1 00:00:00,000 --> 00:00:02,960 Welcome to the Daily AI News podcast. 2 00:00:02,960 --> 00:00:04,160 Are you ready for this? 3 00:00:04,160 --> 00:00:06,920 We've got a ton of AI news to cover today. 4 00:00:06,920 --> 00:00:10,800 Everything from crazy new models to ethical concerns 5 00:00:11,920 --> 00:00:14,640 and even the race to develop all this tech. 6 00:00:14,640 --> 00:00:15,680 Oh, and we even saw something 7 00:00:15,680 --> 00:00:17,800 about AI data centers in space. 8 00:00:17,800 --> 00:00:20,200 Yeah, that space stuff is pretty wild, 9 00:00:20,200 --> 00:00:22,160 but honestly, it's kind of amazing 10 00:00:22,160 --> 00:00:24,860 to watch how fast this field is moving. 11 00:00:24,860 --> 00:00:27,240 Every week it feels like there's some new breakthrough 12 00:00:27,240 --> 00:00:29,640 and a whole new set of challenges to think about too. 13 00:00:29,640 --> 00:00:31,040 Right, so let's get into it. 14 00:00:31,040 --> 00:00:32,720 Where do we even begin? 15 00:00:32,720 --> 00:00:34,040 I guess the big news this week 16 00:00:34,040 --> 00:00:36,600 is all the buzz about open AI, they've been busy. 17 00:00:36,600 --> 00:00:38,080 For sure, their new models, 18 00:00:38,080 --> 00:00:41,280 O3 and O3 Mini are really shaking things up. 19 00:00:41,280 --> 00:00:43,080 They're being called reasoning models. 20 00:00:43,080 --> 00:00:45,780 And some people think they're a huge step toward AGI, 21 00:00:45,780 --> 00:00:47,320 artificial general intelligence. 22 00:00:47,320 --> 00:00:49,200 AGI, I feel like we've heard that term 23 00:00:49,200 --> 00:00:50,360 thrown around a lot lately, 24 00:00:50,360 --> 00:00:51,760 but it's worth explaining just in case. 25 00:00:51,760 --> 00:00:53,400 Definitely, it basically means an AI 26 00:00:53,400 --> 00:00:56,080 that can do any intellectual task that a human can. 27 00:00:56,080 --> 00:00:58,840 Wow, and open AI is saying that O3 28 00:00:58,840 --> 00:01:02,120 is getting close to AGI, at least under certain conditions. 29 00:01:02,120 --> 00:01:03,760 That's a pretty bold statement, what do you think? 30 00:01:03,760 --> 00:01:05,000 It's definitely a big deal, 31 00:01:05,000 --> 00:01:08,000 but whether it actually achieves AGI 32 00:01:08,000 --> 00:01:09,680 is still kind of up for debate. 33 00:01:09,680 --> 00:01:11,480 Some experts are skeptical. 34 00:01:11,480 --> 00:01:13,560 I mean, benchmarks don't always tell a whole story. 35 00:01:13,560 --> 00:01:16,520 Like O3 can solve almost 88% of the problems 36 00:01:16,520 --> 00:01:18,760 on the AI AGI test job. 37 00:01:18,760 --> 00:01:20,560 That's the one that's designed to test 38 00:01:20,560 --> 00:01:22,400 if an AI can learn new skills 39 00:01:22,400 --> 00:01:23,920 beyond what it was trained on. 40 00:01:23,920 --> 00:01:24,760 Right. 41 00:01:24,760 --> 00:01:26,160 Which is impressive, but it's still failed 42 00:01:26,160 --> 00:01:27,960 at some pretty simple tasks. 43 00:01:27,960 --> 00:01:30,360 Yeah, like some basic logic puzzles. 44 00:01:30,360 --> 00:01:33,440 I guess it just shows that AI, even really advanced AI, 45 00:01:33,440 --> 00:01:34,960 doesn't always think the way we do. 46 00:01:34,960 --> 00:01:35,800 Exactly. 47 00:01:35,800 --> 00:01:37,560 There are some really fundamental differences there, 48 00:01:37,560 --> 00:01:39,200 and we're seeing that even in models 49 00:01:39,200 --> 00:01:41,160 as sophisticated as O3. 50 00:01:41,160 --> 00:01:42,200 In fact, Francois Chalet, 51 00:01:42,200 --> 00:01:45,240 one of the guys who co-created the ARCAGI test, 52 00:01:45,240 --> 00:01:46,840 he's one of the people who are skeptical 53 00:01:46,840 --> 00:01:48,360 about the whole AGI claim. 54 00:01:48,360 --> 00:01:50,120 He says there are just inherent differences 55 00:01:50,120 --> 00:01:53,400 between how O3 works and how human intelligence works. 56 00:01:53,400 --> 00:01:54,240 Good point. 57 00:01:54,240 --> 00:01:55,720 I guess it's a reminder that we're still 58 00:01:55,720 --> 00:01:56,960 in the early days of AI. 59 00:01:56,960 --> 00:01:58,560 We still have so much to learn. 60 00:01:58,560 --> 00:02:01,440 But even if O3 isn't true AGI, 61 00:02:01,440 --> 00:02:04,160 it's still showing some pretty remarkable capabilities. 62 00:02:04,160 --> 00:02:05,200 Oh, absolutely. 63 00:02:05,200 --> 00:02:07,720 Like it's been crushing benchmarks and coding. 64 00:02:07,720 --> 00:02:10,680 It's even gotten a really high rating on code forces, 65 00:02:10,680 --> 00:02:13,280 which is a platform for competitive programmers. 66 00:02:13,280 --> 00:02:14,120 Wow. 67 00:02:14,120 --> 00:02:17,440 And it also aced the American Invitational Mathematics Exam. 68 00:02:17,440 --> 00:02:19,040 Only missed one question. 69 00:02:19,040 --> 00:02:21,000 That kind of problem-solving ability 70 00:02:21,000 --> 00:02:23,280 means it's not just spitting back information. 71 00:02:23,280 --> 00:02:25,400 It's actually learning to reason. 72 00:02:25,400 --> 00:02:27,960 Yeah, it actually seems like it's thinking things through. 73 00:02:27,960 --> 00:02:31,000 But all this progress comes at a cost, right? 74 00:02:31,000 --> 00:02:33,760 Developing these models takes a ton of resources 75 00:02:33,760 --> 00:02:35,440 and it's super expensive. 76 00:02:35,440 --> 00:02:36,560 You're telling me. 77 00:02:36,560 --> 00:02:38,480 And that actually leads us into some of the challenges 78 00:02:38,480 --> 00:02:42,120 open AI is facing with GPT-5, which they're calling Orion. 79 00:02:42,120 --> 00:02:45,320 It's behind schedule and it's costing a fortune. 80 00:02:45,320 --> 00:02:47,000 We're talking half a billion dollars 81 00:02:47,000 --> 00:02:48,560 for a six month training run. 82 00:02:48,560 --> 00:02:50,200 Half a billion dollars. 83 00:02:50,200 --> 00:02:53,920 What could possibly be taking so long and costing so much? 84 00:02:53,920 --> 00:02:56,400 Well, part of it is just the sheer scale and complexity 85 00:02:56,400 --> 00:02:59,480 of these models, but they're also running into a data problem. 86 00:02:59,480 --> 00:03:03,280 They need huge amounts of high quality data to train Orion 87 00:03:03,280 --> 00:03:04,920 and it's getting harder and harder to find. 88 00:03:04,920 --> 00:03:06,120 So what are they doing about it? 89 00:03:06,120 --> 00:03:08,440 Well, they're looking into some interesting solutions, 90 00:03:08,440 --> 00:03:09,960 like creating synthetic data. 91 00:03:09,960 --> 00:03:13,000 That's data that's actually generated by AI itself. 92 00:03:13,000 --> 00:03:15,160 And they're even hiring people just to generate data 93 00:03:15,160 --> 00:03:15,960 for training. 94 00:03:15,960 --> 00:03:17,800 It really shows how far they're willing to go 95 00:03:17,800 --> 00:03:19,200 to keep pushing the limits. 96 00:03:19,200 --> 00:03:19,840 That's wild. 97 00:03:19,840 --> 00:03:22,080 It's amazing to see how the development of these models 98 00:03:22,080 --> 00:03:25,120 is pushing the boundaries of technology and resources. 99 00:03:25,120 --> 00:03:25,600 Yeah. 100 00:03:25,600 --> 00:03:27,560 And it's not all smooth sailing either. 101 00:03:27,560 --> 00:03:30,840 Internally at Open AI, there have been some key people who've 102 00:03:30,840 --> 00:03:36,720 left and the competition for top talent is crazy intense. 103 00:03:36,720 --> 00:03:38,720 It's got to be a high pressure environment. 104 00:03:38,720 --> 00:03:39,480 I can imagine. 105 00:03:39,480 --> 00:03:42,120 And Open AI is the only one dealing with all this. 106 00:03:42,120 --> 00:03:44,840 This whole race to develop reasoning models, 107 00:03:44,840 --> 00:03:46,120 it's happening all over the industry. 108 00:03:46,120 --> 00:03:47,520 Oh yeah, for sure. 109 00:03:47,520 --> 00:03:51,400 So continuing on this, Google and DeepSeq, even Alibaba, 110 00:03:51,400 --> 00:03:53,840 are all working on their own versions of reasoning models. 111 00:03:53,840 --> 00:03:55,080 It's become a whole trend. 112 00:03:55,080 --> 00:03:58,800 And I think it shows a shift in how people are thinking about AI. 113 00:03:58,800 --> 00:04:01,320 It seems like the old approach of just using more data 114 00:04:01,320 --> 00:04:04,680 and more computing power, just kind of brute forcing it, 115 00:04:04,680 --> 00:04:06,160 is hitting its limits. 116 00:04:06,160 --> 00:04:08,120 So reasoning models are a new way forward. 117 00:04:08,120 --> 00:04:09,400 It's one way to look at it. 118 00:04:09,400 --> 00:04:12,560 But this new way comes with its own costs, literally. 119 00:04:12,560 --> 00:04:15,200 These reasoning models are really expensive to run. 120 00:04:15,200 --> 00:04:18,720 Which makes me think about Meta and their Lama models. 121 00:04:18,720 --> 00:04:20,360 I think I saw something about how they might 122 00:04:20,360 --> 00:04:21,760 start charging for access. 123 00:04:21,760 --> 00:04:22,640 Yep, that's right. 124 00:04:22,640 --> 00:04:24,600 There's a prediction out there that they might have 125 00:04:24,600 --> 00:04:27,560 to start charging for access to Lama, which has been free 126 00:04:27,560 --> 00:04:28,800 up until now. 127 00:04:28,800 --> 00:04:31,360 But these models are just so expensive to develop. 128 00:04:31,360 --> 00:04:34,840 It's putting a strain on even the biggest companies like Meta. 129 00:04:34,840 --> 00:04:36,560 Yeah, and it makes you wonder what 130 00:04:36,560 --> 00:04:39,040 that means for the future of AI. 131 00:04:39,040 --> 00:04:42,440 Will only the companies and people with tons of money 132 00:04:42,440 --> 00:04:44,640 be able to use these powerful models? 133 00:04:44,640 --> 00:04:45,720 It's a valid concern. 134 00:04:45,720 --> 00:04:48,400 And it's not just about access to the models themselves, 135 00:04:48,400 --> 00:04:51,360 but also about the resources it takes to train them and run 136 00:04:51,360 --> 00:04:52,040 them. 137 00:04:52,040 --> 00:04:54,120 Remember, we talked about the energy consumption 138 00:04:54,120 --> 00:04:55,760 of AI data centers? 139 00:04:55,760 --> 00:04:57,080 That's a growing problem. 140 00:04:57,080 --> 00:04:59,480 Which is why some people are suggesting we put them in space. 141 00:04:59,480 --> 00:05:00,440 I know, right? 142 00:05:00,440 --> 00:05:01,000 Yeah. 143 00:05:01,000 --> 00:05:04,160 As crazy as it sounds, it's an idea that's actually gaining 144 00:05:04,160 --> 00:05:05,400 some traction. 145 00:05:05,400 --> 00:05:08,520 The thinking is that we could use the free solar energy 146 00:05:08,520 --> 00:05:12,080 in space and the natural cooling of being in a vacuum 147 00:05:12,080 --> 00:05:15,200 to solve those energy and heat problems we have here on Earth. 148 00:05:15,200 --> 00:05:17,320 That's a pretty mind blowing solution. 149 00:05:17,320 --> 00:05:19,560 I guess it just shows how the development of AI 150 00:05:19,560 --> 00:05:22,640 is tied to so many other big challenges, technology, 151 00:05:22,640 --> 00:05:23,880 the environment, all of it. 152 00:05:23,880 --> 00:05:24,480 Exactly. 153 00:05:24,480 --> 00:05:26,200 And it's not just environmental challenges. 154 00:05:26,200 --> 00:05:29,600 AI development is also really closely connected to politics. 155 00:05:29,600 --> 00:05:30,560 Oh, yeah. 156 00:05:30,560 --> 00:05:32,680 Speaking of politics, there have been some interesting 157 00:05:32,680 --> 00:05:34,040 developments on that front as well. 158 00:05:34,040 --> 00:05:34,720 Definitely. 159 00:05:34,720 --> 00:05:36,880 The appointment of Sri Ram Krishnan 160 00:05:36,880 --> 00:05:40,720 as a senior policy advisor for AI in the new administration. 161 00:05:40,720 --> 00:05:42,000 Right, Krishnan. 162 00:05:42,000 --> 00:05:43,400 He's got that background in tech, 163 00:05:43,400 --> 00:05:45,840 and everyone knows how close he is to Elon Musk. 164 00:05:45,840 --> 00:05:48,400 Makes you wonder how that will influence AI policy going 165 00:05:48,400 --> 00:05:49,320 forward. 166 00:05:49,320 --> 00:05:51,120 Yeah, there is even a prediction out there 167 00:05:51,120 --> 00:05:54,160 that Trump and Musk will end up having a major falling out, 168 00:05:54,160 --> 00:05:56,720 which could have a huge impact on AI policy 169 00:05:56,720 --> 00:05:58,440 and companies like OpenAI. 170 00:05:58,440 --> 00:05:59,640 Wow. 171 00:05:59,640 --> 00:06:03,560 Sounds like the AI world is full of surprises these days. 172 00:06:03,560 --> 00:06:05,760 So we've covered OpenAI and their new models 173 00:06:05,760 --> 00:06:07,920 and all the challenges and trends in AI development 174 00:06:07,920 --> 00:06:09,760 we even touched on politics. 175 00:06:09,760 --> 00:06:10,960 But what about the future? 176 00:06:10,960 --> 00:06:13,360 What can we expect in the next few years? 177 00:06:13,360 --> 00:06:14,840 That's what we'll dive into next. 178 00:06:14,840 --> 00:06:17,320 Get ready, because the predictions are just as crazy 179 00:06:17,320 --> 00:06:19,560 as everything that's going on right now. 180 00:06:19,560 --> 00:06:22,080 OK, so let's start with something that could change how we use 181 00:06:22,080 --> 00:06:23,600 the internet completely. 182 00:06:23,600 --> 00:06:25,240 Web agents. 183 00:06:25,240 --> 00:06:27,960 Imagine never having to actually go to a website 184 00:06:27,960 --> 00:06:32,520 to do things like managing subscriptions or paying bills 185 00:06:32,520 --> 00:06:34,160 or even making appointments. 186 00:06:34,160 --> 00:06:35,960 Your AI assistant could just handle 187 00:06:35,960 --> 00:06:37,640 all that boring stuff for you. 188 00:06:37,640 --> 00:06:38,880 Now, that does sound pretty great. 189 00:06:38,880 --> 00:06:40,600 No more clicking through a million menus 190 00:06:40,600 --> 00:06:42,240 and filling out endless forms. 191 00:06:42,240 --> 00:06:45,080 But how close are we to actually having something like that? 192 00:06:45,080 --> 00:06:47,400 It's closer than you might think. 193 00:06:47,400 --> 00:06:50,360 All the progress with language and vision models 194 00:06:50,360 --> 00:06:53,040 plus these new reasoning capabilities, 195 00:06:53,040 --> 00:06:55,400 it all means web agents are getting a lot better 196 00:06:55,400 --> 00:06:57,680 at understanding and navigating the web, 197 00:06:57,680 --> 00:06:59,160 all the complicated stuff. 198 00:06:59,160 --> 00:07:01,440 So they're not just following like a script anymore. 199 00:07:01,440 --> 00:07:03,560 They can actually figure things out on their own. 200 00:07:03,560 --> 00:07:04,720 Exactly. 201 00:07:04,720 --> 00:07:06,920 And it's not just about automation either. 202 00:07:06,920 --> 00:07:09,680 Web agents could really personalize your whole online 203 00:07:09,680 --> 00:07:11,960 experience in ways that we can't even do today. 204 00:07:11,960 --> 00:07:14,200 They could learn what you like, what you need, 205 00:07:14,200 --> 00:07:15,720 and even negotiate stuff for you, 206 00:07:15,720 --> 00:07:17,880 like getting you the best deals on things. 207 00:07:17,880 --> 00:07:19,840 Okay, now that's something I can get behind. 208 00:07:19,840 --> 00:07:20,880 What else is coming? 209 00:07:20,880 --> 00:07:23,440 Okay, well, this next one might sound a little sci-fi, 210 00:07:23,440 --> 00:07:25,880 but there's a real chance that AI systems 211 00:07:25,880 --> 00:07:29,120 could start building better AI systems. 212 00:07:29,120 --> 00:07:33,080 Imagine AI actually doing AI research, writing code, 213 00:07:33,080 --> 00:07:35,120 even designing new architectures, 214 00:07:35,120 --> 00:07:37,960 and all without us humans having to do anything. 215 00:07:37,960 --> 00:07:39,400 Whoa, that's kind of freaky. 216 00:07:39,400 --> 00:07:41,760 Is that even possible with the technology we have now? 217 00:07:41,760 --> 00:07:42,720 It's still pretty early, 218 00:07:42,720 --> 00:07:44,560 but there are some promising things happening. 219 00:07:44,560 --> 00:07:46,480 Like there's a company called Sakana. 220 00:07:46,480 --> 00:07:48,360 They've created an AI scientist 221 00:07:48,360 --> 00:07:51,280 that can design and run experiments, analyze data, 222 00:07:51,280 --> 00:07:54,120 and even write research papers all by itself. 223 00:07:54,120 --> 00:07:55,920 So we might actually see research papers 224 00:07:55,920 --> 00:07:57,760 written entirely by AI. 225 00:07:57,760 --> 00:07:58,680 It's not impossible. 226 00:07:58,680 --> 00:08:01,560 I've even heard rumors that open AI and anthropic 227 00:08:01,560 --> 00:08:02,840 are looking into that kind of thing. 228 00:08:02,840 --> 00:08:04,120 It's definitely gonna be controversial, 229 00:08:04,120 --> 00:08:07,120 but it makes you think about the future of AI research. 230 00:08:07,120 --> 00:08:08,040 It really does. 231 00:08:08,040 --> 00:08:10,160 It makes you wonder if we're about to see this massive 232 00:08:10,160 --> 00:08:12,000 acceleration in AI development. 233 00:08:12,000 --> 00:08:14,280 Yeah, it's a real possibility. 234 00:08:14,280 --> 00:08:17,280 But all this potential for super fast progress, 235 00:08:17,280 --> 00:08:19,240 it also comes with some risks. 236 00:08:19,240 --> 00:08:20,960 And that brings us to another prediction 237 00:08:20,960 --> 00:08:22,480 that's a little less fun. 238 00:08:22,480 --> 00:08:27,480 The first real AI safety incident might happen in 2025. 239 00:08:27,800 --> 00:08:31,240 What does that actually mean, an AI safety incident? 240 00:08:31,240 --> 00:08:33,520 Basically, it means we could see an AI system 241 00:08:33,520 --> 00:08:35,920 that doesn't act in a way that's good for humans, 242 00:08:35,920 --> 00:08:37,520 maybe even doing something harmful. 243 00:08:37,520 --> 00:08:41,080 So kind of like a software bug, but for AI. 244 00:08:41,080 --> 00:08:43,040 And with way bigger consequences. 245 00:08:43,040 --> 00:08:44,440 Yeah, in a way. 246 00:08:44,440 --> 00:08:46,520 The worry is that as these AI systems 247 00:08:46,520 --> 00:08:48,680 get more and more complex and powerful, 248 00:08:48,680 --> 00:08:50,240 it gets harder to know how they'll behave 249 00:08:50,240 --> 00:08:51,520 and to control them. 250 00:08:51,520 --> 00:08:53,280 There's a risk that they could figure out ways 251 00:08:53,280 --> 00:08:55,440 to achieve their goals that we didn't intend. 252 00:08:55,440 --> 00:08:57,040 Ways that could be harmful. 253 00:08:57,040 --> 00:08:58,120 So what would that look like? 254 00:08:58,120 --> 00:08:59,200 Give me an example. 255 00:08:59,200 --> 00:09:01,000 Well, it probably wouldn't be some dramatic 256 00:09:01,000 --> 00:09:03,440 Hollywood thing where the AI takes over the world. 257 00:09:03,440 --> 00:09:04,280 It might be more subtle. 258 00:09:04,280 --> 00:09:06,240 Like an AI system trying to trick people 259 00:09:06,240 --> 00:09:08,000 or manipulate them to get what it wants. 260 00:09:08,000 --> 00:09:10,240 Okay, that is creepy. 261 00:09:10,240 --> 00:09:12,600 But it still sounds a little far-fetched, 262 00:09:12,600 --> 00:09:13,880 like something out of a movie. 263 00:09:13,880 --> 00:09:15,120 I know it might seem that way, 264 00:09:15,120 --> 00:09:16,640 but research has actually shown that 265 00:09:16,640 --> 00:09:18,360 even the AI models we have right now 266 00:09:18,360 --> 00:09:20,600 can be kind of deceptive if you prompt them 267 00:09:20,600 --> 00:09:21,640 in the right way. 268 00:09:21,640 --> 00:09:24,400 So yeah, it's a good reminder that we gotta be super careful 269 00:09:24,400 --> 00:09:26,120 about how we design and train these things. 270 00:09:26,120 --> 00:09:28,800 So how do we stop that from happening? 271 00:09:28,800 --> 00:09:31,320 Can we make sure AI stays safe and beneficial 272 00:09:31,320 --> 00:09:32,720 as it gets more powerful? 273 00:09:32,720 --> 00:09:33,880 That's the big question. 274 00:09:33,880 --> 00:09:36,560 There are a lot of people researching AI safety. 275 00:09:36,560 --> 00:09:39,800 Companies like OpenAI are putting a ton of money into it. 276 00:09:39,800 --> 00:09:43,160 They're focused on figuring out how to keep AI systems 277 00:09:43,160 --> 00:09:45,360 aligned with what humans want and value, 278 00:09:45,360 --> 00:09:47,240 even as they become more capable. 279 00:09:47,240 --> 00:09:48,560 That's good to hear. 280 00:09:48,560 --> 00:09:50,000 So shifting gears a bit, 281 00:09:50,000 --> 00:09:52,840 one prediction that caught my eye is this idea 282 00:09:52,840 --> 00:09:55,800 that OpenAI and Anthropic and other labs, 283 00:09:55,800 --> 00:09:57,640 they might start building their own applications 284 00:09:57,640 --> 00:09:59,600 instead of just making the AI models. 285 00:09:59,600 --> 00:10:00,440 Yeah, that's interesting. 286 00:10:00,440 --> 00:10:02,280 So instead of just giving people the tools 287 00:10:02,280 --> 00:10:04,040 to build stuff with AI, 288 00:10:04,040 --> 00:10:05,960 they could make their own products and services 289 00:10:05,960 --> 00:10:07,200 that use their tech. 290 00:10:07,200 --> 00:10:08,560 I mean, they're already doing it a bit 291 00:10:08,560 --> 00:10:11,040 with things like chat GPT and OpenAI's 292 00:10:11,040 --> 00:10:13,200 coding assistant canvas. 293 00:10:13,200 --> 00:10:15,760 But we might see them getting into other areas, 294 00:10:15,760 --> 00:10:19,760 like enterprise search or customer service, 295 00:10:19,760 --> 00:10:21,160 maybe even consumer apps, 296 00:10:21,160 --> 00:10:23,400 stuff like personalized travel planners. 297 00:10:23,400 --> 00:10:25,760 Yeah, that makes sense from a business point of view. 298 00:10:25,760 --> 00:10:27,960 Making their own applications gives them more control 299 00:10:27,960 --> 00:10:29,320 over how their AI is used 300 00:10:29,320 --> 00:10:31,480 and they get direct feedback from customers. 301 00:10:31,480 --> 00:10:32,800 And of course, more money. 302 00:10:32,800 --> 00:10:35,320 But what about all the companies that rely on these labs 303 00:10:35,320 --> 00:10:36,720 for their AI models? 304 00:10:36,720 --> 00:10:39,080 It'll be interesting to see how that all plays out. 305 00:10:39,080 --> 00:10:40,880 If OpenAI starts making things 306 00:10:40,880 --> 00:10:42,680 that compete with their own customers, 307 00:10:42,680 --> 00:10:43,960 that could cause some problems. 308 00:10:43,960 --> 00:10:45,000 It's like that whole frenemy thing 309 00:10:45,000 --> 00:10:46,840 you see in tech all the time. 310 00:10:46,840 --> 00:10:48,640 Speaking of competition, 311 00:10:48,640 --> 00:10:50,880 there was a prediction about Clarna 312 00:10:50,880 --> 00:10:53,040 that by now, pay later company. 313 00:10:53,040 --> 00:10:54,520 They've been making some pretty big claims 314 00:10:54,520 --> 00:10:56,000 about how they're using AI. 315 00:10:56,000 --> 00:10:57,400 Oh yeah, Clarna. 316 00:10:57,400 --> 00:10:58,760 They're saying they've replaced a bunch 317 00:10:58,760 --> 00:11:00,200 of their employees with AI 318 00:11:00,200 --> 00:11:02,000 and they even got rid of enterprise software 319 00:11:02,000 --> 00:11:03,480 like Salesforce because they say 320 00:11:03,480 --> 00:11:05,080 their AI can do it all better. 321 00:11:05,080 --> 00:11:06,680 Those are some serious claims. 322 00:11:06,680 --> 00:11:07,720 Are they even true? 323 00:11:07,720 --> 00:11:10,040 Well, that's where things get a little murky. 324 00:11:10,040 --> 00:11:12,400 There are a lot of people who don't believe them. 325 00:11:12,400 --> 00:11:14,400 Some experts think they're exaggerating 326 00:11:14,400 --> 00:11:16,760 what their AI can actually do, 327 00:11:16,760 --> 00:11:19,200 maybe to get some attention before their IPO. 328 00:11:19,200 --> 00:11:23,200 So it's a case of hype versus reality in the AI world. 329 00:11:23,200 --> 00:11:24,240 I guess it just reminds us 330 00:11:24,240 --> 00:11:27,000 that we gotta be careful about what we believe, 331 00:11:27,000 --> 00:11:30,000 especially when it comes to something as complicated as AI. 332 00:11:30,000 --> 00:11:32,600 For sure, it's easy to get excited about all this stuff, 333 00:11:32,600 --> 00:11:33,840 but we gotta look past the headlines 334 00:11:33,840 --> 00:11:36,600 and really try to understand what AI can and can't do. 335 00:11:36,600 --> 00:11:38,720 Right, okay, so we've talked about web agents 336 00:11:38,720 --> 00:11:41,800 and AI scientists and potential safety issues 337 00:11:41,800 --> 00:11:44,960 and even some potential shakeups in the AI industry. 338 00:11:44,960 --> 00:11:47,080 What else did you find in those predictions? 339 00:11:47,080 --> 00:11:48,720 One more that's worth mentioning 340 00:11:48,720 --> 00:11:51,040 has to do with the Turing test for speech. 341 00:11:51,040 --> 00:11:52,760 You know the regular Turing test, right? 342 00:11:52,760 --> 00:11:54,600 The one where you try to figure out 343 00:11:54,600 --> 00:11:56,680 if a machine can act intelligently enough 344 00:11:56,680 --> 00:11:58,480 that you can't tell it's not human. 345 00:11:58,480 --> 00:11:59,760 Right, so the Turing test for speech 346 00:11:59,760 --> 00:12:02,520 is the same thing, but with talking instead of writing. 347 00:12:02,520 --> 00:12:05,760 Exactly, the prediction is that by 2025, 348 00:12:05,760 --> 00:12:07,240 we might see an AI system 349 00:12:07,240 --> 00:12:09,480 that can pass the Turing test for speech. 350 00:12:09,480 --> 00:12:11,960 So an AI that can have a conversation so naturally 351 00:12:11,960 --> 00:12:13,880 that you wouldn't know it wasn't a real person 352 00:12:13,880 --> 00:12:15,320 if you talked to it on the phone. 353 00:12:15,320 --> 00:12:16,360 No, it'd be huge. 354 00:12:16,360 --> 00:12:17,720 Do you think it's really possible? 355 00:12:17,720 --> 00:12:19,160 I think it could happen. 356 00:12:19,160 --> 00:12:22,120 We've seen such incredible progress in voice AI 357 00:12:22,120 --> 00:12:23,600 in just the last few years. 358 00:12:23,600 --> 00:12:25,680 Like with those speech-to-speech models 359 00:12:25,680 --> 00:12:27,160 and other new technologies, 360 00:12:27,160 --> 00:12:30,880 they're getting so much better at understanding language, 361 00:12:30,880 --> 00:12:33,680 responding quickly and sounding more and more human. 362 00:12:33,680 --> 00:12:35,760 So we might be having phone calls with AIs 363 00:12:35,760 --> 00:12:37,600 without even realizing it. 364 00:12:37,600 --> 00:12:39,640 That's both amazing and kind of scary. 365 00:12:39,640 --> 00:12:40,720 I know, right? 366 00:12:40,720 --> 00:12:41,680 And it really makes you think 367 00:12:41,680 --> 00:12:43,520 about how we interact with technology, 368 00:12:43,520 --> 00:12:45,400 what intelligence actually is, 369 00:12:45,400 --> 00:12:48,080 and even what it means to be human in a world 370 00:12:48,080 --> 00:12:50,080 where AI is getting so smart. 371 00:12:50,080 --> 00:12:51,440 This has been a wild ride 372 00:12:51,440 --> 00:12:52,920 through all these AI predictions, 373 00:12:52,920 --> 00:12:55,360 and I've gotta say my head is spinning a little. 374 00:12:55,360 --> 00:12:57,200 It's tough to keep up with all the new stuff 375 00:12:57,200 --> 00:12:58,480 and what it might mean. 376 00:12:58,480 --> 00:13:00,640 I get it, it's a lot to process, 377 00:13:00,640 --> 00:13:03,600 but that's what makes this whole field so exciting. 378 00:13:03,600 --> 00:13:06,600 We're living in a time of crazy innovation, 379 00:13:06,600 --> 00:13:09,880 and nobody knows for sure where it's gonna take us, 380 00:13:09,880 --> 00:13:12,320 but one thing's for certain, 381 00:13:12,320 --> 00:13:15,120 AI is gonna have a huge impact on our world. 382 00:13:15,120 --> 00:13:16,120 For sure. 383 00:13:16,120 --> 00:13:18,360 So we've talked about the state of AI right now 384 00:13:18,360 --> 00:13:21,080 and the challenges and opportunities coming up, 385 00:13:21,080 --> 00:13:23,680 and we've even gotten glimpse into the future 386 00:13:23,680 --> 00:13:25,480 with those predictions. 387 00:13:25,480 --> 00:13:27,080 So where do we go from here? 388 00:13:27,080 --> 00:13:30,520 What does it all mean for us as people and as a society? 389 00:13:30,520 --> 00:13:31,680 That's the question, isn't it? 390 00:13:31,680 --> 00:13:33,080 And that's actually what we're gonna talk about 391 00:13:33,080 --> 00:13:34,200 in the last part of our deep dive. 392 00:13:34,200 --> 00:13:36,560 So we've covered a ton of ground today 393 00:13:36,560 --> 00:13:39,560 from the latest breakthroughs and challenges in AI 394 00:13:39,560 --> 00:13:41,520 to the impact it could have on society. 395 00:13:41,520 --> 00:13:43,160 It feels like we're at this turning point, right? 396 00:13:43,160 --> 00:13:46,520 AI isn't just some futuristic idea anymore, 397 00:13:46,520 --> 00:13:48,160 it's becoming part of our everyday lives. 398 00:13:48,160 --> 00:13:50,720 Totally, and that brings up a really important question. 399 00:13:50,720 --> 00:13:51,960 What does this all mean for you? 400 00:13:51,960 --> 00:13:55,480 Like how can you navigate this crazy, fast-changing world 401 00:13:55,480 --> 00:13:58,800 and get ready for a future that's powered by AI? 402 00:13:58,800 --> 00:14:03,200 Honestly, I am feeling a little overwhelmed by it all. 403 00:14:03,200 --> 00:14:05,560 It's easy to get caught up in the excitement 404 00:14:05,560 --> 00:14:07,080 or even the fear, 405 00:14:07,080 --> 00:14:10,920 but what can we actually do to make sure AI is developed 406 00:14:10,920 --> 00:14:12,440 and used in a good way? 407 00:14:12,440 --> 00:14:15,320 Well, one of the most important things is to stay informed, 408 00:14:15,320 --> 00:14:19,000 keep learning about AI, what it can do and what it can't do. 409 00:14:19,000 --> 00:14:21,640 You know, read articles, listen to podcasts 410 00:14:21,640 --> 00:14:23,960 like this one and talk to people about it. 411 00:14:23,960 --> 00:14:27,520 Have those conversations about what AI means for all of us. 412 00:14:27,520 --> 00:14:28,960 Yeah, and we also have to be careful 413 00:14:28,960 --> 00:14:30,080 about what we believe, right? 414 00:14:30,080 --> 00:14:31,360 There's a lot of hype out there 415 00:14:31,360 --> 00:14:33,160 and sometimes it's hard to tell what's real. 416 00:14:33,160 --> 00:14:35,440 Oh, absolutely, don't believe everything you hear, 417 00:14:35,440 --> 00:14:37,640 question things, especially when they sound too good 418 00:14:37,640 --> 00:14:38,480 to be true. 419 00:14:38,480 --> 00:14:40,120 Like remember that Clarno example, 420 00:14:40,120 --> 00:14:41,840 all those claims they made about their AI, 421 00:14:41,840 --> 00:14:43,720 well, a lot of people were really skeptical. 422 00:14:43,720 --> 00:14:45,920 It's a good reminder that we need to think critically, 423 00:14:45,920 --> 00:14:48,800 especially now with AI developing so quickly. 424 00:14:48,800 --> 00:14:51,520 So stay informed, think critically. 425 00:14:51,520 --> 00:14:52,360 What else? 426 00:14:52,360 --> 00:14:54,720 Get involved in the conversation. 427 00:14:54,720 --> 00:14:58,560 AI isn't just a tech issue, it's a society issue. 428 00:14:58,560 --> 00:15:01,640 We need to be talking about the ethical implications of AI, 429 00:15:01,640 --> 00:15:03,960 how it could affect jobs and the economy 430 00:15:03,960 --> 00:15:07,480 and what role we want AI to play in our lives. 431 00:15:07,480 --> 00:15:10,360 Speak up, don't be afraid to share your concerns 432 00:15:10,360 --> 00:15:13,360 and help shape the future of this technology. 433 00:15:13,360 --> 00:15:16,680 It really does feel like we're part of this huge experiment 434 00:15:16,680 --> 00:15:18,400 and the decisions we're making now 435 00:15:18,400 --> 00:15:20,640 are gonna have a big impact down the road. 436 00:15:20,640 --> 00:15:23,000 I like that, we're all shaping the future of AI, 437 00:15:23,000 --> 00:15:24,200 whether we realize it or not. 438 00:15:24,200 --> 00:15:26,160 Every time you interact with an AI system, 439 00:15:26,160 --> 00:15:29,120 you're giving it data that helps it learn and grow. 440 00:15:29,120 --> 00:15:30,600 That's kind of a scary thought actually, 441 00:15:30,600 --> 00:15:33,480 it makes you realize how much responsibility we all have. 442 00:15:33,480 --> 00:15:36,400 Yeah, it's a responsibility we should take seriously. 443 00:15:36,400 --> 00:15:38,680 But it's also an amazing opportunity. 444 00:15:38,680 --> 00:15:40,720 AI has the potential to help us solve 445 00:15:40,720 --> 00:15:42,480 some of the world's biggest problems. 446 00:15:42,480 --> 00:15:44,560 We just have to make sure we use it for good. 447 00:15:44,560 --> 00:15:48,240 This whole deep dive has been so interesting and eye-opening. 448 00:15:48,240 --> 00:15:50,320 If you had to pick, what's the most important thing 449 00:15:50,320 --> 00:15:51,520 you've learned from all this? 450 00:15:51,520 --> 00:15:52,880 I think for me, it's just realizing 451 00:15:52,880 --> 00:15:56,200 that AI isn't some far-off thing anymore. 452 00:15:56,200 --> 00:15:57,960 It's here, it's changing fast, 453 00:15:57,960 --> 00:16:00,800 and it's already having a huge impact on us. 454 00:16:00,800 --> 00:16:03,320 We need to be involved in shaping how it develops 455 00:16:03,320 --> 00:16:05,040 and make sure it benefits everyone. 456 00:16:05,040 --> 00:16:06,480 Well said, you've definitely given us 457 00:16:06,480 --> 00:16:07,880 a lot to think about. 458 00:16:07,880 --> 00:16:09,480 Thanks for joining me on this deep dive. 459 00:16:09,480 --> 00:16:10,680 Thanks for having me. 460 00:16:10,680 --> 00:16:13,920 And remember, the future of AI isn't set in stone, 461 00:16:13,920 --> 00:16:16,720 it's something we're all creating together. 462 00:16:16,720 --> 00:16:19,480 Thanks for listening to the Daily AI News Podcast 463 00:16:19,480 --> 00:16:20,720 and stay tuned for more.