1 00:00:00,000 --> 00:00:01,640 Hey everyone and welcome back. 2 00:00:01,640 --> 00:00:05,000 We're diving deep today into generative AI 3 00:00:05,000 --> 00:00:08,960 and specifically how it's changing the game for businesses. 4 00:00:08,960 --> 00:00:09,800 Yeah. 5 00:00:09,800 --> 00:00:12,640 We've got some really interesting insights to unpack 6 00:00:12,640 --> 00:00:15,600 from some big players like Ryder, Twilio and ServiceNow. 7 00:00:15,600 --> 00:00:16,480 Right. 8 00:00:16,480 --> 00:00:18,560 And maybe even a sneak peek 9 00:00:18,560 --> 00:00:20,760 at what NVIDIA has been working on. 10 00:00:20,760 --> 00:00:22,200 So by the end of this deep dive, 11 00:00:22,200 --> 00:00:25,640 you'll know what it really takes to build AI you can trust. 12 00:00:25,640 --> 00:00:26,480 Yeah. 13 00:00:26,480 --> 00:00:27,920 And how things are changing like really fast. 14 00:00:27,920 --> 00:00:32,360 Yeah. It really is a fascinating time to be working in AI. 15 00:00:32,360 --> 00:00:35,440 There are three major trends converging right now 16 00:00:35,440 --> 00:00:38,280 that are reshaping generative AI as we know it. 17 00:00:38,280 --> 00:00:39,880 Wow. Three big trends. 18 00:00:39,880 --> 00:00:40,880 Okay. Let's break them down. 19 00:00:40,880 --> 00:00:42,520 What are the big three we need to know about? 20 00:00:42,520 --> 00:00:43,840 Well, first we're moving towards 21 00:00:43,840 --> 00:00:45,720 what are called agentic systems. 22 00:00:45,720 --> 00:00:49,640 These are AI systems that can manage like entire workflows. 23 00:00:49,640 --> 00:00:51,600 I think like multi-snap processes. 24 00:00:51,600 --> 00:00:52,440 Right. 25 00:00:52,440 --> 00:00:54,200 Almost like having a team of digital assistants 26 00:00:54,200 --> 00:00:55,040 working for you. 27 00:00:55,040 --> 00:00:55,920 Oh wow. 28 00:00:55,920 --> 00:00:58,400 This represents a huge leap in complexity 29 00:00:58,400 --> 00:01:00,560 from the single task AI that we're used to. 30 00:01:00,560 --> 00:01:02,880 So we're talking about AI that can like figure things out 31 00:01:02,880 --> 00:01:05,560 on its own, not just follow instructions that's exciting, 32 00:01:05,560 --> 00:01:07,120 but also a little intimidating. 33 00:01:07,120 --> 00:01:07,960 Absolutely. 34 00:01:07,960 --> 00:01:10,600 And the shift toward agentic systems 35 00:01:10,600 --> 00:01:12,240 goes hand in hand with the second trend. 36 00:01:12,240 --> 00:01:13,080 Right. 37 00:01:13,080 --> 00:01:14,920 The rise of bigger, more capable models. 38 00:01:14,920 --> 00:01:19,920 We're seeing models like GPT-4 Mini and Gemini 1.5 Flash 39 00:01:20,600 --> 00:01:23,560 packing more power and accuracy than ever before. 40 00:01:23,560 --> 00:01:24,400 Wow. 41 00:01:24,400 --> 00:01:26,520 All while becoming more cost effective. 42 00:01:26,520 --> 00:01:31,520 And this is really crucial because it means more businesses 43 00:01:31,520 --> 00:01:35,320 can access and utilize these advanced AI systems. 44 00:01:35,320 --> 00:01:36,560 Yeah, that makes sense. 45 00:01:36,560 --> 00:01:38,920 What good is amazing AI if it's so expensive 46 00:01:38,920 --> 00:01:40,200 no one can afford to use it, right? 47 00:01:40,200 --> 00:01:41,040 Exactly. 48 00:01:41,040 --> 00:01:42,320 Okay, but let's talk about the people 49 00:01:42,320 --> 00:01:44,200 who are actually building these systems. 50 00:01:44,200 --> 00:01:45,040 Yeah. 51 00:01:45,040 --> 00:01:46,200 Who are the key players now? 52 00:01:46,200 --> 00:01:47,840 That's where the third trend comes in. 53 00:01:47,840 --> 00:01:48,680 Okay. 54 00:01:48,680 --> 00:01:50,920 The rise of the software engineer. 55 00:01:50,920 --> 00:01:53,800 Traditionally, AI development has been the domain 56 00:01:53,800 --> 00:01:54,880 of data scientists. 57 00:01:54,880 --> 00:01:55,720 Right. 58 00:01:55,720 --> 00:01:58,720 Now we're seeing software engineers taking a leading role, 59 00:01:58,720 --> 00:02:01,800 especially when it comes to building these agentic systems. 60 00:02:01,800 --> 00:02:04,920 They're driving the adoption of languages like TypeScript, 61 00:02:04,920 --> 00:02:08,200 which works alongside Python to manage the complexities 62 00:02:08,200 --> 00:02:09,880 of these AI agents. 63 00:02:09,880 --> 00:02:10,720 I see. 64 00:02:10,720 --> 00:02:12,600 It's becoming a real collaborative effort. 65 00:02:12,600 --> 00:02:13,440 Okay. 66 00:02:13,440 --> 00:02:15,720 Here's where things get really interesting for me. 67 00:02:15,720 --> 00:02:16,560 Okay. 68 00:02:16,560 --> 00:02:17,520 The ROI piece. 69 00:02:17,520 --> 00:02:18,360 Yeah. 70 00:02:18,360 --> 00:02:20,680 All of this AI advancement is great. 71 00:02:20,680 --> 00:02:21,520 Right. 72 00:02:21,520 --> 00:02:24,440 But how are companies actually making money with it? 73 00:02:24,440 --> 00:02:26,200 That's the million dollar question. 74 00:02:26,200 --> 00:02:27,040 Yeah. 75 00:02:27,040 --> 00:02:29,480 And frankly, it's one that many businesses 76 00:02:29,480 --> 00:02:30,560 are still struggling with. 77 00:02:30,560 --> 00:02:31,880 Take banks, for example. 78 00:02:31,880 --> 00:02:32,720 Okay. 79 00:02:32,720 --> 00:02:35,400 A lot of them have generative AI in production, 80 00:02:35,400 --> 00:02:38,240 but almost none are seeing a positive return 81 00:02:38,240 --> 00:02:39,440 on their investment yet. 82 00:02:39,440 --> 00:02:40,520 Wow, that's surprising. 83 00:02:40,520 --> 00:02:41,360 Yeah. 84 00:02:41,360 --> 00:02:43,200 You'd think AI would be like a no-brainer 85 00:02:43,200 --> 00:02:45,360 for optimizing financial processes. 86 00:02:45,360 --> 00:02:46,280 What's holding them back? 87 00:02:46,280 --> 00:02:48,120 Well, there are a few key factors. 88 00:02:48,120 --> 00:02:49,840 Integrating these complex systems 89 00:02:49,840 --> 00:02:52,960 into existing infrastructure is a dig hurdle. 90 00:02:52,960 --> 00:02:53,800 Okay. 91 00:02:53,800 --> 00:02:56,880 Then there's the need for really robust security 92 00:02:56,880 --> 00:02:59,080 and governance, which is no small feat. 93 00:02:59,080 --> 00:02:59,920 Yeah. 94 00:02:59,920 --> 00:03:01,840 But maybe the biggest challenge is figuring out 95 00:03:01,840 --> 00:03:05,080 how to measure the intangible benefits of generative AI. 96 00:03:05,080 --> 00:03:07,840 It's not always easy to quantify things like 97 00:03:07,840 --> 00:03:10,960 improved customer experience or better decision making. 98 00:03:10,960 --> 00:03:14,520 Okay, so where are we seeing generative AI 99 00:03:14,520 --> 00:03:17,080 deliver tangible value? 100 00:03:17,080 --> 00:03:17,920 Yeah. 101 00:03:17,920 --> 00:03:19,800 Are there any bright spots? 102 00:03:19,800 --> 00:03:21,880 There are definitely some exciting success stories 103 00:03:21,880 --> 00:03:23,040 emerging. 104 00:03:23,040 --> 00:03:27,360 One surprising area is insurance underwriting. 105 00:03:27,360 --> 00:03:30,880 Think about the sheer volume of data involved 106 00:03:30,880 --> 00:03:33,880 property records, financial statements, 107 00:03:33,880 --> 00:03:35,320 environmental reports. 108 00:03:35,320 --> 00:03:38,120 It's overwhelming for humans to process. 109 00:03:38,120 --> 00:03:40,400 Generative AI can analyze all of this data 110 00:03:40,400 --> 00:03:42,560 with incredible efficiency. 111 00:03:42,560 --> 00:03:45,000 And it's actually leading to significant improvements 112 00:03:45,000 --> 00:03:46,160 in risk prediction. 113 00:03:46,160 --> 00:03:48,560 We're talking about a 4%, 5% improvement, 114 00:03:48,560 --> 00:03:50,680 which is huge for the insurance industry. 115 00:03:50,680 --> 00:03:51,960 That's a major improvement. 116 00:03:51,960 --> 00:03:52,800 Yeah. 117 00:03:52,800 --> 00:03:54,000 But with such high stakes, 118 00:03:54,000 --> 00:03:55,840 I imagine there are also major risks 119 00:03:55,840 --> 00:03:57,360 if the AI makes a mistake. 120 00:03:57,360 --> 00:03:58,200 Absolutely. 121 00:03:58,200 --> 00:03:59,320 This highlights a crucial point. 122 00:03:59,320 --> 00:04:01,640 For generative AI to be truly valuable, 123 00:04:01,640 --> 00:04:04,240 especially in industries like finance and insurance, 124 00:04:04,240 --> 00:04:05,960 it needs to be trustworthy. 125 00:04:05,960 --> 00:04:08,080 We need systems that are not only accurate, 126 00:04:08,080 --> 00:04:10,920 but also explainable and reliable. 127 00:04:10,920 --> 00:04:13,320 Which brings us to something I've always wondered about. 128 00:04:13,320 --> 00:04:16,000 Those AI hallucinations, 129 00:04:16,000 --> 00:04:19,240 how do we make sure the AI doesn't just make stuff up? 130 00:04:19,240 --> 00:04:21,840 Hallucinations are a major roadblock, 131 00:04:21,840 --> 00:04:24,840 especially when trust is paramount. 132 00:04:24,840 --> 00:04:28,800 These are instances where the AI generates outputs 133 00:04:28,800 --> 00:04:30,760 that are just plain wrong, 134 00:04:30,760 --> 00:04:33,720 either factually incorrect or nonsensical. 135 00:04:33,720 --> 00:04:36,920 It's a problem that arises even when the AI is trained 136 00:04:36,920 --> 00:04:40,480 on good data because these large language models 137 00:04:40,480 --> 00:04:42,800 can still generate these unexpected 138 00:04:42,800 --> 00:04:45,400 and sometimes wildly inaccurate outputs. 139 00:04:45,400 --> 00:04:47,720 So even with all the progress we've made, 140 00:04:47,720 --> 00:04:51,040 we can't blindly trust the output of these AI systems. 141 00:04:51,040 --> 00:04:51,880 Exactly. 142 00:04:51,880 --> 00:04:53,560 Understanding how the system work 143 00:04:53,560 --> 00:04:55,240 and its limitations is crucial. 144 00:04:55,240 --> 00:04:58,040 We're even seeing this with AI judges systems 145 00:04:58,040 --> 00:05:00,680 designed to evaluate the outputs of other AI. 146 00:05:00,680 --> 00:05:01,600 Interesting. 147 00:05:01,600 --> 00:05:04,800 Sometimes being right isn't the same as being helpful. 148 00:05:04,800 --> 00:05:07,720 Imagine a chatbot telling a job applicant, 149 00:05:07,720 --> 00:05:09,840 no, you're a terrible fit. 150 00:05:09,840 --> 00:05:11,720 Technically accurate maybe, 151 00:05:11,720 --> 00:05:14,120 but definitely not the best user experience. 152 00:05:14,120 --> 00:05:15,160 That's a good point. 153 00:05:15,160 --> 00:05:19,360 So how do we move towards more reliable AI, 154 00:05:19,360 --> 00:05:21,840 especially in these high stakes scenarios? 155 00:05:21,840 --> 00:05:23,600 That's where companies like Galileo 156 00:05:23,600 --> 00:05:25,720 are doing some fascinating work. 157 00:05:25,720 --> 00:05:30,240 Their Luna Evaluation Suite is tackling this problem head on. 158 00:05:30,240 --> 00:05:34,520 They're developing specialized metrics and tools 159 00:05:34,520 --> 00:05:38,240 that go beyond simple accuracy to assess things 160 00:05:38,240 --> 00:05:41,920 like context adherence, instruction following, 161 00:05:41,920 --> 00:05:43,280 and even bias detection. 162 00:05:43,280 --> 00:05:45,960 It's like building a whole immune system for AI 163 00:05:45,960 --> 00:05:48,720 to make it more robust and less prone to errors. 164 00:05:48,720 --> 00:05:49,960 I love that analogy. 165 00:05:49,960 --> 00:05:51,040 It sounds like they're trying to create 166 00:05:51,040 --> 00:05:55,280 a more nuanced understanding of what good AI looks like, 167 00:05:55,280 --> 00:05:57,520 not just focusing on getting the right answer, 168 00:05:57,520 --> 00:05:59,240 but also on making sure the process 169 00:05:59,240 --> 00:06:00,400 and the output are trustworthy. 170 00:06:00,400 --> 00:06:01,360 That's exactly right. 171 00:06:01,360 --> 00:06:04,360 And it's essential because as AI becomes more powerful 172 00:06:04,360 --> 00:06:06,280 and integrated into our lives, 173 00:06:06,280 --> 00:06:09,560 the consequences of errors become much more significant. 174 00:06:09,560 --> 00:06:12,480 We need to make sure we're building sisters we can rely on. 175 00:06:12,480 --> 00:06:15,880 Speaking of powerful AI, let's talk about AI agents. 176 00:06:15,880 --> 00:06:17,320 They sound straight out of science fiction. 177 00:06:17,320 --> 00:06:18,280 Kind of, yeah. 178 00:06:18,280 --> 00:06:20,120 What makes them different from the AI 179 00:06:20,120 --> 00:06:21,600 we've been talking about so far? 180 00:06:21,600 --> 00:06:26,280 What sets agents apart is their ability to act independently. 181 00:06:26,280 --> 00:06:29,640 They're not just following pre-programmed instructions. 182 00:06:29,640 --> 00:06:32,120 They can make decisions and take actions 183 00:06:32,120 --> 00:06:34,040 to achieve specific goals. 184 00:06:34,040 --> 00:06:37,040 Think of it like giving AI its own agency. 185 00:06:37,040 --> 00:06:40,080 It's a huge leap forward in terms of capability 186 00:06:40,080 --> 00:06:41,000 and complexity. 187 00:06:41,000 --> 00:06:42,560 Okay, so we're talking about AI 188 00:06:42,560 --> 00:06:45,480 that can actually make choices and act on them. 189 00:06:45,480 --> 00:06:48,520 That's pretty mind blowing. 190 00:06:48,520 --> 00:06:50,800 But that must make them a lot harder to control, right? 191 00:06:50,800 --> 00:06:53,400 It certainly adds a new layer of complexity. 192 00:06:53,400 --> 00:06:54,920 That's why companies like Rider 193 00:06:54,920 --> 00:06:58,480 are focusing on building full stack AI platforms 194 00:06:58,480 --> 00:07:01,200 that offer a balance of power and control. 195 00:07:01,200 --> 00:07:03,480 They're essentially making it easier for businesses 196 00:07:03,480 --> 00:07:05,840 to leverage this powerful technology 197 00:07:05,840 --> 00:07:07,760 safely and effectively. 198 00:07:07,760 --> 00:07:09,960 So let's get down to brass tacks here. 199 00:07:09,960 --> 00:07:12,440 What are some examples of how these AI agents 200 00:07:12,440 --> 00:07:13,600 are being used today? 201 00:07:13,600 --> 00:07:15,560 What kinds of problems are they solving? 202 00:07:15,560 --> 00:07:19,000 One common application is customer service chatbots. 203 00:07:19,000 --> 00:07:19,840 Okay. 204 00:07:19,840 --> 00:07:22,240 Agents can handle a wider range of inquiries 205 00:07:22,240 --> 00:07:26,320 and even personalize responses based on customer history. 206 00:07:26,320 --> 00:07:27,800 But there's still work to be done 207 00:07:27,800 --> 00:07:29,600 on making the handoff to human agents 208 00:07:29,600 --> 00:07:31,080 seamless when needed. 209 00:07:31,080 --> 00:07:32,560 Right, because sometimes you just need 210 00:07:32,560 --> 00:07:33,880 to talk to a real person. 211 00:07:33,880 --> 00:07:34,840 Exactly. 212 00:07:34,840 --> 00:07:37,240 What about other areas where agents are making an impact? 213 00:07:37,240 --> 00:07:39,360 Code generation is another area 214 00:07:39,360 --> 00:07:42,240 where we're seeing a lot of potential. 215 00:07:42,240 --> 00:07:47,000 Agents can generate code that's not only syntactically correct, 216 00:07:47,000 --> 00:07:49,480 but also contextually relevant. 217 00:07:49,480 --> 00:07:50,320 I see. 218 00:07:50,320 --> 00:07:52,360 However, there are still challenges 219 00:07:52,360 --> 00:07:54,480 in making sure the generated code 220 00:07:54,480 --> 00:07:56,480 truly meets the developer's intent 221 00:07:56,480 --> 00:07:58,960 and fits seamlessly into the larger project. 222 00:07:58,960 --> 00:07:59,800 That makes sense. 223 00:07:59,800 --> 00:08:01,120 It's one thing to generate code. 224 00:08:01,120 --> 00:08:03,760 It's another to generate code that's actually useful 225 00:08:03,760 --> 00:08:05,120 and solves the right problem. 226 00:08:05,120 --> 00:08:05,960 Exactly. 227 00:08:05,960 --> 00:08:07,840 Are there any companies tackling this challenge 228 00:08:07,840 --> 00:08:10,000 in particularly innovative ways? 229 00:08:10,000 --> 00:08:12,120 Twilio is doing some really interesting work 230 00:08:12,120 --> 00:08:14,160 with what they call dynamic pipelines. 231 00:08:14,160 --> 00:08:15,000 Okay. 232 00:08:15,000 --> 00:08:17,440 They're using AI agents to automate tasks, 233 00:08:17,440 --> 00:08:20,440 like selecting the best data chunking method 234 00:08:20,440 --> 00:08:24,840 and choosing the optimal embedding model for a given task. 235 00:08:24,840 --> 00:08:26,200 Hold on, I'm trying to picture this. 236 00:08:26,200 --> 00:08:30,200 So the AI is making decisions about how to process data 237 00:08:30,200 --> 00:08:32,680 before it even starts working on the actual task. 238 00:08:32,680 --> 00:08:33,520 Exactly. 239 00:08:33,520 --> 00:08:36,080 It's about optimizing the entire workflow, 240 00:08:36,080 --> 00:08:37,800 or not just individual steps. 241 00:08:37,800 --> 00:08:38,640 I see. 242 00:08:38,640 --> 00:08:41,280 This can lead to significant gains in efficiency 243 00:08:41,280 --> 00:08:43,080 and accuracy. 244 00:08:43,080 --> 00:08:47,480 Instead of a human having to manually configure these settings, 245 00:08:47,480 --> 00:08:50,000 the AI agent can learn and adapt 246 00:08:50,000 --> 00:08:52,760 to choose the best approach for each situation. 247 00:08:52,760 --> 00:08:54,880 So it's like having an AI project manager 248 00:08:54,880 --> 00:08:56,400 overseeing everything 249 00:08:56,400 --> 00:08:58,360 and making sure the project runs smoothly. 250 00:08:58,360 --> 00:08:59,520 That's a great way to put it. 251 00:08:59,520 --> 00:09:02,520 And this approach of using AI for optimization 252 00:09:02,520 --> 00:09:04,440 isn't limited to Twilio. 253 00:09:04,440 --> 00:09:07,360 It's becoming increasingly common in other areas as well. 254 00:09:07,360 --> 00:09:10,360 Yeah, I'm really starting to see how this agentic AI 255 00:09:10,360 --> 00:09:12,440 can take things to a whole new level. 256 00:09:12,440 --> 00:09:15,160 What other areas are ripe for this kind of disruption? 257 00:09:15,160 --> 00:09:17,400 Data analytics is a prime example. 258 00:09:17,400 --> 00:09:21,400 Imagine an AI agent that can sift through mountains of data, 259 00:09:21,400 --> 00:09:23,000 understand the business context, 260 00:09:23,000 --> 00:09:25,600 and then generate actionable insights. 261 00:09:25,600 --> 00:09:28,240 That's a game changer for decision making. 262 00:09:28,240 --> 00:09:29,400 Wow, that would be incredible. 263 00:09:29,400 --> 00:09:31,080 Instead of just drowning in data, 264 00:09:31,080 --> 00:09:33,720 you'd have an AI partner helping you make sense of it all. 265 00:09:33,720 --> 00:09:34,400 Exactly. 266 00:09:34,400 --> 00:09:37,280 And this ties into another fascinating development, 267 00:09:37,280 --> 00:09:40,440 the emergence of autonomous agents. 268 00:09:40,440 --> 00:09:44,840 These are AI systems that can actually learn and adapt 269 00:09:44,840 --> 00:09:48,800 over time, becoming more efficient and effective 270 00:09:48,800 --> 00:09:51,520 without constant human intervention. 271 00:09:51,520 --> 00:09:54,200 Okay, now that's a bit mind blowing. 272 00:09:54,200 --> 00:09:57,600 So the AI is basically learning how to learn. 273 00:09:57,600 --> 00:09:58,800 Doesn't that make it even harder 274 00:09:58,800 --> 00:10:00,120 to predict what it will do? 275 00:10:00,120 --> 00:10:02,520 It does introduce a new level of complexity. 276 00:10:02,520 --> 00:10:04,680 That's why the focus on trust and explainability 277 00:10:04,680 --> 00:10:06,320 is more important than ever. 278 00:10:06,320 --> 00:10:09,600 We need to develop AI systems that not only demonstrate 279 00:10:09,600 --> 00:10:12,920 their capabilities, but also provide clear insights 280 00:10:12,920 --> 00:10:15,080 into their decision making processes. 281 00:10:15,080 --> 00:10:17,760 If we can't understand how an AI system arrived 282 00:10:17,760 --> 00:10:19,360 at a particular conclusion, 283 00:10:19,360 --> 00:10:22,280 how can we trust it to make critical decisions? 284 00:10:22,280 --> 00:10:23,760 That makes perfect sense. 285 00:10:23,760 --> 00:10:25,640 It's not just about the AI being ripe, 286 00:10:25,640 --> 00:10:29,240 but also about us humans understanding why it's ripe. 287 00:10:29,240 --> 00:10:30,160 Precisely. 288 00:10:30,160 --> 00:10:33,160 And this becomes even more critical 289 00:10:33,160 --> 00:10:35,760 as we move towards more complex systems 290 00:10:35,760 --> 00:10:37,840 with multiple agents working together. 291 00:10:37,840 --> 00:10:38,960 Multi-agent systems. 292 00:10:38,960 --> 00:10:40,120 Yes. 293 00:10:40,120 --> 00:10:41,720 Okay, you've officially lost me. 294 00:10:41,720 --> 00:10:42,840 What does that even mean? 295 00:10:42,840 --> 00:10:46,560 Imagine a network of AI agents 296 00:10:46,560 --> 00:10:50,920 each with its own specialized skills and knowledge 297 00:10:50,920 --> 00:10:53,080 collaborating to solve a problem. 298 00:10:53,080 --> 00:10:53,920 I see. 299 00:10:53,920 --> 00:10:57,560 Think of it like a team of human experts working together. 300 00:10:57,560 --> 00:10:59,880 But with the speed and scalability of AI. 301 00:10:59,880 --> 00:11:01,760 Okay, I'm starting to get the picture. 302 00:11:01,760 --> 00:11:03,680 That sounds incredibly powerful, 303 00:11:03,680 --> 00:11:05,680 but also incredibly complex to manage. 304 00:11:05,680 --> 00:11:06,520 It is. 305 00:11:06,520 --> 00:11:08,760 And this complexity brings new challenges 306 00:11:08,760 --> 00:11:11,000 in terms of coordination, communication, 307 00:11:11,000 --> 00:11:13,280 and even conflict resolution among agents. 308 00:11:13,280 --> 00:11:14,120 Wow. 309 00:11:14,120 --> 00:11:16,560 It's like we're creating a whole AI society 310 00:11:16,560 --> 00:11:18,640 with its own rules and dynamics. 311 00:11:18,640 --> 00:11:20,160 That's a fascinating way to think about it. 312 00:11:20,160 --> 00:11:21,720 It sounds like we're not just building AI, 313 00:11:21,720 --> 00:11:23,720 we're building a whole new way of interacting 314 00:11:23,720 --> 00:11:25,920 with technology and solving problems. 315 00:11:25,920 --> 00:11:28,320 You're hitting on a crucial point. 316 00:11:28,320 --> 00:11:31,280 As we build increasingly intelligent 317 00:11:31,280 --> 00:11:33,600 and autonomous systems, 318 00:11:33,600 --> 00:11:35,880 we need to ensure that they align 319 00:11:35,880 --> 00:11:38,600 with our values and goals as a society. 320 00:11:38,600 --> 00:11:40,640 The ethical implications of AI 321 00:11:40,640 --> 00:11:42,800 are becoming increasingly important. 322 00:11:42,800 --> 00:11:45,360 This brings us back to that idea of trust, doesn't it? 323 00:11:45,360 --> 00:11:48,520 If we're gonna hand over more and more control to AI, 324 00:11:48,520 --> 00:11:51,160 we need to be confident that it's working for us, 325 00:11:51,160 --> 00:11:52,200 not against us. 326 00:11:52,200 --> 00:11:53,040 Absolutely. 327 00:11:53,040 --> 00:11:54,040 Trust is the foundation 328 00:11:54,040 --> 00:11:56,840 of any successful AI implementation. 329 00:11:56,840 --> 00:11:59,280 Without it, we're just building sophisticated tools 330 00:11:59,280 --> 00:12:00,560 that no one will use. 331 00:12:00,560 --> 00:12:02,160 So how do we build that trust? 332 00:12:02,160 --> 00:12:05,360 What are the key ingredients that make AI trustworthy? 333 00:12:05,360 --> 00:12:07,160 There's no single answer. 334 00:12:07,160 --> 00:12:09,600 But there are several factors that contribute 335 00:12:09,600 --> 00:12:11,320 to trustworthy AI. 336 00:12:11,320 --> 00:12:13,200 Transparency is crucial. 337 00:12:13,200 --> 00:12:16,840 Being able to understand how an AI system works, 338 00:12:16,840 --> 00:12:18,680 what data it uses, 339 00:12:18,680 --> 00:12:21,040 and how it arrives at its conclusions. 340 00:12:21,040 --> 00:12:23,240 This requires clear documentation, 341 00:12:23,240 --> 00:12:25,440 explainable models, and tools 342 00:12:25,440 --> 00:12:28,440 that allow us to audit and monitor AI behavior. 343 00:12:28,440 --> 00:12:30,520 So it's like having a glass box AI 344 00:12:30,520 --> 00:12:32,160 where we can see what's going on inside. 345 00:12:32,160 --> 00:12:33,000 Exactly. 346 00:12:33,000 --> 00:12:35,240 And robustness is another key factor. 347 00:12:35,240 --> 00:12:39,080 We need systems that can handle unexpected inputs, 348 00:12:39,080 --> 00:12:41,480 recover from errors gracefully, 349 00:12:41,480 --> 00:12:43,440 and continue to perform reliably 350 00:12:43,440 --> 00:12:45,480 even in challenging environments. 351 00:12:45,480 --> 00:12:47,760 It's about building AI that's not only smart, 352 00:12:47,760 --> 00:12:49,600 but also resilient. 353 00:12:49,600 --> 00:12:51,320 Then there's fairness. 354 00:12:51,320 --> 00:12:53,440 This means ensuring that AI systems 355 00:12:53,440 --> 00:12:57,240 don't perpetuate or amplify existing biases. 356 00:12:57,240 --> 00:12:59,320 This requires careful attention to the data 357 00:12:59,320 --> 00:13:00,920 used to train these systems, 358 00:13:00,920 --> 00:13:02,800 as well as ongoing monitoring, 359 00:13:02,800 --> 00:13:04,160 to detect and mitigate 360 00:13:04,160 --> 00:13:06,920 any unintended discriminatory outcomes. 361 00:13:06,920 --> 00:13:09,480 So it's not just about building AI that's smart. 362 00:13:09,480 --> 00:13:12,120 It's about building AI that's ethical and responsible. 363 00:13:12,120 --> 00:13:13,240 It sounds like a tall order. 364 00:13:13,240 --> 00:13:16,400 It is, and that's why human oversight remains crucial. 365 00:13:16,400 --> 00:13:19,880 We can't simply hand over all decision-making to AI. 366 00:13:19,880 --> 00:13:22,240 We need to establish clear guidelines 367 00:13:22,240 --> 00:13:25,520 and ethical frameworks and mechanisms for accountability 368 00:13:25,520 --> 00:13:27,480 to ensure that AI is used in a way 369 00:13:27,480 --> 00:13:29,400 that benefits society as a whole. 370 00:13:29,400 --> 00:13:34,360 This takes us back to that idea of AI as a team sport. 371 00:13:34,360 --> 00:13:36,920 It's not just about the technology. 372 00:13:36,920 --> 00:13:40,160 It's about humans and AI working together 373 00:13:40,160 --> 00:13:41,760 to achieve shared goals. 374 00:13:41,760 --> 00:13:42,680 You're absolutely right. 375 00:13:42,680 --> 00:13:44,560 And this collaboration is essential 376 00:13:44,560 --> 00:13:48,720 for navigating the ethical and societal implications of AI. 377 00:13:48,720 --> 00:13:52,880 As we build increasingly intelligent and autonomous systems, 378 00:13:52,880 --> 00:13:55,200 we need to ensure that they align with our value 379 00:13:55,200 --> 00:13:56,960 and goals as a society. 380 00:13:56,960 --> 00:13:59,680 This leads us nicely to the next part of our conversation, 381 00:13:59,680 --> 00:14:03,320 how to actually build and deploy AI agents effectively. 382 00:14:03,320 --> 00:14:04,840 Okay, let's get practical. 383 00:14:04,840 --> 00:14:08,120 What are the key steps involved in building an AI agent 384 00:14:08,120 --> 00:14:11,000 that can actually deliver value in the real world? 385 00:14:11,000 --> 00:14:13,440 The first step is surprisingly simple, 386 00:14:13,440 --> 00:14:14,920 but often overlooked, 387 00:14:14,920 --> 00:14:16,960 defining the problem you wanna solve. 388 00:14:16,960 --> 00:14:20,280 What specific tasks or processes do you wanna automate? 389 00:14:20,280 --> 00:14:22,280 What are the key challenges you're facing? 390 00:14:22,280 --> 00:14:25,040 And most importantly, what are the desired outcomes? 391 00:14:25,040 --> 00:14:26,960 So it's like any other project, really. 392 00:14:26,960 --> 00:14:28,560 Start with a clear understanding 393 00:14:28,560 --> 00:14:30,240 of the problem you're trying to solve 394 00:14:30,240 --> 00:14:31,880 and the goals you wanna achieve. 395 00:14:31,880 --> 00:14:35,640 Exactly, once you have a well-defined problem statement, 396 00:14:35,640 --> 00:14:38,400 you can start thinking about the specific capabilities 397 00:14:38,400 --> 00:14:39,640 your agent will need. 398 00:14:39,640 --> 00:14:42,960 Will it need to access external data sources, 399 00:14:42,960 --> 00:14:46,400 interact with other systems or APIs, 400 00:14:46,400 --> 00:14:50,000 make decisions based on complex rules or logic? 401 00:14:50,000 --> 00:14:51,720 This is where it starts to get a bit more technical, right? 402 00:14:51,720 --> 00:14:52,560 Yeah. 403 00:14:52,560 --> 00:14:53,400 We're talking about the nuts and bolts 404 00:14:53,400 --> 00:14:54,600 of building the agent itself. 405 00:14:54,600 --> 00:14:56,560 And here's where choosing the right tools 406 00:14:56,560 --> 00:14:58,280 and frameworks is crucial. 407 00:14:58,280 --> 00:15:00,240 There are many different options available. 408 00:15:00,240 --> 00:15:01,080 Right. 409 00:15:01,080 --> 00:15:02,840 Each with its own strengths and weaknesses. 410 00:15:02,840 --> 00:15:03,680 Okay. 411 00:15:03,680 --> 00:15:06,400 You'll need to consider factors like ease of use, 412 00:15:06,400 --> 00:15:10,120 scalability, security, and integration 413 00:15:10,120 --> 00:15:11,800 with your existing systems. 414 00:15:11,800 --> 00:15:13,560 Sounds like there's a lot to consider. 415 00:15:13,560 --> 00:15:14,400 There is. 416 00:15:14,400 --> 00:15:16,960 And this is where having a strong understanding 417 00:15:16,960 --> 00:15:20,320 of the AI landscape and the available technologies 418 00:15:20,320 --> 00:15:21,400 is essential. 419 00:15:21,400 --> 00:15:22,240 Okay. 420 00:15:22,240 --> 00:15:24,320 You don't need to be an expert in every detail, 421 00:15:24,320 --> 00:15:27,720 but you should have a good grasp of the key concepts. 422 00:15:27,720 --> 00:15:30,960 And the trade-offs involve in different approaches. 423 00:15:30,960 --> 00:15:34,440 It's like being a savvy consumer in the world of AI, 424 00:15:34,440 --> 00:15:36,800 knowing enough to ask the right questions 425 00:15:36,800 --> 00:15:38,400 and make informed decisions. 426 00:15:38,400 --> 00:15:39,240 That makes sense. 427 00:15:39,240 --> 00:15:41,120 So we've got our problem defined. 428 00:15:41,120 --> 00:15:42,680 We've chosen our tools. 429 00:15:42,680 --> 00:15:43,520 What's next? 430 00:15:43,520 --> 00:15:45,480 I want to remember that building an AI agent 431 00:15:45,480 --> 00:15:47,600 is an iterative process. 432 00:15:47,600 --> 00:15:50,160 You'll likely start with a simple prototype, 433 00:15:50,160 --> 00:15:53,400 test it rigorously, gather feedback, 434 00:15:53,400 --> 00:15:56,200 and then refine and improve it over time. 435 00:15:56,200 --> 00:15:58,080 It's like that classic Agile approach 436 00:15:58,080 --> 00:15:59,320 to software development, right? 437 00:15:59,320 --> 00:16:00,160 Yeah. 438 00:16:00,160 --> 00:16:01,880 Build fast, fail fast, learn fast. 439 00:16:01,880 --> 00:16:03,600 That's a great way to think about it. 440 00:16:03,600 --> 00:16:06,400 And this iterative approach is especially important 441 00:16:06,400 --> 00:16:08,760 when working with agentic systems. 442 00:16:08,760 --> 00:16:09,600 Okay. 443 00:16:09,600 --> 00:16:13,320 As their behavior can be unpredictable and complex, 444 00:16:13,320 --> 00:16:15,040 we're not just building an AI. 445 00:16:15,040 --> 00:16:17,960 We're also learning how to interact with it 446 00:16:17,960 --> 00:16:21,320 and adapt to its capabilities as it evolves. 447 00:16:21,320 --> 00:16:23,280 And this brings us back to the importance 448 00:16:23,280 --> 00:16:25,240 of human oversight. 449 00:16:25,240 --> 00:16:29,080 We can't simply set and forget these systems. 450 00:16:29,080 --> 00:16:33,000 We need to establish clear guidelines, ethical frameworks, 451 00:16:33,000 --> 00:16:35,440 and mechanisms for accountability 452 00:16:35,440 --> 00:16:39,080 to ensure that they're used responsibly and effectively. 453 00:16:39,080 --> 00:16:40,760 So it's a continuous learning process 454 00:16:40,760 --> 00:16:43,280 for both the AI and the humans involved. 455 00:16:43,280 --> 00:16:44,120 Exactly. 456 00:16:44,120 --> 00:16:44,960 You're in this together. 457 00:16:44,960 --> 00:16:45,800 Exactly. 458 00:16:45,800 --> 00:16:47,600 And this partnership is crucial 459 00:16:47,600 --> 00:16:51,640 for navigating the ethical and societal implications of AI. 460 00:16:51,640 --> 00:16:52,480 Okay. 461 00:16:52,480 --> 00:16:54,400 As we build increasingly intelligent 462 00:16:54,400 --> 00:16:56,080 and autonomous systems, 463 00:16:56,080 --> 00:16:58,640 we need to ensure that they align with our values 464 00:16:58,640 --> 00:17:00,840 and goals as a society. 465 00:17:00,840 --> 00:17:03,640 This has been an incredibly thought provoking conversation. 466 00:17:03,640 --> 00:17:06,480 I'm feeling both excited and a bit overwhelmed 467 00:17:06,480 --> 00:17:08,960 by the possibilities and the challenges ahead. 468 00:17:08,960 --> 00:17:10,400 I think that's a natural reaction. 469 00:17:10,400 --> 00:17:11,600 We're living through a time 470 00:17:11,600 --> 00:17:13,960 of profound technological change. 471 00:17:13,960 --> 00:17:16,440 And AI is at the forefront of that change. 472 00:17:16,440 --> 00:17:19,320 It's up to all of us to embrace the opportunities 473 00:17:19,320 --> 00:17:21,520 while also being mindful of the risks. 474 00:17:21,520 --> 00:17:22,360 So what's next? 475 00:17:22,360 --> 00:17:23,600 Where do we go from here? 476 00:17:23,600 --> 00:17:26,240 What does the future hold for AI agents? 477 00:17:26,240 --> 00:17:28,480 I think the next few years will be crucial 478 00:17:28,480 --> 00:17:31,400 in determining how AI shapes our world. 479 00:17:31,400 --> 00:17:35,240 We'll see continued advancements in agent technology, 480 00:17:35,240 --> 00:17:38,680 leading to more sophisticated and autonomous systems. 481 00:17:38,680 --> 00:17:42,560 But we'll also see a growing emphasis on trust, 482 00:17:42,560 --> 00:17:45,760 explainability, and ethical considerations. 483 00:17:45,760 --> 00:17:47,480 It sounds like the future of AI 484 00:17:47,480 --> 00:17:50,040 is as much about human values 485 00:17:50,040 --> 00:17:52,560 as it is about technological breakthroughs. 486 00:17:52,560 --> 00:17:55,360 But before we get too far ahead of ourselves, 487 00:17:55,360 --> 00:17:56,560 let's bring in some folks 488 00:17:56,560 --> 00:17:58,920 who are actually building and deploying these systems 489 00:17:58,920 --> 00:18:00,400 in the real world. 490 00:18:00,400 --> 00:18:01,960 We've talked a lot about the theory 491 00:18:01,960 --> 00:18:05,280 and the big picture of AI agents. 492 00:18:05,280 --> 00:18:07,040 Now let's get down to the nitty gritty 493 00:18:07,040 --> 00:18:08,680 with some folks who are actually building 494 00:18:08,680 --> 00:18:11,160 and deploying these systems in the real world. 495 00:18:11,160 --> 00:18:13,360 We've got an awesome panel joining us today. 496 00:18:13,360 --> 00:18:16,400 Mimit from Service Titan, Vinnie from Tilly-O, 497 00:18:16,400 --> 00:18:18,720 and Grant from Indeed's AI Platform team. 498 00:18:18,720 --> 00:18:19,720 What come, everyone? 499 00:18:19,720 --> 00:18:22,120 Yeah, I'm really excited to hear from our panelists 500 00:18:22,120 --> 00:18:24,280 about what's working, what's challenging, 501 00:18:24,280 --> 00:18:27,080 and what the future holds for AI agents. 502 00:18:27,080 --> 00:18:28,560 Mimit, let's start with you. 503 00:18:28,560 --> 00:18:31,320 Service Titan is tackling some complex problems 504 00:18:31,320 --> 00:18:33,440 in the home services industry. 505 00:18:33,440 --> 00:18:35,080 What are some key lessons you've learned 506 00:18:35,080 --> 00:18:38,040 about building AI agents that actually deliver value 507 00:18:38,040 --> 00:18:39,320 in a field like that? 508 00:18:39,320 --> 00:18:41,600 One of the biggest lessons has been the importance 509 00:18:41,600 --> 00:18:44,080 of really understanding the nuances 510 00:18:44,080 --> 00:18:45,840 of our specific domain. 511 00:18:45,840 --> 00:18:47,920 We're dealing with a ton of unstructured data 512 00:18:47,920 --> 00:18:51,680 in home services, technician notes, customer conversations, 513 00:18:51,680 --> 00:18:53,640 even photos of job sites. 514 00:18:53,640 --> 00:18:57,000 So you've got this wild mix of information coming in. 515 00:18:57,000 --> 00:18:59,600 How do you even begin to make sense of it all with AI? 516 00:18:59,600 --> 00:19:00,920 That's exactly the challenge. 517 00:19:00,920 --> 00:19:03,440 It's not enough to just have a powerful AI model. 518 00:19:03,440 --> 00:19:06,800 You have to tailor it to the specific needs of your industry 519 00:19:06,800 --> 00:19:08,040 and your data. 520 00:19:08,040 --> 00:19:11,080 We had to develop ways to extract meaningful insights 521 00:19:11,080 --> 00:19:13,400 from all that messy, unstructured data, 522 00:19:13,400 --> 00:19:16,360 and that took a lot of experimentation and fine-tuning. 523 00:19:16,360 --> 00:19:18,920 Vinnie over at Twilio, you're all about communication. 524 00:19:18,920 --> 00:19:22,680 How are you seeing AI agents changing the game in that space? 525 00:19:22,680 --> 00:19:25,120 Oh, it's been a complete transformation. 526 00:19:25,120 --> 00:19:28,040 We're seeing AI agents being used to personalize messages, 527 00:19:28,040 --> 00:19:31,520 automate customer support, even generate marketing content. 528 00:19:31,520 --> 00:19:33,720 But one of the biggest hurdles we face 529 00:19:33,720 --> 00:19:37,400 is making sure these interactions feel natural and engaging. 530 00:19:37,400 --> 00:19:39,240 Nobody wants to talk to a robot, right? 531 00:19:39,240 --> 00:19:39,760 Definitely not. 532 00:19:39,760 --> 00:19:41,760 It's got to feel human and relatable. 533 00:19:41,760 --> 00:19:43,760 How do you achieve that with AI? 534 00:19:43,760 --> 00:19:46,400 That's where natural language processing and conversational 535 00:19:46,400 --> 00:19:47,280 AI come in. 536 00:19:47,280 --> 00:19:49,320 We're investing a lot in these areas 537 00:19:49,320 --> 00:19:52,200 to make AI interactions more seamless and effective. 538 00:19:52,200 --> 00:19:55,120 It's all about understanding the subtleties of language, tone, 539 00:19:55,120 --> 00:19:56,560 and context. 540 00:19:56,560 --> 00:19:59,160 Grant, over at Indeed, you're in the business of connecting 541 00:19:59,160 --> 00:20:00,800 people with jobs. 542 00:20:00,800 --> 00:20:03,240 How are you leveraging AI agents to make that process 543 00:20:03,240 --> 00:20:04,720 smarter and more efficient? 544 00:20:04,720 --> 00:20:07,200 AI is playing a huge role in how we match candidates 545 00:20:07,200 --> 00:20:08,760 with the right opportunities. 546 00:20:08,760 --> 00:20:11,560 We're using agents to personalize job recommendations, 547 00:20:11,560 --> 00:20:13,400 help recruiters identify top talent, 548 00:20:13,400 --> 00:20:15,520 and even provide insights to job seekers 549 00:20:15,520 --> 00:20:17,680 to help them improve their applications. 550 00:20:17,680 --> 00:20:20,880 It sounds like AI is touching almost every aspect of the job 551 00:20:20,880 --> 00:20:22,800 search process. 552 00:20:22,800 --> 00:20:24,520 But I imagine there are also challenges 553 00:20:24,520 --> 00:20:27,680 in ensuring that these systems are fair and unbiased. 554 00:20:27,680 --> 00:20:29,520 How do you approach that at Indeed? 555 00:20:29,520 --> 00:20:31,800 It's definitely a top priority for us. 556 00:20:31,800 --> 00:20:34,720 We have a dedicated team focused on responsible AI, 557 00:20:34,720 --> 00:20:37,000 and we're constantly working to ensure our systems don't 558 00:20:37,000 --> 00:20:39,960 perpetuate or amplify existing biases. 559 00:20:39,960 --> 00:20:43,080 We're committed to transparency, giving users control 560 00:20:43,080 --> 00:20:47,320 over their data, and making sure our AI is used ethically. 561 00:20:47,320 --> 00:20:49,440 Mehmet, you mentioned earlier the importance of managing 562 00:20:49,440 --> 00:20:51,560 expectations around AI. 563 00:20:51,560 --> 00:20:54,160 What are some common misconceptions you've encountered? 564 00:20:54,160 --> 00:20:55,680 One of the biggest misconceptions 565 00:20:55,680 --> 00:20:59,240 is that AI is like a magic bullet that can solve any problem. 566 00:20:59,240 --> 00:21:01,560 It's important to remember that AI is a tool, 567 00:21:01,560 --> 00:21:03,880 and like any tool, it has its limitations. 568 00:21:03,880 --> 00:21:06,560 We need to be realistic about what AI can and cannot do, 569 00:21:06,560 --> 00:21:09,200 and focus on using it to augment human capabilities, 570 00:21:09,200 --> 00:21:10,880 not replace them entirely. 571 00:21:10,880 --> 00:21:12,080 That's a great point. 572 00:21:12,080 --> 00:21:13,720 AI isn't about replacing humans. 573 00:21:13,720 --> 00:21:16,120 It's about empowering them to do their jobs better. 574 00:21:16,120 --> 00:21:18,440 This has been such a rich discussion. 575 00:21:18,440 --> 00:21:21,000 It's inspiring to see how thoughtfully our panelists are 576 00:21:21,000 --> 00:21:22,680 approaching AI development. 577 00:21:22,680 --> 00:21:24,040 Absolutely. 578 00:21:24,040 --> 00:21:26,120 It's clear that building trustworthy AI 579 00:21:26,120 --> 00:21:28,840 requires more than just technical expertise. 580 00:21:28,840 --> 00:21:32,440 It demands a deep understanding of human needs, values, 581 00:21:32,440 --> 00:21:34,720 and ethics. 582 00:21:34,720 --> 00:21:37,200 Any final thoughts from our panelists 583 00:21:37,200 --> 00:21:39,600 on what the future holds for AI agents? 584 00:21:39,600 --> 00:21:42,800 I think we'll see a shift from single-purpose agents 585 00:21:42,800 --> 00:21:45,480 to more versatile systems that can handle 586 00:21:45,480 --> 00:21:46,960 a wider range of tasks. 587 00:21:46,960 --> 00:21:49,320 Imagine an agent that can help you schedule appointments, 588 00:21:49,320 --> 00:21:51,520 book travel, manage your finances, 589 00:21:51,520 --> 00:21:53,040 all while learning your preferences 590 00:21:53,040 --> 00:21:54,520 and adapting to your needs. 591 00:21:54,520 --> 00:21:56,960 That sounds like the ultimate personal assistant. 592 00:21:56,960 --> 00:21:59,000 I agree. We'll also see a greater emphasis 593 00:21:59,000 --> 00:22:01,240 on collaboration between agents. 594 00:22:01,240 --> 00:22:02,960 Instead of isolated systems, we'll 595 00:22:02,960 --> 00:22:05,840 see agents working together to solve complex problems, 596 00:22:05,840 --> 00:22:07,720 much like teams of humans do today. 597 00:22:07,720 --> 00:22:10,320 So we're talking about a whole ecosystem of AI agents 598 00:22:10,320 --> 00:22:11,040 working together. 599 00:22:11,040 --> 00:22:12,320 That's a pretty wild thought. 600 00:22:12,320 --> 00:22:15,160 I believe there will be a growing demand for explainable AI, 601 00:22:15,160 --> 00:22:17,640 especially as these systems become more integrated 602 00:22:17,640 --> 00:22:18,880 into our lives. 603 00:22:18,880 --> 00:22:22,160 We need to be able to understand how AI agents make decisions, 604 00:22:22,160 --> 00:22:23,480 what factors they're considering, 605 00:22:23,480 --> 00:22:25,600 especially when it comes to important things like health 606 00:22:25,600 --> 00:22:27,760 finance and legal decisions. 607 00:22:27,760 --> 00:22:28,720 That makes a lot of sense. 608 00:22:28,720 --> 00:22:31,520 Transparency and accountability are key. 609 00:22:31,520 --> 00:22:34,400 Well, we've covered a ton of ground in this deep dive. 610 00:22:34,400 --> 00:22:37,640 It's clear that the world of AI agents is evolving rapidly, 611 00:22:37,640 --> 00:22:41,320 and it's going to have a profound impact on how we live and work. 612 00:22:41,320 --> 00:22:44,040 It's been an incredible journey exploring this topic with you, 613 00:22:44,040 --> 00:22:45,760 and I'm so grateful to our panelists 614 00:22:45,760 --> 00:22:47,360 for sharing their insights. 615 00:22:47,360 --> 00:22:50,120 Thank you to our expert and our amazing panelists 616 00:22:50,120 --> 00:22:52,000 for such a fascinating conversation. 617 00:22:52,000 --> 00:22:54,920 And to you, dear listener, for joining us on this deep dive 618 00:22:54,920 --> 00:22:56,880 into the world of AI agents. 619 00:22:56,880 --> 00:22:58,800 Keep exploring, keep questioning, 620 00:22:58,800 --> 00:23:00,440 and keep pushing the boundaries of what's 621 00:23:00,440 --> 00:23:03,160 possible with this incredible technology. 622 00:23:03,160 --> 00:23:25,520 Until next time, happy learning.