29,080 Yeah, yeah. 14 00:00:29,080 --> 00:00:31,680 And for this deep dive, we've got some amazing material 15 00:00:31,680 --> 00:00:32,520 to work with. 16 00:00:32,520 --> 00:00:33,360 Yeah, we do. 17 00:00:33,360 --> 00:00:35,680 Excerpts from a talk given by a Stanford professor. 18 00:00:35,680 --> 00:00:36,960 A real expert in the field. 19 00:00:36,960 --> 00:00:37,800 Exactly. 20 00:00:37,800 --> 00:00:39,600 And, you know, our mission here, 21 00:00:39,600 --> 00:00:41,840 what we wanna really understand is the impact 22 00:00:41,840 --> 00:00:43,320 of AI in healthcare. 23 00:00:43,320 --> 00:00:46,600 Not just the tech itself, but how it's actually affecting, 24 00:00:46,600 --> 00:00:49,000 you know, patients, doctors. 25 00:00:49,000 --> 00:00:50,440 Right, the whole healthcare system. 26 00:00:50,440 --> 00:00:51,800 Yeah, the whole ecosystem. 27 00:00:51,800 --> 00:00:54,440 So some big questions we wanna explore. 28 00:00:54,440 --> 00:00:58,240 How can AI help doctors make better decisions 29 00:00:58,240 --> 00:00:59,560 and even faster? 30 00:00:59,560 --> 00:01:01,360 Faster, yeah. 31 00:01:01,360 --> 00:01:02,840 More efficient, hopefully. 32 00:01:02,840 --> 00:01:03,800 Right. 33 00:01:03,800 --> 00:01:06,760 Can it help us predict things like, 34 00:01:06,760 --> 00:01:09,560 who might benefit from end-of-life care conversations 35 00:01:09,560 --> 00:01:10,400 sooner? 36 00:01:10,400 --> 00:01:11,240 That's a big one. 37 00:01:11,240 --> 00:01:12,320 It is. 38 00:01:12,320 --> 00:01:16,760 And can AI actually, you know, help healthcare systems 39 00:01:16,760 --> 00:01:17,960 save money? 40 00:01:17,960 --> 00:01:19,120 That's the hope, right. 41 00:01:19,120 --> 00:01:21,440 While at the same time improving patient care. 42 00:01:21,440 --> 00:01:23,320 Right, finding that sweet spot. 43 00:01:23,320 --> 00:01:25,160 Okay, so to help us unpack all of this, 44 00:01:25,160 --> 00:01:26,800 we have, well, you. 45 00:01:26,800 --> 00:01:27,640 Oh, me. 46 00:01:27,640 --> 00:01:29,600 I'm an expert in, well, you tell us. 47 00:01:29,600 --> 00:01:32,920 I focus on the intersection of AI and healthcare policy. 48 00:01:32,920 --> 00:01:34,400 Perfect, perfect. 49 00:01:34,400 --> 00:01:36,880 Okay, so where do we even begin with this? 50 00:01:36,880 --> 00:01:39,320 Well, you can't really talk about AI and healthcare 51 00:01:39,320 --> 00:01:40,840 without talking about data. 52 00:01:40,840 --> 00:01:42,160 Okay, data. 53 00:01:42,160 --> 00:01:43,320 AI loves data. 54 00:01:43,320 --> 00:01:44,640 Right, it's the fuel. 55 00:01:44,640 --> 00:01:47,880 But healthcare data, is that a special kind of data? 56 00:01:47,880 --> 00:01:49,120 Oh, it's a whole different animal. 57 00:01:49,120 --> 00:01:49,960 How so? 58 00:01:49,960 --> 00:01:52,000 Think of it this way, instead of like a snapshot, 59 00:01:52,000 --> 00:01:54,640 it's more like a long winding novel. 60 00:01:54,640 --> 00:01:55,480 A novel. 61 00:01:55,480 --> 00:01:56,320 Interesting. 62 00:01:56,320 --> 00:01:59,000 You're dealing with a patient's entire history. 63 00:01:59,000 --> 00:02:00,680 Every hospital visit. 64 00:02:00,680 --> 00:02:03,480 Every medication they've taken, every test result, 65 00:02:03,480 --> 00:02:06,200 every doctor's note, I mean, everything. 66 00:02:06,200 --> 00:02:09,000 Wow, okay, so it's not just analyzing like, 67 00:02:09,000 --> 00:02:10,760 you know, a single moment in time. 68 00:02:10,760 --> 00:02:11,600 No, yeah. 69 00:02:11,600 --> 00:02:13,600 It's like understanding their whole, the whole narrative. 70 00:02:13,600 --> 00:02:14,840 The whole, yeah. 71 00:02:14,840 --> 00:02:16,600 The whole journey, their health journey. 72 00:02:16,600 --> 00:02:17,440 Right. 73 00:02:17,440 --> 00:02:19,120 And that's where it gets really, really tricky. 74 00:02:19,120 --> 00:02:19,960 Right. 75 00:02:19,960 --> 00:02:22,400 Because this novel, it's often scattered across 76 00:02:22,400 --> 00:02:24,680 different systems collected over years and years 77 00:02:24,680 --> 00:02:27,720 with potential gaps in consistencies. 78 00:02:27,720 --> 00:02:28,560 Oh, I see. 79 00:02:28,560 --> 00:02:30,360 You rarely have a complete like, 80 00:02:30,360 --> 00:02:33,120 up to the minute picture of a patient's health. 81 00:02:33,120 --> 00:02:36,520 That sounds like kind of a nightmare for AI. 82 00:02:36,520 --> 00:02:37,360 It is. 83 00:02:37,360 --> 00:02:40,240 How can it build like accurate models with all that, 84 00:02:40,240 --> 00:02:41,280 that mess? 85 00:02:41,280 --> 00:02:43,400 Well, Stanford realized this early on. 86 00:02:43,400 --> 00:02:44,240 Yeah. 87 00:02:44,240 --> 00:02:47,000 They started investing heavily in data infrastructure 88 00:02:47,000 --> 00:02:48,160 back in 2005. 89 00:02:48,160 --> 00:02:50,280 Wow, that's really forward thinking. 90 00:02:50,280 --> 00:02:51,880 Yeah, they were ahead of the curve. 91 00:02:51,880 --> 00:02:54,240 They knew that high quality data was gonna be 92 00:02:54,240 --> 00:02:56,000 the foundation for everything. 93 00:02:56,000 --> 00:02:57,920 So they were building the foundation before they even knew 94 00:02:57,920 --> 00:02:59,480 what kind of house they were building. 95 00:02:59,480 --> 00:03:01,280 Exactly, exactly. 96 00:03:01,280 --> 00:03:04,680 And now they have this unified data platform 97 00:03:04,680 --> 00:03:07,720 that allows them to kind of connect the dots, 98 00:03:07,720 --> 00:03:11,200 pull out meaningful insights from all of this data. 99 00:03:11,200 --> 00:03:14,800 Okay, so what's AI actually doing with all this data? 100 00:03:14,800 --> 00:03:16,120 Well, think of it this way. 101 00:03:16,120 --> 00:03:17,840 It has two main jobs. 102 00:03:17,840 --> 00:03:18,680 Okay. 103 00:03:18,680 --> 00:03:20,520 One, helping doctors decide whether to treat. 104 00:03:20,520 --> 00:03:22,960 And two, how to treat. 105 00:03:22,960 --> 00:03:25,480 And these decisions, they're being made all the time, 106 00:03:25,480 --> 00:03:26,320 you know? 107 00:03:26,320 --> 00:03:28,200 Every day in every clinical setting. 108 00:03:28,200 --> 00:03:29,040 Right. 109 00:03:29,040 --> 00:03:31,920 Can you give me like a real world example of how that work? 110 00:03:31,920 --> 00:03:36,920 Sure, so let's say a patient comes in with a cough, a fever. 111 00:03:37,440 --> 00:03:38,280 Okay. 112 00:03:38,280 --> 00:03:39,480 The doctor needs to figure out, okay, 113 00:03:39,480 --> 00:03:41,320 is this something that needs treatment? 114 00:03:41,320 --> 00:03:42,160 Right. 115 00:03:42,160 --> 00:03:43,920 And if so, what's the best treatment? 116 00:03:43,920 --> 00:03:46,400 AI can analyze that patient's data, 117 00:03:46,400 --> 00:03:49,720 consider similar cases, and offer recommendations. 118 00:03:49,720 --> 00:03:52,600 Okay, so it's like having like a super smart assistant 119 00:03:52,600 --> 00:03:54,800 who's read every medical textbook. 120 00:03:54,800 --> 00:03:55,640 Yeah. 121 00:03:55,640 --> 00:03:58,480 And can instantly pull up relevant information. 122 00:03:58,480 --> 00:03:59,440 It's a great analogy. 123 00:03:59,440 --> 00:04:01,160 Okay, so it's helping the doctor make those decisions, 124 00:04:01,160 --> 00:04:03,880 but how does it actually do that? 125 00:04:03,880 --> 00:04:08,680 Well, one way is by acting as a kind of medical detective, 126 00:04:08,680 --> 00:04:09,520 right? 127 00:04:09,520 --> 00:04:11,640 We call that classification or diagnosis. 128 00:04:11,640 --> 00:04:13,800 So think about like analyzing an image 129 00:04:13,800 --> 00:04:15,720 to see if there are signs of pneumonia. 130 00:04:15,720 --> 00:04:16,560 Okay. 131 00:04:16,560 --> 00:04:19,320 That's AI using data to figure out what's already there. 132 00:04:19,320 --> 00:04:20,160 Right. 133 00:04:20,160 --> 00:04:22,680 AI can also predict what might happen. 134 00:04:22,680 --> 00:04:23,520 Oh, interesting. 135 00:04:23,520 --> 00:04:25,400 That's prediction or prognosis. 136 00:04:25,400 --> 00:04:26,240 Okay. 137 00:04:26,240 --> 00:04:30,560 So here it's using data to forecast future events. 138 00:04:30,560 --> 00:04:31,400 Bro. 139 00:04:31,400 --> 00:04:35,360 Like a patient's risk of developing a certain condition 140 00:04:35,360 --> 00:04:37,400 or their chances of recovery. 141 00:04:37,400 --> 00:04:40,040 So it's like having a crystal ball kind of. 142 00:04:40,040 --> 00:04:41,280 It is pretty remarkable. 143 00:04:41,280 --> 00:04:42,200 Yeah, it really is. 144 00:04:42,200 --> 00:04:43,640 And we're also seeing this new thing 145 00:04:43,640 --> 00:04:46,160 called generative AI being used. 146 00:04:46,160 --> 00:04:49,080 Oh yeah, generative AI, that's everywhere now. 147 00:04:49,080 --> 00:04:50,480 It is, and in healthcare, 148 00:04:50,480 --> 00:04:53,000 it can take all this complex data 149 00:04:53,000 --> 00:04:56,320 and summarize it into like these really digestible reports 150 00:04:56,320 --> 00:04:57,240 for doctors. 151 00:04:57,240 --> 00:04:58,080 Oh, I see. 152 00:04:58,080 --> 00:04:59,880 So instead of wading through, you know, 153 00:04:59,880 --> 00:05:01,280 pages and pages of notes, 154 00:05:01,280 --> 00:05:04,680 they get this concise summary that's AI generated. 155 00:05:04,680 --> 00:05:06,920 Okay, that makes a lot of sense. 156 00:05:06,920 --> 00:05:09,640 But how does all of this, this amazing technology 157 00:05:09,640 --> 00:05:12,280 actually translate into benefits for patients? 158 00:05:12,280 --> 00:05:14,360 All right, so it's not just about the technology itself, 159 00:05:14,360 --> 00:05:15,360 it's about the impact. 160 00:05:15,360 --> 00:05:16,200 Yeah. 161 00:05:16,200 --> 00:05:17,720 Well, the professor uses this great example, 162 00:05:17,720 --> 00:05:19,520 a heart condition called heart failure 163 00:05:19,520 --> 00:05:21,520 with preserved ejection fraction. 164 00:05:21,520 --> 00:05:22,360 Okay. 165 00:05:22,360 --> 00:05:23,800 Or H-F-P-E-F. 166 00:05:23,800 --> 00:05:24,640 Okay. 167 00:05:24,640 --> 00:05:26,920 Now this used to be considered just one disease. 168 00:05:26,920 --> 00:05:27,760 Okay. 169 00:05:27,760 --> 00:05:29,600 But AI helped researchers discover 170 00:05:29,600 --> 00:05:32,320 that it's actually three distinct conditions. 171 00:05:32,320 --> 00:05:33,160 Oh wow. 172 00:05:33,160 --> 00:05:35,120 By analyzing all this patient data. 173 00:05:35,120 --> 00:05:37,640 So AI helped solve a medical mystery. 174 00:05:37,640 --> 00:05:38,480 It did. 175 00:05:38,480 --> 00:05:39,320 That's amazing. 176 00:05:39,320 --> 00:05:42,560 And it shows how AI impacts healthcare in three ways, 177 00:05:42,560 --> 00:05:44,880 science, practice, and delivery. 178 00:05:44,880 --> 00:05:45,840 Okay, break that down for me. 179 00:05:45,840 --> 00:05:48,200 Sure, so first the science level. 180 00:05:48,200 --> 00:05:51,360 AI advanced our understanding of H-F-P-E-F 181 00:05:51,360 --> 00:05:53,280 by revealing these subtypes. 182 00:05:53,280 --> 00:05:54,120 Right. 183 00:05:54,120 --> 00:05:55,440 Each with a different prognosis. 184 00:05:55,440 --> 00:05:56,880 Then the practice level, 185 00:05:56,880 --> 00:05:59,360 we can now develop specific diagnostic tests 186 00:05:59,360 --> 00:06:00,400 for each subtype. 187 00:06:00,400 --> 00:06:01,240 Okay. 188 00:06:01,240 --> 00:06:03,120 Create tailored treatment guidelines. 189 00:06:03,120 --> 00:06:04,680 So it's like AI gave doctors 190 00:06:04,680 --> 00:06:07,040 a much more detailed map of the disease. 191 00:06:07,040 --> 00:06:07,880 Exactly. 192 00:06:07,880 --> 00:06:08,720 Okay. 193 00:06:08,720 --> 00:06:10,320 But the final piece, and it's often overlooked, 194 00:06:10,320 --> 00:06:11,360 is delivery. 195 00:06:11,360 --> 00:06:13,440 It's not enough to have the knowledge and the tool, 196 00:06:13,440 --> 00:06:15,560 the healthcare system actually has to implement 197 00:06:15,560 --> 00:06:17,120 these discoveries. 198 00:06:17,120 --> 00:06:17,960 Yeah. 199 00:06:17,960 --> 00:06:19,440 Adopt new tests and treatments, 200 00:06:19,440 --> 00:06:21,160 improve care coordination. 201 00:06:21,160 --> 00:06:23,200 Right, so everything has to change 202 00:06:23,200 --> 00:06:24,680 to support these new discoveries. 203 00:06:24,680 --> 00:06:25,520 Exactly. 204 00:06:25,520 --> 00:06:26,840 Okay, so that makes a lot of sense. 205 00:06:26,840 --> 00:06:29,760 And it sounds like it's pushing the healthcare system 206 00:06:29,760 --> 00:06:30,720 to evolve. 207 00:06:30,720 --> 00:06:31,560 It is. 208 00:06:31,560 --> 00:06:33,640 It's pushing it to be more efficient, 209 00:06:33,640 --> 00:06:35,240 more patient centered. 210 00:06:35,240 --> 00:06:36,080 That's great. 211 00:06:37,280 --> 00:06:40,600 So can you give me another example of AI in action? 212 00:06:40,600 --> 00:06:41,640 Absolutely. 213 00:06:41,640 --> 00:06:43,360 The professor talks about this project 214 00:06:43,360 --> 00:06:44,520 called the green button. 215 00:06:44,520 --> 00:06:45,360 The green button. 216 00:06:45,360 --> 00:06:48,120 Yeah, it was this AI powered service at Stanford. 217 00:06:48,120 --> 00:06:48,960 Okay. 218 00:06:48,960 --> 00:06:51,480 It could give doctors data driven insights 219 00:06:51,480 --> 00:06:53,240 on similar patients. 220 00:06:53,240 --> 00:06:54,600 All within a single day. 221 00:06:54,600 --> 00:06:57,040 A single day, that's really fast. 222 00:06:57,040 --> 00:06:57,880 It is. 223 00:06:57,880 --> 00:06:58,960 How is that even possible? 224 00:06:58,960 --> 00:07:02,040 I will remember how Stanford has been investing 225 00:07:02,040 --> 00:07:04,000 in that data infrastructure. 226 00:07:04,000 --> 00:07:05,440 Right, right, the foundation. 227 00:07:05,440 --> 00:07:06,640 Yeah, it paid off here. 228 00:07:06,640 --> 00:07:07,480 Yeah. 229 00:07:07,480 --> 00:07:09,080 This project was designed to address 230 00:07:09,080 --> 00:07:11,200 a pretty shocking statistic. 231 00:07:11,200 --> 00:07:14,080 80% of medical decisions 232 00:07:14,080 --> 00:07:17,200 are made without solid evidence. 233 00:07:17,200 --> 00:07:18,040 80%. 234 00:07:19,520 --> 00:07:20,360 Wait, really? 235 00:07:20,360 --> 00:07:21,760 That's a little unsettling. 236 00:07:21,760 --> 00:07:22,600 It is. 237 00:07:22,600 --> 00:07:25,600 Doctors often have to rely on experience, 238 00:07:25,600 --> 00:07:26,840 intuition, 239 00:07:26,840 --> 00:07:28,480 limited information. 240 00:07:28,480 --> 00:07:30,120 The green button gave them access 241 00:07:30,120 --> 00:07:32,680 to all this data on similar patients 242 00:07:32,680 --> 00:07:34,800 so they can make more informed decisions. 243 00:07:34,800 --> 00:07:39,400 Okay, so it's like AI is giving doctors a cheat sheet. 244 00:07:39,400 --> 00:07:40,240 But in a good way. 245 00:07:40,240 --> 00:07:42,360 Yeah, a cheat sheet for better patient care. 246 00:07:42,360 --> 00:07:43,440 Okay. 247 00:07:43,440 --> 00:07:45,480 Is this green button still in use? 248 00:07:45,480 --> 00:07:47,280 Well, it's evolved. 249 00:07:47,280 --> 00:07:50,200 They created a company called Atropos Health 250 00:07:50,200 --> 00:07:51,960 to make this technology available 251 00:07:51,960 --> 00:07:53,320 to other healthcare systems. 252 00:07:53,320 --> 00:07:54,160 That's great. 253 00:07:54,160 --> 00:07:55,800 So they're not just keeping it to themselves. 254 00:07:55,800 --> 00:07:57,320 No, they wanna make a real impact. 255 00:07:57,320 --> 00:07:58,240 Yeah. 256 00:07:58,240 --> 00:08:00,080 Okay, so we've talked about helping doctors 257 00:08:00,080 --> 00:08:01,360 make better decisions. 258 00:08:01,360 --> 00:08:03,640 But what about the ethics of all this? 259 00:08:03,640 --> 00:08:06,040 Right, the professor talks about another example. 260 00:08:06,040 --> 00:08:08,880 Using AI to predict mortality. 261 00:08:08,880 --> 00:08:10,200 Okay, that sounds a little bit ominous. 262 00:08:10,200 --> 00:08:11,040 I know. 263 00:08:11,040 --> 00:08:13,840 But the goal wasn't to make some grim prediction 264 00:08:13,840 --> 00:08:15,320 about someone's lifespan. 265 00:08:15,320 --> 00:08:18,280 It was to identify patients who might benefit 266 00:08:18,280 --> 00:08:22,960 from having end of life care conversations sooner. 267 00:08:22,960 --> 00:08:24,560 So not predicting death. 268 00:08:24,560 --> 00:08:25,400 Yeah. 269 00:08:25,400 --> 00:08:27,840 But improving the quality of life in the final stages. 270 00:08:27,840 --> 00:08:30,400 Right, and making sure that their wishes are respected. 271 00:08:30,400 --> 00:08:31,240 Exactly. 272 00:08:31,240 --> 00:08:34,400 Okay, so how did they address the potential 273 00:08:34,400 --> 00:08:37,800 for misuse of this kind of technology? 274 00:08:37,800 --> 00:08:38,920 They were really careful. 275 00:08:38,920 --> 00:08:41,640 They worked with the Patient Advisory Council 276 00:08:41,640 --> 00:08:44,360 to make sure it was focused on improving care. 277 00:08:44,360 --> 00:08:48,120 And that it wouldn't be used for things like insurance pricing 278 00:08:48,120 --> 00:08:51,280 or selective enrollment in healthcare plans. 279 00:08:51,280 --> 00:08:52,560 Yeah, okay, that makes sense. 280 00:08:52,560 --> 00:08:54,480 They're trying to walk that line 281 00:08:54,480 --> 00:08:56,040 between the benefits of AI 282 00:08:56,040 --> 00:08:57,840 and making sure it's used responsibly. 283 00:08:57,840 --> 00:09:00,800 Absolutely, and it brings up another important point. 284 00:09:00,800 --> 00:09:03,840 Having these powerful AI insights is great. 285 00:09:03,840 --> 00:09:06,560 But only if you can actually act on them. 286 00:09:06,560 --> 00:09:07,400 What do you mean? 287 00:09:07,400 --> 00:09:10,560 He uses this analogy of the Mercedes-Benz logo, 288 00:09:10,560 --> 00:09:11,720 you know, the three-pointed star. 289 00:09:11,720 --> 00:09:12,560 Okay, yeah. 290 00:09:12,560 --> 00:09:15,640 He says the three points represent models, 291 00:09:15,640 --> 00:09:18,000 workforce capacity, and policy. 292 00:09:18,000 --> 00:09:21,040 All three have to work together to actually drive change. 293 00:09:21,040 --> 00:09:24,320 So you could have a fancy car with a powerful engine. 294 00:09:24,320 --> 00:09:25,160 Right. 295 00:09:25,160 --> 00:09:28,160 But if there's no driver and no roads to go on, 296 00:09:28,160 --> 00:09:29,120 it's not going anywhere. 297 00:09:29,120 --> 00:09:30,880 Exactly, and that's the trap 298 00:09:30,880 --> 00:09:32,400 that a lot of people fall into at AI. 299 00:09:32,400 --> 00:09:34,960 They focus on building these sophisticated models, 300 00:09:34,960 --> 00:09:36,920 but they forget about the implementation. 301 00:09:36,920 --> 00:09:37,760 Right. 302 00:09:37,760 --> 00:09:38,720 The resources, the people. 303 00:09:38,720 --> 00:09:40,400 Okay, so it's not just about the technology, 304 00:09:40,400 --> 00:09:41,640 it's about the whole ecosystem. 305 00:09:41,640 --> 00:09:42,480 Exactly. 306 00:09:42,480 --> 00:09:43,320 Right. 307 00:09:43,320 --> 00:09:45,360 So speaking of exciting new technology, 308 00:09:45,360 --> 00:09:47,400 what about generative AI? 309 00:09:47,400 --> 00:09:52,400 Ah, yes, generative AI, the buzzword of the moment. 310 00:09:52,640 --> 00:09:53,480 It's everywhere. 311 00:09:53,480 --> 00:09:54,840 Everybody's talking about chat, GP. 312 00:09:54,840 --> 00:09:56,520 It's gonna revolutionize everything, right? 313 00:09:56,520 --> 00:10:00,600 Right, so what's the professor's take on all this hype? 314 00:10:00,600 --> 00:10:02,600 Well, he acknowledges the excitement, 315 00:10:02,600 --> 00:10:04,840 but he also urges caution. 316 00:10:04,840 --> 00:10:09,840 He did his own deep dive, analyzing over 520 research papers 317 00:10:09,960 --> 00:10:12,680 on generative AI and healthcare. 318 00:10:12,680 --> 00:10:14,520 And the findings are pretty interesting. 319 00:10:14,520 --> 00:10:18,000 Most studies are using non-patient data, 320 00:10:18,000 --> 00:10:19,800 and they're focusing on things 321 00:10:19,800 --> 00:10:22,040 like answering those USMLE questions, 322 00:10:22,040 --> 00:10:24,440 you know, the exams medical students take. 323 00:10:24,440 --> 00:10:25,280 Right, right. 324 00:10:25,280 --> 00:10:28,160 So they're testing AI's medical knowledge, 325 00:10:28,160 --> 00:10:30,680 but not necessarily its ability to function 326 00:10:30,680 --> 00:10:31,520 in the real world, 327 00:10:31,520 --> 00:10:33,760 and that's where we start to see the limitations. 328 00:10:33,760 --> 00:10:37,680 For example, chat GPT can give inconsistent answers. 329 00:10:37,680 --> 00:10:39,680 There are high error rates 330 00:10:39,680 --> 00:10:41,840 when it's trying to complete medical tasks. 331 00:10:41,840 --> 00:10:42,680 Okay. 332 00:10:42,680 --> 00:10:46,240 And surprisingly, there were no real productivity gains 333 00:10:46,240 --> 00:10:48,440 when it was responding to patient messages. 334 00:10:48,440 --> 00:10:50,560 So it's not quite living up to the hype? 335 00:10:50,560 --> 00:10:51,560 Not yet. 336 00:10:51,560 --> 00:10:53,800 He emphasizes the importance of, you know, 337 00:10:53,800 --> 00:10:57,520 rigorous evaluation and setting realistic expectations. 338 00:10:57,520 --> 00:10:59,120 It's still early days. 339 00:10:59,120 --> 00:10:59,960 Right. 340 00:10:59,960 --> 00:11:01,320 We need to be careful about getting swept up 341 00:11:01,320 --> 00:11:02,160 in the excitement. 342 00:11:02,160 --> 00:11:03,560 Yeah, we don't wanna get ahead of ourselves. 343 00:11:03,560 --> 00:11:07,400 Okay, so if chat GPT isn't the answer, what is? 344 00:11:07,400 --> 00:11:10,200 What is the future of AI and healthcare look like? 345 00:11:10,200 --> 00:11:11,360 Well, the professor's work 346 00:11:11,360 --> 00:11:13,320 offers a really intriguing direction. 347 00:11:13,320 --> 00:11:17,720 They're training AI on the language of patient timelines. 348 00:11:17,720 --> 00:11:19,760 Language, what does that even mean? 349 00:11:19,760 --> 00:11:21,160 I know, it sounds a little strange, 350 00:11:21,160 --> 00:11:24,600 but think back to how we talked healthcare data 351 00:11:24,600 --> 00:11:26,280 as this timeline of events. 352 00:11:26,280 --> 00:11:29,120 Right, diagnoses, medications, lab results. 353 00:11:29,120 --> 00:11:29,960 Exactly. 354 00:11:29,960 --> 00:11:33,720 To a computer, that sequence looks a lot like a language. 355 00:11:33,720 --> 00:11:34,560 Oh, okay. 356 00:11:34,560 --> 00:11:36,000 With its own grammar and patterns. 357 00:11:36,000 --> 00:11:40,120 So AI is learning to read this story of a patient's health. 358 00:11:40,120 --> 00:11:41,280 You got it. 359 00:11:41,280 --> 00:11:43,080 And by learning this language, 360 00:11:43,080 --> 00:11:46,880 AI can actually predict future events. 361 00:11:46,880 --> 00:11:47,720 Wow. 362 00:11:47,720 --> 00:11:50,000 Outcomes with incredible accuracy. 363 00:11:50,000 --> 00:11:50,920 That's pretty amazing. 364 00:11:50,920 --> 00:11:51,760 It is. 365 00:11:51,760 --> 00:11:52,600 So how is this different 366 00:11:52,600 --> 00:11:54,120 from the models we've already talked about? 367 00:11:54,120 --> 00:11:56,000 Well, the key here is efficiency. 368 00:11:56,000 --> 00:11:56,840 Okay. 369 00:11:56,840 --> 00:12:00,040 AI is driving the data to the most significant extent 370 00:12:00,040 --> 00:12:03,640 that it needs far less data to train effective model. 371 00:12:03,640 --> 00:12:04,480 Less data. 372 00:12:04,480 --> 00:12:05,800 That's a big deal. 373 00:12:05,800 --> 00:12:06,640 It is. 374 00:12:06,640 --> 00:12:08,920 They were able to get the same level of performance 375 00:12:08,920 --> 00:12:10,520 with a small sample of cases 376 00:12:10,520 --> 00:12:12,840 as a model trained on mountains of data. 377 00:12:12,840 --> 00:12:13,680 That's incredible. 378 00:12:13,680 --> 00:12:14,520 It is. 379 00:12:14,520 --> 00:12:16,480 And as with any powerful technology, 380 00:12:16,480 --> 00:12:17,720 there are considerations. 381 00:12:17,720 --> 00:12:19,600 Stanford is very aware of that. 382 00:12:19,600 --> 00:12:21,840 And they're working to ensure this technology 383 00:12:21,840 --> 00:12:23,720 is developed and deployed responsibly. 384 00:12:23,720 --> 00:12:24,560 That's good to hear. 385 00:12:24,560 --> 00:12:26,360 So we've covered a lot of ground here, 386 00:12:26,360 --> 00:12:28,800 AI is being used, diagnosis, treatment, 387 00:12:28,800 --> 00:12:30,080 even end of life care. 388 00:12:30,080 --> 00:12:35,000 And we talked about the importance of workforce capacity, 389 00:12:35,000 --> 00:12:37,440 policy, making sure everything works together. 390 00:12:37,440 --> 00:12:38,280 Right. 391 00:12:38,280 --> 00:12:40,920 Not just the technology itself, the whole ecosystem. 392 00:12:40,920 --> 00:12:42,760 And we even touched on generative AI, 393 00:12:42,760 --> 00:12:44,560 the hype, the reality. 394 00:12:44,560 --> 00:12:46,160 It's a lot to take in. 395 00:12:46,160 --> 00:12:49,560 But it's fascinating to see how AI 396 00:12:49,560 --> 00:12:51,400 is already making a difference. 397 00:12:51,400 --> 00:12:52,240 Absolutely. 398 00:12:52,240 --> 00:12:53,600 The Green Button Project, 399 00:12:53,600 --> 00:12:56,120 helping doctors with real-time insights, 400 00:12:56,120 --> 00:12:58,720 discovering new subtypes of heart conditions. 401 00:12:58,720 --> 00:12:59,560 Pretty amazing. 402 00:12:59,560 --> 00:13:00,400 It is. 403 00:13:00,400 --> 00:13:01,920 And it makes you wonder, what's next? 404 00:13:01,920 --> 00:13:03,520 What does the future hold? 405 00:13:03,520 --> 00:13:06,160 Yeah, it does feel like we're just scratching the surface. 406 00:13:06,160 --> 00:13:07,000 Absolutely. 407 00:13:07,000 --> 00:13:09,280 The field is evolving so rapidly. 408 00:13:09,280 --> 00:13:12,240 What's cutting edge today might be commonplace tomorrow. 409 00:13:12,240 --> 00:13:13,160 It's exciting. 410 00:13:13,160 --> 00:13:14,640 And a little bit daunting too, right? 411 00:13:14,640 --> 00:13:15,480 It is. 412 00:13:15,480 --> 00:13:17,280 AI has so much potential. 413 00:13:17,280 --> 00:13:19,120 But we need to make sure it's used for good. 414 00:13:19,120 --> 00:13:20,280 Absolutely. 415 00:13:20,280 --> 00:13:23,800 The professor really emphasizes that AI in health care 416 00:13:23,800 --> 00:13:26,400 should be about more than just the technology. 417 00:13:26,400 --> 00:13:27,240 Right. 418 00:13:27,240 --> 00:13:28,920 It's about making a real impact on people's lives. 419 00:13:28,920 --> 00:13:29,760 Right. 420 00:13:29,760 --> 00:13:32,560 It's about empowering doctors, improving patient outcomes, 421 00:13:32,560 --> 00:13:34,840 creating a health care system that works for everyone. 422 00:13:34,840 --> 00:13:36,280 Well said. 423 00:13:36,280 --> 00:13:39,960 So for you, the listener, as AI continues to evolve, 424 00:13:39,960 --> 00:13:43,440 how do you think it will shape your health care experience? 425 00:13:43,440 --> 00:13:47,720 Will you be more likely to seek out AI-powered tools? 426 00:13:47,720 --> 00:13:50,400 Will you feel more confident in your doctor's decisions 427 00:13:50,400 --> 00:13:52,680 knowing they have access to this technology? 428 00:13:52,680 --> 00:13:54,760 These are all important questions to think about. 429 00:13:54,760 --> 00:13:57,000 AI is already changing health care. 430 00:13:57,000 --> 00:13:57,440 Right. 431 00:13:57,440 --> 00:13:59,200 And its influence is only going to grow. 432 00:13:59,200 --> 00:14:00,960 So it's important to stay informed, 433 00:14:00,960 --> 00:14:03,720 engage in these conversations, and advocate 434 00:14:03,720 --> 00:14:06,240 for the kind of health care system you want to see. 435 00:14:06,240 --> 00:14:10,480 This deep dive has taken us beyond the nuts and bolts of AI. 436 00:14:10,480 --> 00:14:11,240 Absolutely. 437 00:14:11,240 --> 00:14:14,360 It's about understanding how this technology is 438 00:14:14,360 --> 00:14:16,680 shaping the very future of health care. 439 00:14:16,680 --> 00:14:17,120 It is. 440 00:14:17,120 --> 00:14:18,720 And ultimately, our own health. 441 00:14:18,720 --> 00:14:19,480 Our own health. 442 00:14:19,480 --> 00:14:23,000 And as Stanford continues to lead the way 443 00:14:23,000 --> 00:14:26,080 in responsible AI development, I think 444 00:14:26,080 --> 00:14:28,240 we can be optimistic about the future. 445 00:14:28,240 --> 00:14:29,440 Absolutely. 446 00:14:29,440 --> 00:14:31,360 Now, before we wrap up, I'm curious. 447 00:14:31,360 --> 00:14:33,120 The professor mentioned that Stanford already 448 00:14:33,120 --> 00:14:36,320 has over 30 applications of AI in health care. 449 00:14:36,320 --> 00:14:37,080 That's right. 450 00:14:37,080 --> 00:14:39,080 Across a wide range of areas. 451 00:14:39,080 --> 00:14:40,600 Wow, that's amazing. 452 00:14:40,600 --> 00:14:42,320 Can you give us a taste of what those are? 453 00:14:42,320 --> 00:14:44,240 Well, on the administrative side, 454 00:14:44,240 --> 00:14:47,920 think about things like predicting patient no shows, 455 00:14:47,920 --> 00:14:50,600 optimizing hospital bed allocation, 456 00:14:50,600 --> 00:14:54,000 even improving the accuracy of medical billing and coding. 457 00:14:54,000 --> 00:14:57,040 Oh, so streamlining a lot of those back end processes. 458 00:14:57,040 --> 00:14:58,880 Exactly, making things more efficient. 459 00:14:58,880 --> 00:14:59,440 Yeah. 460 00:14:59,440 --> 00:15:01,200 What about on the patient facing side? 461 00:15:01,200 --> 00:15:02,880 Well, you've got AI-powered chatbots 462 00:15:02,880 --> 00:15:06,400 that can answer basic questions, schedule appointments, 463 00:15:06,400 --> 00:15:08,760 even provide personalized health information. 464 00:15:08,760 --> 00:15:09,360 That's pretty cool. 465 00:15:09,360 --> 00:15:11,040 So freeing up time for health care workers 466 00:15:11,040 --> 00:15:12,680 to focus on the more complex stuff. 467 00:15:12,680 --> 00:15:12,960 Right. 468 00:15:12,960 --> 00:15:14,680 And of course, there are all the clinical applications 469 00:15:14,680 --> 00:15:16,680 we've talked about, diagnosing conditions, 470 00:15:16,680 --> 00:15:18,120 predicting outcomes. 471 00:15:18,120 --> 00:15:22,640 But AI is also being used to personalize treatment plans. 472 00:15:22,640 --> 00:15:24,440 Monitor patients remotely. 473 00:15:24,440 --> 00:15:24,960 Wow. 474 00:15:24,960 --> 00:15:26,720 Even assist with surgeries. 475 00:15:26,720 --> 00:15:30,400 It really is being integrated into so many different aspects 476 00:15:30,400 --> 00:15:31,280 of health care. 477 00:15:31,280 --> 00:15:31,760 It is. 478 00:15:31,760 --> 00:15:35,240 It feels like we're on the cusp of a real revolution. 479 00:15:35,240 --> 00:15:36,600 It does, doesn't it? 480 00:15:36,600 --> 00:15:40,640 And as the technology continues to evolve, 481 00:15:40,640 --> 00:15:42,760 I mean, who knows what we'll see next? 482 00:15:42,760 --> 00:15:44,760 The possibilities are pretty much limitless. 483 00:15:44,760 --> 00:15:45,400 They really are. 484 00:15:45,400 --> 00:15:47,720 It's an exciting time to be watching this space. 485 00:15:47,720 --> 00:15:50,520 So if chatGPT isn't the answer, what is? 486 00:15:50,520 --> 00:15:53,480 Well, the professor's work at Stanford, 487 00:15:53,480 --> 00:15:56,080 it offers a kind of different direction. 488 00:15:56,080 --> 00:15:58,880 They're training AI on the, well, they 489 00:15:58,880 --> 00:16:01,200 call it the language of patient timelines. 490 00:16:01,200 --> 00:16:02,560 Language. 491 00:16:02,560 --> 00:16:05,120 Are patients speaking in some kind of secret code? 492 00:16:05,120 --> 00:16:05,760 No, no, no. 493 00:16:05,760 --> 00:16:07,080 Not exactly. 494 00:16:07,080 --> 00:16:09,240 But remember how we were talking about health care data? 495 00:16:09,240 --> 00:16:09,800 Yeah. 496 00:16:09,800 --> 00:16:12,280 How it's like a timeline of events. 497 00:16:12,280 --> 00:16:16,600 Diagnoses, medications, lab results, all those things. 498 00:16:16,600 --> 00:16:19,600 Well, to a computer, that sequence, 499 00:16:19,600 --> 00:16:23,440 it's kind of like a language with its own grammar and patterns. 500 00:16:23,440 --> 00:16:24,400 Oh, most, yeah, I see. 501 00:16:24,400 --> 00:16:29,000 So it's like AI is learning to read the story of a patient's 502 00:16:29,000 --> 00:16:29,360 health. 503 00:16:29,360 --> 00:16:30,560 Exactly, you got it. 504 00:16:30,560 --> 00:16:31,040 OK. 505 00:16:31,040 --> 00:16:34,720 And by learning this language, AI can actually 506 00:16:34,720 --> 00:16:36,160 predict future events. 507 00:16:36,160 --> 00:16:36,560 Really? 508 00:16:36,560 --> 00:16:37,360 Yeah. 509 00:16:37,360 --> 00:16:40,800 Outcomes with really impressive accuracy. 510 00:16:40,800 --> 00:16:41,320 Wow. 511 00:16:41,320 --> 00:16:43,760 So how is this different from the other models 512 00:16:43,760 --> 00:16:44,800 we've been talking about? 513 00:16:44,800 --> 00:16:46,680 Well, the big thing here is efficiency. 514 00:16:46,680 --> 00:16:52,160 This method, it needs a lot less data to train these models. 515 00:16:52,160 --> 00:16:53,040 Less data. 516 00:16:53,040 --> 00:16:54,400 OK, so that's a big deal. 517 00:16:54,400 --> 00:16:55,520 Yeah, huge. 518 00:16:55,520 --> 00:16:59,840 They found that they could get the same level of performance 519 00:16:59,840 --> 00:17:03,200 with just a small sample of cases compared to a model that 520 00:17:03,200 --> 00:17:05,440 was trained on mountains of data. 521 00:17:05,440 --> 00:17:06,160 That's amazing. 522 00:17:06,160 --> 00:17:09,520 So less data, faster training, sounds like a win-win. 523 00:17:09,520 --> 00:17:10,200 It is. 524 00:17:10,200 --> 00:17:15,160 It's definitely promising, but as with anything powerful, 525 00:17:15,160 --> 00:17:15,960 got to be careful. 526 00:17:15,960 --> 00:17:17,560 Stanford's very aware of that. 527 00:17:17,560 --> 00:17:21,000 They're working hard to ensure this technology is developed 528 00:17:21,000 --> 00:17:22,240 and used responsibly. 529 00:17:22,240 --> 00:17:23,040 That's good to hear. 530 00:17:23,040 --> 00:17:25,240 They're taking a cautious approach. 531 00:17:25,240 --> 00:17:28,280 OK, so we've talked about the challenges with health care 532 00:17:28,280 --> 00:17:28,800 data. 533 00:17:28,800 --> 00:17:32,840 The different ways AI is being used, I mean, diagnosis, 534 00:17:32,840 --> 00:17:35,480 treatment, end of life care. 535 00:17:35,480 --> 00:17:38,800 We talked about the need for workforce capacity, policies, 536 00:17:38,800 --> 00:17:40,440 making sure everything works together. 537 00:17:40,440 --> 00:17:42,720 Making sure it's not just about the tech itself. 538 00:17:42,720 --> 00:17:43,560 Yeah, exactly. 539 00:17:43,560 --> 00:17:47,720 And we even touched on generative AI, the hype, the reality. 540 00:17:47,720 --> 00:17:48,880 The good and the bad. 541 00:17:48,880 --> 00:17:50,360 Right, it's a lot. 542 00:17:50,360 --> 00:17:50,880 It is. 543 00:17:50,880 --> 00:17:54,080 But it's fascinating to see how AI is already 544 00:17:54,080 --> 00:17:55,520 having an impact. 545 00:17:55,520 --> 00:17:56,080 It is. 546 00:17:56,080 --> 00:17:58,160 I mean, think about the Green Button Project. 547 00:17:58,160 --> 00:18:01,120 Helping doctors make decisions in real time, 548 00:18:01,120 --> 00:18:03,920 discovering new subtypes of heart conditions. 549 00:18:03,920 --> 00:18:04,640 Pretty incredible. 550 00:18:04,640 --> 00:18:06,840 Yeah, it really makes you wonder what's next, 551 00:18:06,840 --> 00:18:07,920 what's around the corner. 552 00:18:07,920 --> 00:18:10,200 It does feel like we're just at the beginning. 553 00:18:10,200 --> 00:18:11,360 We are, yeah. 554 00:18:11,360 --> 00:18:13,520 It's a rapidly evolving field. 555 00:18:13,520 --> 00:18:17,400 What's cutting edge today could be completely outdated tomorrow. 556 00:18:17,400 --> 00:18:18,120 Absolutely. 557 00:18:18,120 --> 00:18:18,680 It's exciting. 558 00:18:18,680 --> 00:18:20,040 It's a little bit scary, too. 559 00:18:20,040 --> 00:18:21,080 Yeah, it is. 560 00:18:21,080 --> 00:18:23,560 AI has so much potential. 561 00:18:23,560 --> 00:18:26,320 But we've got to make sure it's used for the right reasons. 562 00:18:26,320 --> 00:18:27,240 Right, for good. 563 00:18:27,240 --> 00:18:27,480 Yeah. 564 00:18:27,480 --> 00:18:32,800 And the professor, he really, really drives this point home. 565 00:18:32,800 --> 00:18:36,840 AI in health care, it can't just be about the technology. 566 00:18:36,840 --> 00:18:39,120 It has to be about people, about actually 567 00:18:39,120 --> 00:18:40,280 making a positive impact. 568 00:18:40,280 --> 00:18:44,040 Empowering doctors, improving outcomes for patients, 569 00:18:44,040 --> 00:18:47,120 creating a system that's fairer for everyone. 570 00:18:47,120 --> 00:18:48,520 Yeah, that's the goal. 571 00:18:48,520 --> 00:18:52,720 OK, so for our listeners out there, as all this continues, 572 00:18:52,720 --> 00:18:56,520 how do you see AI changing your experience with health care? 573 00:18:56,520 --> 00:18:57,360 Right, think about it. 574 00:18:57,360 --> 00:19:01,280 Will you be looking for AI-powered tools? 575 00:19:01,280 --> 00:19:04,080 Will you feel more confident in the decisions 576 00:19:04,080 --> 00:19:06,280 your doctor is making, knowing that they have access 577 00:19:06,280 --> 00:19:07,440 to all this information? 578 00:19:07,440 --> 00:19:08,320 Those are good questions. 579 00:19:08,320 --> 00:19:10,840 Yeah, AI is already changing things. 580 00:19:10,840 --> 00:19:11,440 It is. 581 00:19:11,440 --> 00:19:13,520 And that influence is only going to grow. 582 00:19:13,520 --> 00:19:17,800 So it's on us to stay informed, to be part of the conversation, 583 00:19:17,800 --> 00:19:20,320 and really advocate for the kind of health care we want. 584 00:19:20,320 --> 00:19:21,520 I think that's a great point. 585 00:19:21,520 --> 00:19:24,080 We've gone way beyond just the basics here. 586 00:19:24,080 --> 00:19:24,640 We have. 587 00:19:24,640 --> 00:19:27,360 It's not just about understanding how AI works. 588 00:19:27,360 --> 00:19:29,280 It's about how this is all going to impact 589 00:19:29,280 --> 00:19:31,160 the future of health care. 590 00:19:31,160 --> 00:19:32,200 Absolutely. 591 00:19:32,200 --> 00:19:34,160 And in the end, our own health. 592 00:19:34,160 --> 00:19:35,400 Our own health. 593 00:19:35,400 --> 00:19:36,240 Exactly. 594 00:19:36,240 --> 00:19:38,200 And with Stanford leading the way, 595 00:19:38,200 --> 00:19:41,800 doing this work responsibly, it makes me feel hopeful. 596 00:19:41,800 --> 00:19:42,400 Me too. 597 00:19:42,400 --> 00:19:44,760 Optimistic about what's to come. 598 00:19:44,760 --> 00:19:45,520 Yeah. 599 00:19:45,520 --> 00:19:47,680 Now, before we go, I'm curious. 600 00:19:47,680 --> 00:19:49,560 The professor mentioned that Stanford already 601 00:19:49,560 --> 00:19:53,560 has over 30 applications of AI in health care. 602 00:19:53,560 --> 00:19:54,240 That's right, yeah. 603 00:19:54,240 --> 00:19:54,880 30. 604 00:19:54,880 --> 00:19:56,560 30, all sorts of things. 605 00:19:56,560 --> 00:19:57,400 Wow. 606 00:19:57,400 --> 00:19:58,440 What are some of those? 607 00:19:58,440 --> 00:19:59,800 Can you give us some examples? 608 00:19:59,800 --> 00:20:00,280 Sure. 609 00:20:00,280 --> 00:20:02,400 I mean, on the administrative side, 610 00:20:02,400 --> 00:20:07,320 you've got things like predicting patient no shows, 611 00:20:07,320 --> 00:20:10,640 optimizing hospital bed allocation, even things 612 00:20:10,640 --> 00:20:14,000 like improving the accuracy of billing and coding. 613 00:20:14,000 --> 00:20:16,120 Oh, so making all that back end stuff more efficient. 614 00:20:16,120 --> 00:20:17,120 Exactly, yeah. 615 00:20:17,120 --> 00:20:18,560 Just streamlining everything. 616 00:20:18,560 --> 00:20:19,200 OK, cool. 617 00:20:19,200 --> 00:20:21,600 What about on the patient-facing side? 618 00:20:21,600 --> 00:20:22,480 What's happening there? 619 00:20:22,480 --> 00:20:25,120 Well, there are AI-powered chatbots. 620 00:20:25,120 --> 00:20:28,800 They can answer basic questions, schedule appointments, 621 00:20:28,800 --> 00:20:32,160 even provide personalized health information. 622 00:20:32,160 --> 00:20:33,240 Oh, wow. 623 00:20:33,240 --> 00:20:36,360 So it's freeing up health care workers' time. 624 00:20:36,360 --> 00:20:37,200 Exactly. 625 00:20:37,200 --> 00:20:40,560 To focus on the things that really require that human touch. 626 00:20:40,560 --> 00:20:41,600 Exactly, yeah. 627 00:20:41,600 --> 00:20:44,040 Let them focus on the complex cases. 628 00:20:44,040 --> 00:20:44,520 OK. 629 00:20:44,520 --> 00:20:45,320 Yeah. 630 00:20:45,320 --> 00:20:47,320 And then, of course, all the clinical applications. 631 00:20:47,320 --> 00:20:50,520 We talked about diagnosis, predicting outcomes. 632 00:20:50,520 --> 00:20:53,640 But I mean, AI is being used to personalize treatment plans 633 00:20:53,640 --> 00:20:54,120 now. 634 00:20:54,120 --> 00:20:54,560 That's right. 635 00:20:54,560 --> 00:20:56,080 Monitor patients remotely. 636 00:20:56,080 --> 00:20:58,320 Even assist with surgeries. 637 00:20:58,320 --> 00:20:59,320 It's amazing. 638 00:20:59,320 --> 00:20:59,680 It is. 639 00:20:59,680 --> 00:21:00,600 It's really amazing. 640 00:21:00,600 --> 00:21:03,760 I mean, it's being woven into every aspect of health care. 641 00:21:03,760 --> 00:21:04,920 Yeah, it really is. 642 00:21:04,920 --> 00:21:07,760 It does feel like we're on the edge of a real revolution. 643 00:21:07,760 --> 00:21:08,920 It does, doesn't it? 644 00:21:08,920 --> 00:21:10,440 I mean, the possibility. 645 00:21:10,440 --> 00:21:11,160 It does? 646 00:21:11,160 --> 00:21:12,360 They really are endless. 647 00:21:12,360 --> 00:21:16,320 I mean, as this all keeps developing, who knows what we'll see? 648 00:21:16,320 --> 00:21:17,520 It's an exciting time. 649 00:21:17,520 --> 00:21:19,640 It is an exciting time to be in this field. 650 00:21:19,640 --> 00:21:19,960 Yeah. 651 00:21:19,960 --> 00:21:22,760 And it's not just about inventing something new. 652 00:21:22,760 --> 00:21:24,160 It's about doing it the right way. 653 00:21:24,160 --> 00:21:25,480 Exactly, yeah. 654 00:21:25,480 --> 00:21:26,840 Thinking through all the implications. 655 00:21:26,840 --> 00:21:28,680 And that's what's so impressive about what 656 00:21:28,680 --> 00:21:29,840 Stanford's doing, right? 657 00:21:29,840 --> 00:21:33,880 They're not just racing to create the next big AI thing. 658 00:21:33,880 --> 00:21:35,880 They're really taking the time to make sure it benefits 659 00:21:35,880 --> 00:21:36,360 everybody. 660 00:21:36,360 --> 00:21:38,840 Right, patients, doctors, the whole system. 661 00:21:38,840 --> 00:21:41,680 It's that combination, the research and the responsibility 662 00:21:41,680 --> 00:21:44,360 that makes them, well, leaders in this field. 663 00:21:44,360 --> 00:21:45,080 Absolutely. 664 00:21:45,080 --> 00:21:47,600 It's what we need more of, honestly. 665 00:21:47,600 --> 00:21:52,160 You can't just build powerful AI and then hope for the best. 666 00:21:52,160 --> 00:21:53,560 Got to think about the consequences. 667 00:21:53,560 --> 00:21:54,160 Exactly. 668 00:21:54,160 --> 00:21:56,680 Make sure it's aligned with our values, 669 00:21:56,680 --> 00:21:58,560 that it's serving a good purpose. 670 00:21:58,560 --> 00:21:59,800 Yeah, yeah. 671 00:21:59,800 --> 00:22:01,600 OK, well, we've covered a lot of ground today. 672 00:22:01,600 --> 00:22:02,320 We have. 673 00:22:02,320 --> 00:22:04,040 It's been quite a journey. 674 00:22:04,040 --> 00:22:06,360 The challenges of health care data, 675 00:22:06,360 --> 00:22:09,160 all the amazing ways AI is being used, 676 00:22:09,160 --> 00:22:12,680 I mean, diagnosis, treatment, even end of life care. 677 00:22:12,680 --> 00:22:15,480 Right, things that a few years ago we wouldn't have even 678 00:22:15,480 --> 00:22:16,000 imagined. 679 00:22:16,000 --> 00:22:16,560 Exactly. 680 00:22:16,560 --> 00:22:19,200 And we talked about the importance of having everything 681 00:22:19,200 --> 00:22:19,840 worked together. 682 00:22:19,840 --> 00:22:22,240 The technology, the people, the policies. 683 00:22:22,240 --> 00:22:23,960 The whole, yeah, the whole ecosystem. 684 00:22:23,960 --> 00:22:25,160 The ecosystem, yeah. 685 00:22:25,160 --> 00:22:27,280 And we even touched on generative AI, 686 00:22:27,280 --> 00:22:29,000 all the excitement around that. 687 00:22:29,000 --> 00:22:30,720 But also some of the reality. 688 00:22:30,720 --> 00:22:31,720 We're doing that. 689 00:22:31,720 --> 00:22:36,120 Yeah, so for our listeners, the takeaway here is, 690 00:22:36,120 --> 00:22:37,880 stay curious. 691 00:22:37,880 --> 00:22:38,720 Stay informed. 692 00:22:38,720 --> 00:22:39,640 We're getting engaged. 693 00:22:39,640 --> 00:22:41,640 Yeah, this is happening now. 694 00:22:41,640 --> 00:22:45,360 The future of health care is being shaped right in front of us. 695 00:22:45,360 --> 00:22:47,600 And it's up to all of us to make sure 696 00:22:47,600 --> 00:22:50,400 AI is used to build the kind of system we want. 697 00:22:50,400 --> 00:22:54,560 A system that's more equitable, more accessible, more human. 698 00:22:54,560 --> 00:22:54,960 Exactly. 699 00:22:54,960 --> 00:22:56,360 That's what it's all about. 700 00:22:56,360 --> 00:22:58,600 And on that note, I think it's time 701 00:22:58,600 --> 00:22:59,400 to wrap things up. 702 00:22:59,400 --> 00:23:00,640 I think so, yeah. 703 00:23:00,640 --> 00:23:02,320 But keep exploring. 704 00:23:02,320 --> 00:23:03,440 Keep asking questions. 705 00:23:03,440 --> 00:23:05,040 Keep diving deep. 706 00:23:05,040 --> 00:23:06,040 And we'll see you next time. 707 00:23:06,040 --> 00:23:28,800 See you then.