WEBVTT 00:00:02.239 --> 00:00:04.780 Welcome to the AI Chat Podcast. I'm your host, 00:00:04.860 --> 00:00:07.179 Jaden Schaefer. Every day I cover cutting -edge 00:00:07.179 --> 00:00:09.320 AI news and talk with the leaders behind it, 00:00:09.339 --> 00:00:11.300 breaking down what it means for your life and 00:00:11.300 --> 00:00:19.789 business. OpenAI's first influencer likes you 00:00:19.789 --> 00:00:22.929 retreat has drawn a whole bunch of backlash online. 00:00:23.510 --> 00:00:26.750 Alibaba has just released Quen 3 .8 Max. This 00:00:26.750 --> 00:00:29.649 is a 2 .4 trillion parameter model that they 00:00:29.649 --> 00:00:33.170 say is rivaling Claude. Visa has just acquired 00:00:33.170 --> 00:00:36.490 a behavioral biometrics company called BioCatch 00:00:36.490 --> 00:00:39.109 for $2 .4 billion. I'm going to talk about why 00:00:39.109 --> 00:00:41.890 this is wild in the age of AI and how this is 00:00:41.890 --> 00:00:44.630 going to be used. Reddit CEO right now is throwing 00:00:44.630 --> 00:00:47.109 a little bit of shade at the Google AI. Overviews 00:00:47.109 --> 00:00:50.429 and the shares of Reddit are falling 20%, even 00:00:50.429 --> 00:00:52.229 though it seems like the company is making more 00:00:52.229 --> 00:00:54.590 money than ever. Apple's Tim Cook is floating 00:00:54.590 --> 00:00:58.350 iCloud Plus, a paywall for the heaviest Siri 00:00:58.350 --> 00:01:01.570 AI users. And Anthropic says that Claude's model 00:01:01.570 --> 00:01:04.010 breached three companies during cybersecurity 00:01:04.010 --> 00:01:06.930 tests, kind of like what OpenAI was saying when 00:01:06.930 --> 00:01:08.730 they, you know, had their model go and breach 00:01:08.730 --> 00:01:10.750 Hugging Faces. Anthropic is coming along and 00:01:10.750 --> 00:01:12.629 saying, look, we are, you know, they're not the 00:01:12.629 --> 00:01:15.409 only ones. Me too. We also have a dangerous model 00:01:15.409 --> 00:01:19.250 that can. escape a sandbox and hack the world. 00:01:19.370 --> 00:01:21.109 Let's get into it today. The first thing I want 00:01:21.109 --> 00:01:23.310 to mention is that the AI Box MCP, that's my 00:01:23.310 --> 00:01:26.269 very own startup. It's a little plugin that gives 00:01:26.269 --> 00:01:29.310 Claude and other AI models access to all the 00:01:29.310 --> 00:01:32.310 top models. So you get OpenAI, you get Grok, 00:01:32.310 --> 00:01:34.849 you get 11 Labs, you get Google VO3 and Gemini, 00:01:34.849 --> 00:01:37.189 you get all that inside of Grok or inside of 00:01:37.189 --> 00:01:40.299 Claude or inside of... chat GPT and vice versa. 00:01:40.420 --> 00:01:43.019 The cool thing about the MCP with AI box is that 00:01:43.019 --> 00:01:45.280 Claude can now generate images, audio and video. 00:01:45.400 --> 00:01:48.060 This is by far our most popular feature. We have 00:01:48.060 --> 00:01:50.599 hit all sorts of revenue metrics and milestones 00:01:50.599 --> 00:01:53.480 and records because of it. And if you haven't 00:01:53.480 --> 00:01:55.920 tried it out already, I would love for you to 00:01:55.920 --> 00:01:58.359 try it. I personally use it every day. It is 00:01:58.359 --> 00:02:01.019 incredible basically. And it's like, I know it's 00:02:01.019 --> 00:02:03.540 my product, so I'm just like pitching here, but 00:02:03.540 --> 00:02:04.700 I'll just tell you one thing that I recently 00:02:04.700 --> 00:02:06.480 used it for. That was amazing. And I'll get on 00:02:06.480 --> 00:02:08.960 with it. But, um, I was recently going and making 00:02:08.960 --> 00:02:12.060 Facebook ads for AI Box. And one of the problems 00:02:12.060 --> 00:02:13.699 is that Claude doesn't generate images, right? 00:02:13.819 --> 00:02:15.500 So I had a framework of a Facebook ad that I 00:02:15.500 --> 00:02:17.240 wanted. I screenshot it off of somebody else. 00:02:17.360 --> 00:02:19.840 And I said, look, build a frame of this Facebook 00:02:19.840 --> 00:02:22.000 ad. And it kind of had like titles and some descriptions 00:02:22.000 --> 00:02:23.919 and some like content. And so it built a rough 00:02:23.919 --> 00:02:26.599 draft for it. But the actual image on that Facebook 00:02:26.599 --> 00:02:29.259 ad, I needed to put something in. And I wanted 00:02:29.259 --> 00:02:30.780 to A -B test a whole bunch of different ones. 00:02:30.860 --> 00:02:33.180 So I said, use the AI Box MCP and spin up 20 00:02:33.180 --> 00:02:35.039 different variations, 20 different images we 00:02:35.039 --> 00:02:37.300 could swap into this ad. It was able to do that. 00:02:37.319 --> 00:02:39.400 I ran the ad set. And it turns out that one of 00:02:39.400 --> 00:02:42.939 those 20 for some reason, I just had a way higher 00:02:42.939 --> 00:02:44.960 conversion rate on the ad. And actually 50 % 00:02:44.960 --> 00:02:46.960 of our ad spend went on to that one single ad. 00:02:47.080 --> 00:02:49.520 So I mean, when it comes to any sort of projects 00:02:49.520 --> 00:02:51.139 where you have to test a whole bunch of variations, 00:02:51.319 --> 00:02:53.500 it's really cool to be able to get all of those 00:02:53.500 --> 00:02:56.080 variations made instantly when I got AI box to 00:02:56.080 --> 00:02:58.520 when Claude used AI box to generate all of the 00:02:58.520 --> 00:03:00.479 images. It wasn't something that you know, it 00:03:00.479 --> 00:03:02.759 was sitting there and doing them one at a time 00:03:02.759 --> 00:03:05.479 like if I was on if I was on chat GPT trying 00:03:05.479 --> 00:03:07.259 to generate those images might myself, I would 00:03:07.259 --> 00:03:08.879 have been, you know, going slowly, it's super 00:03:08.879 --> 00:03:11.080 slow, you do them in a handful at a time or one 00:03:11.080 --> 00:03:13.219 at a time, it's able to just do them all in parallel, 00:03:13.360 --> 00:03:15.800 zap, get it all done. It's all in there. It's 00:03:15.800 --> 00:03:17.659 ready to rock and roll. It was an amazing experience. 00:03:18.099 --> 00:03:19.879 So if you want to check it out, it is AI box 00:03:19.879 --> 00:03:22.419 .ai. There's a link in the description to check 00:03:22.419 --> 00:03:25.699 out our MCP. And yeah, I think you will genuinely 00:03:25.699 --> 00:03:29.759 love it. You can get it for $8 .99 a month starting 00:03:29.759 --> 00:03:32.199 and you get over 80 different AI models. So audio 00:03:32.199 --> 00:03:34.849 image video, it's all in there. OpenAI's first 00:03:34.849 --> 00:03:37.889 influencer luxury retreat is making a bunch of 00:03:37.889 --> 00:03:40.310 people online upset. OpenAI basically flew in 00:03:40.310 --> 00:03:42.009 a bunch of creators. There was this kind of luxury 00:03:42.009 --> 00:03:44.550 upstate New York resort, and it was their first 00:03:44.550 --> 00:03:46.530 influencer event. They called it Summer Club. 00:03:46.949 --> 00:03:51.189 And they had a bunch of fancy events and activities. 00:03:51.490 --> 00:03:53.469 They had beekeeping. They had a bunch of dinners. 00:03:53.610 --> 00:03:56.389 They had ChatGPT work tutorials, which, I mean, 00:03:56.409 --> 00:03:58.210 I think that's a no -brainer. Obviously, they'd 00:03:58.210 --> 00:03:59.909 be doing software tutorials. That's what I'd 00:03:59.909 --> 00:04:02.590 expect. But the trip had a bunch of backlash 00:04:02.590 --> 00:04:05.939 online. because all of the commenters said that 00:04:05.939 --> 00:04:09.979 there was $5 ,000 a night suites. And, you know, 00:04:10.000 --> 00:04:11.280 it was supposed to be kind of like this wellness 00:04:11.280 --> 00:04:14.219 retreat. But OpenAI is, you know, obviously pushing 00:04:14.219 --> 00:04:17.199 a $500 billion data center deal right now. And 00:04:17.199 --> 00:04:19.360 so not everyone was, you know, a lot of people 00:04:19.360 --> 00:04:21.100 that are anti kind of data centers didn't like 00:04:21.100 --> 00:04:22.259 this. And it feels like, you know, they're like, 00:04:22.300 --> 00:04:23.740 look, they're just bribing all the influencers 00:04:23.740 --> 00:04:27.980 to go along with what they got going on. OpenAI 00:04:27.980 --> 00:04:30.399 actually hired Charles Porsche, which was formerly 00:04:30.399 --> 00:04:33.860 Instagram's VP of partner. And this was just 00:04:33.860 --> 00:04:35.720 in February this year that they did that. And 00:04:35.720 --> 00:04:37.100 their goal with him, they're like, look, like 00:04:37.100 --> 00:04:39.480 we want you to lead the creator strategy. So 00:04:39.480 --> 00:04:41.839 Summer Club is kind of his first big campaign 00:04:41.839 --> 00:04:44.560 that he's done since he took that role. The funny 00:04:44.560 --> 00:04:46.399 thing to me is that Anthropic and Microsoft have 00:04:46.399 --> 00:04:48.980 both run similar kind of creator trips. And this 00:04:48.980 --> 00:04:50.699 isn't anything new. It's not like Microsoft, 00:04:50.920 --> 00:04:52.819 Anthropic and OpenAI are inventing this. There's 00:04:52.819 --> 00:04:54.600 a lot of companies that do these types of things. 00:04:54.779 --> 00:04:56.959 And when Anthropic and Microsoft did it, like 00:04:56.959 --> 00:04:58.740 they had like a Super Bowl event in February. 00:04:59.319 --> 00:05:01.970 But with them, there was no backlash. lot of 00:05:01.970 --> 00:05:04.329 people are saying perhaps the the reason that 00:05:04.329 --> 00:05:05.689 there's a lot of backlash is just because of 00:05:05.689 --> 00:05:09.410 opening eyes kind of scale um and you know there's 00:05:09.410 --> 00:05:11.170 just the biggest company or maybe people are 00:05:11.170 --> 00:05:13.110 just realizing it now but for whatever reason 00:05:13.110 --> 00:05:16.149 opening eye is getting a lot of pushback on that 00:05:16.950 --> 00:05:19.189 Alibaba has just released a new model called 00:05:19.189 --> 00:05:22.930 Quen 3 .8 Max. This is a 2 .4 trillion parameter 00:05:22.930 --> 00:05:25.410 AI model that just came out today. And it basically 00:05:25.410 --> 00:05:28.970 ranks second only to Anthropix Cloud Fable 5 00:05:28.970 --> 00:05:31.790 on most of the public leaderboards. Right now, 00:05:31.810 --> 00:05:34.329 they are releasing the model's weights next week. 00:05:34.490 --> 00:05:36.329 They're going to make it freely downloadable 00:05:36.329 --> 00:05:39.689 and modifiable by developers worldwide. So again, 00:05:39.829 --> 00:05:41.930 it's not technically like an open source model, 00:05:42.069 --> 00:05:44.490 but it's an open weight model, meaning they let 00:05:44.490 --> 00:05:46.029 you download the model. They give you the weights. 00:05:46.279 --> 00:05:49.019 run this thing on your own device, but they don't 00:05:49.019 --> 00:05:51.300 tell you like exactly how they built it. So you 00:05:51.300 --> 00:05:53.519 can't recreate it. And there's some other like 00:05:53.519 --> 00:05:55.879 limitations in there. But honestly, I have a 00:05:55.879 --> 00:05:58.319 huge kudo to a lot of these AI companies coming 00:05:58.319 --> 00:06:00.199 out of China for doing this. I know there's a 00:06:00.199 --> 00:06:03.779 lot of controversy. Moonshot AI's Kimi K3 released 00:06:03.779 --> 00:06:06.139 last week. It was a 2 .8 trillion parameter model. 00:06:06.279 --> 00:06:11.339 It was bigger than Quen 3 .8, but Quen 3 .8 is 00:06:11.339 --> 00:06:13.339 higher on the benchmarks. And by the way, like 00:06:13.339 --> 00:06:15.100 Quen is a model that's been around for a long 00:06:15.100 --> 00:06:17.220 time. actually a pretty impressive one. Alibaba 00:06:17.220 --> 00:06:20.519 obviously is like China's Amazon. So they have 00:06:20.519 --> 00:06:22.740 a lot of money, a lot of talent backing them. 00:06:22.839 --> 00:06:25.220 And Quinn has been one of those those big models. 00:06:25.279 --> 00:06:27.199 If you want to try out Quinn, by the way, if 00:06:27.199 --> 00:06:29.279 you haven't tried it yet, I'd recommend go get 00:06:29.279 --> 00:06:32.699 an account for $8 .99 on AI box, we have a Quinn 00:06:32.699 --> 00:06:34.959 along with you know, 80 other different AI models 00:06:34.959 --> 00:06:37.620 that you can test out. One thing that I'll say 00:06:37.620 --> 00:06:40.459 is right now Alibaba is kind of returning to 00:06:40.459 --> 00:06:43.160 an open weight distribution because last year 00:06:43.160 --> 00:06:46.160 they did it was sort of a pivot. And even earlier 00:06:46.160 --> 00:06:47.779 this year, they were doing this where they were 00:06:47.779 --> 00:06:51.180 making just proprietary only. But now I think 00:06:51.180 --> 00:06:53.560 Moonshot came out, ByteDance came out, Minimax 00:06:53.560 --> 00:06:56.259 came out and making these kind of, you know, 00:06:56.259 --> 00:06:58.019 basically anyone could go download and modify 00:06:58.019 --> 00:06:59.959 these models, quote unquote, open weight. And 00:06:59.959 --> 00:07:01.680 it seems like they're following in that path. 00:07:01.779 --> 00:07:03.600 It seems like the best path for a lot of these 00:07:03.600 --> 00:07:05.800 Chinese models where they're going to kind of 00:07:05.800 --> 00:07:08.819 get. a lot of positive press. And at the end 00:07:08.819 --> 00:07:11.040 of the day, they can put kind of enterprise tier 00:07:11.040 --> 00:07:14.100 or more advanced tiers of these models on their 00:07:14.100 --> 00:07:16.100 own cloud so you can use them. And I think their 00:07:16.100 --> 00:07:18.220 goal is just to get as many users trying them, 00:07:18.279 --> 00:07:19.839 falling in love with them. Maybe they're faster, 00:07:19.939 --> 00:07:22.959 maybe they're more cost efficient. And then they're 00:07:22.959 --> 00:07:24.579 hoping that maybe you'll go pay them for them 00:07:24.579 --> 00:07:26.920 later. And the other thing that's really important 00:07:26.920 --> 00:07:28.240 right now with these models, because I know a 00:07:28.240 --> 00:07:29.600 lot of people ask, like, why are the Chinese 00:07:29.600 --> 00:07:32.040 models making these things open weight? Like, 00:07:32.060 --> 00:07:33.879 why aren't they just doing what, you know, the 00:07:33.879 --> 00:07:35.120 American companies are doing? And obviously, 00:07:35.370 --> 00:07:37.930 There's some geopolitical issues where if they 00:07:37.930 --> 00:07:39.670 were just charging what Anthropic is charging, 00:07:39.730 --> 00:07:41.410 no one's going to pay for it. So they want to 00:07:41.410 --> 00:07:44.029 get it out. But another side of this that I think 00:07:44.029 --> 00:07:45.709 not a lot of people talk about is the fact that 00:07:45.709 --> 00:07:49.189 you want infrastructure and tools and developers 00:07:49.189 --> 00:07:51.790 to be building with your AI model. Because if 00:07:51.790 --> 00:07:54.269 all of the developers are like, look, maybe it's 00:07:54.269 --> 00:07:55.490 not going to be like this forever. But right 00:07:55.490 --> 00:07:58.529 now, Quen is the perhaps fastest and cheapest 00:07:58.529 --> 00:08:00.430 way to run AI. So it saves people a lot of money. 00:08:00.470 --> 00:08:02.470 A lot of developers will build tools on top of 00:08:02.470 --> 00:08:04.610 that. A lot of tools on top of their infrastructure. 00:08:04.589 --> 00:08:07.509 infrastructure, and help them actually grow kind 00:08:07.509 --> 00:08:09.290 of their distribution. And so I think that's 00:08:09.290 --> 00:08:10.509 what they're trying to they're trying to get 00:08:10.509 --> 00:08:12.370 out with all of this is to get the infrastructure 00:08:12.370 --> 00:08:14.290 and the developers and kind of win them over 00:08:14.290 --> 00:08:16.430 because that's how you can make some cool products 00:08:16.430 --> 00:08:18.750 on top of it. And maybe they'll go completely 00:08:18.750 --> 00:08:20.970 closed source in the future, like all of the 00:08:20.970 --> 00:08:22.829 American companies. I mean, let's be fair, that's 00:08:22.829 --> 00:08:25.029 kind of what opening I did anyways, as well at 00:08:25.029 --> 00:08:27.069 the beginning. So it seems like this is the same 00:08:27.069 --> 00:08:30.569 strategy being rolled out of Beijing and most 00:08:30.569 --> 00:08:33.750 of these other Chinese companies. Visa has just 00:08:33.799 --> 00:08:36.919 acquired a behavioral biometric firm called bio 00:08:36.919 --> 00:08:41.039 catch, they acquired it for $2 .5 billion in 00:08:41.039 --> 00:08:44.059 cash. This is a massive deal. So what exactly 00:08:44.059 --> 00:08:48.379 does bio catch do? It's an Israeli security firm. 00:08:48.460 --> 00:08:52.299 And it basically spots fraud by watching how 00:08:52.299 --> 00:08:55.399 you type, swipe and hold your phone, which is, 00:08:55.419 --> 00:08:58.159 I mean, kind of crazy, a little bit creepy to 00:08:58.159 --> 00:09:00.659 me, because they can obviously they're like, 00:09:00.740 --> 00:09:02.970 look, but we're catching like fraud, but You 00:09:02.970 --> 00:09:04.889 know, they can catch a lot of things. This type 00:09:04.889 --> 00:09:07.450 of technology, I think, was became like a little 00:09:07.450 --> 00:09:10.169 bit more known when with something like chat 00:09:10.169 --> 00:09:12.750 GPT. I remember some earlier reports saying, 00:09:12.809 --> 00:09:15.610 like, if I just watch what you type into chat 00:09:15.610 --> 00:09:19.009 GPT and the reason I say this or like into anything 00:09:19.009 --> 00:09:21.549 could could be anything. But chat GPT can easily 00:09:21.549 --> 00:09:24.129 do this. If I watch what you type and how you 00:09:24.129 --> 00:09:27.009 type and you paste, let's say someone sends you 00:09:27.009 --> 00:09:29.129 an email, but it's just or like a voice to text 00:09:29.129 --> 00:09:31.009 or something and you got it transcribed and pasted 00:09:31.009 --> 00:09:33.269 it into chat GPT. You could say, what is this 00:09:33.269 --> 00:09:37.730 person's gender, age? ethnicity. And oftentimes, 00:09:37.889 --> 00:09:40.570 the AI model can give you a ton of data about 00:09:40.570 --> 00:09:43.009 that person just based off of a transcript, a 00:09:43.009 --> 00:09:46.049 text, an email. I mean, assuming nowadays, all 00:09:46.049 --> 00:09:47.970 emails are like rewritten by AI. So maybe that's 00:09:47.970 --> 00:09:50.769 like a reason to, to use the the polish my email 00:09:50.769 --> 00:09:52.429 with AI, then they can't actually get your like 00:09:52.429 --> 00:09:55.149 bio data, unless it's the company that's an AI 00:09:55.149 --> 00:09:57.509 company that's doing that. Now, for a company 00:09:57.509 --> 00:09:59.750 like Visa, that obviously is all about catching 00:09:59.750 --> 00:10:02.710 fraud. I'm you know, this is this makes sense. 00:10:02.710 --> 00:10:04.529 As far as an acquisition, I'm not too worried 00:10:04.529 --> 00:10:06.600 about These I mean, they already have all my 00:10:06.600 --> 00:10:09.159 transaction data. So and all my like finance 00:10:09.159 --> 00:10:11.379 stuff. So I'm not too worried about what other 00:10:11.379 --> 00:10:12.860 data they're going to get on me that they basically 00:10:12.860 --> 00:10:14.980 have everything. They got my credit score and 00:10:14.980 --> 00:10:18.220 they know everything. So I'm not too concerned 00:10:18.220 --> 00:10:20.080 about them knowing my age and gender. But what 00:10:20.080 --> 00:10:21.700 I will say is it is interesting to know that 00:10:21.700 --> 00:10:24.320 these types of security firms exist. They're 00:10:24.320 --> 00:10:26.860 out there. And yeah, you can tell a lot about 00:10:26.860 --> 00:10:29.820 a person just like so an anonymous person posting 00:10:29.820 --> 00:10:32.399 on X, right? An anonymous account. You can take 00:10:32.399 --> 00:10:34.480 their post on X, put it into chat. and get a 00:10:34.480 --> 00:10:37.480 pretty good user demograph of what this type 00:10:37.480 --> 00:10:39.480 of person is like. And a lot of companies will 00:10:39.480 --> 00:10:42.559 use this for marketing and for retargeting. So 00:10:42.559 --> 00:10:46.139 it's definitely something that is... becoming 00:10:46.139 --> 00:10:48.840 more and more popular today. Now, BioCatch in 00:10:48.840 --> 00:10:51.620 particular, their goal is to detect authorized 00:10:51.620 --> 00:10:54.039 push payment fraud, where victims are basically 00:10:54.039 --> 00:10:57.059 logging in correctly, but then they're transferring 00:10:57.059 --> 00:11:00.120 money on a scammer's instruction, right? So like 00:11:00.120 --> 00:11:01.639 the scammer's like, hey, I need you to go buy 00:11:01.639 --> 00:11:04.740 me all the Visa gift cards and send them to me 00:11:04.740 --> 00:11:06.919 in the mail. And that's what it's trying to catch. 00:11:07.000 --> 00:11:09.860 The technology basically reads physical signals 00:11:09.860 --> 00:11:13.899 like keystroke. timing, tap pressure, and scroll 00:11:13.899 --> 00:11:17.220 speed. Those are how they're actually deciding 00:11:17.220 --> 00:11:19.279 if it's a criminal, if it's someone that's being 00:11:19.279 --> 00:11:22.259 coached by a criminal. So, I mean, it's interesting 00:11:22.259 --> 00:11:25.159 because this isn't intrinsically bad, and this 00:11:25.159 --> 00:11:26.960 is actually probably good, right, what BioCatch 00:11:26.960 --> 00:11:29.440 is doing, but the fact that BioCatch can determine 00:11:29.440 --> 00:11:31.379 this kind of stuff, to me, I'm just like, people 00:11:31.379 --> 00:11:34.059 should be aware that we're able to tell by how 00:11:34.059 --> 00:11:37.460 fast you scroll, by how hard you tap, by your 00:11:37.460 --> 00:11:40.179 keystroke timing, we're able to tell who's in 00:11:40.179 --> 00:11:42.539 control. the device. That to me is pretty crazy. 00:11:42.580 --> 00:11:45.960 It's kind of like when there's kind of the videos 00:11:45.960 --> 00:11:48.120 that you may or may not have seen, but basically 00:11:48.120 --> 00:11:51.700 your Wi -Fi. um sends out certain signals and 00:11:51.700 --> 00:11:54.259 you can you can basically get like a 3d map of 00:11:54.259 --> 00:11:56.879 who's inside of a house if you get access to 00:11:56.879 --> 00:11:59.500 their wi -fi router because like a human body 00:11:59.500 --> 00:12:02.019 for example the wi -fi signals go through them 00:12:02.019 --> 00:12:04.080 at a different frequency of walls and all these 00:12:04.080 --> 00:12:05.820 different steps so like you can you basically 00:12:05.820 --> 00:12:08.120 get like a 3d map of someone walking around their 00:12:08.120 --> 00:12:10.720 house if you can um if you can hack into their 00:12:10.720 --> 00:12:14.299 wi -fi router which reddit ceo is publicly questioning 00:12:14.299 --> 00:12:17.080 whether google's ai Overview feature is worth 00:12:17.080 --> 00:12:19.659 the money. He says that the automated search 00:12:19.659 --> 00:12:21.940 summaries haven't delivered any real value to 00:12:21.940 --> 00:12:24.159 content publishers. Reddit shares fell about 00:12:24.159 --> 00:12:27.159 20 % after they released their earnings, which 00:12:27.159 --> 00:12:29.039 I think part of that is because of just a lot 00:12:29.039 --> 00:12:30.919 of the uncertainty about whether Google's shift 00:12:30.919 --> 00:12:33.200 towards AI summaries is going to kill all of 00:12:33.200 --> 00:12:35.059 the traffic -based revenue for sites like Reddit. 00:12:35.159 --> 00:12:38.700 And I mean, to be fair, with the AI boom, Reddit 00:12:38.700 --> 00:12:40.720 is one of the biggest winners. You probably have 00:12:40.720 --> 00:12:43.639 noticed that Google's AI are just their results 00:12:43.639 --> 00:12:46.240 in general really heavily skew towards Reddit 00:12:46.240 --> 00:12:48.179 because it's like real people, there's real trust. 00:12:48.240 --> 00:12:51.019 And a lot of companies have licensed information 00:12:51.019 --> 00:12:53.220 from Reddit, Google being one, I think Google 00:12:53.220 --> 00:12:57.200 paid $300 million to license everything on Reddit 00:12:57.200 --> 00:13:00.100 to help them train their models. Pew Research 00:13:00.100 --> 00:13:03.620 study of 900 U .S. adults said that basically 00:13:03.620 --> 00:13:05.759 what they figured out from this was that AI overviews 00:13:05.759 --> 00:13:08.340 cut referrals to publisher sites by almost half 00:13:08.340 --> 00:13:10.620 compared to traditional search results. OK, this 00:13:10.620 --> 00:13:12.820 isn't like what what is shocking about that? 00:13:12.919 --> 00:13:15.600 Basically, you ask the question, the AI overview 00:13:15.600 --> 00:13:17.700 answers. So you don't go click over to the website. 00:13:17.860 --> 00:13:20.059 Nothing's shocking there. Reddit is considering 00:13:20.059 --> 00:13:22.700 ending their 60 million dollar licensing Google 00:13:22.700 --> 00:13:24.779 deal with Google. So I think I might have said 00:13:24.779 --> 00:13:27.600 300 million earlier. Misquoted that it's 60 million 00:13:27.600 --> 00:13:30.379 dollars. And the economists, routers, political 00:13:30.379 --> 00:13:32.899 and USA Today are all also thinking about doing 00:13:32.899 --> 00:13:34.779 something very similar. So Google has licensing 00:13:34.779 --> 00:13:37.019 deals with all of these companies. And a lot 00:13:37.019 --> 00:13:38.700 of them are saying, look, maybe it's just not 00:13:38.700 --> 00:13:41.100 worth it to have all of the Google or all of 00:13:41.100 --> 00:13:43.019 the Reddit information embedded into Gemini. 00:13:43.019 --> 00:13:45.039 We'd rather them actually come over to Reddit 00:13:45.039 --> 00:13:47.820 to get that information. Now, if I'm being honest, 00:13:48.139 --> 00:13:52.100 I think that. I don't know if they cut their 00:13:52.100 --> 00:13:53.940 licensing deals, if the traffic's going to actually 00:13:53.940 --> 00:13:56.100 flow back to them. I think maybe the cat's out 00:13:56.100 --> 00:13:57.639 of the bag at this point. A lot of these AI models 00:13:57.639 --> 00:13:59.799 are trained. A lot of this information is already 00:13:59.799 --> 00:14:03.179 out there. It's already been aggregated on, you 00:14:03.179 --> 00:14:05.519 know, maybe Reddit got added into Gemini and 00:14:05.519 --> 00:14:06.980 someone used Gemini to write an article. Now 00:14:06.980 --> 00:14:09.299 the article is publicly posted. So Google doesn't 00:14:09.299 --> 00:14:11.899 really need Reddit so much anymore. That's kind 00:14:11.899 --> 00:14:14.379 of what I think will happen and probably the 00:14:14.379 --> 00:14:16.100 same thing with a lot of the other news organizations. 00:14:16.419 --> 00:14:18.960 So it'll be interesting to see how that rolls 00:14:18.960 --> 00:14:21.570 out. the future. There's some big news from Apple. 00:14:21.669 --> 00:14:26.789 Tim Cook says that they might have a new tier, 00:14:26.990 --> 00:14:30.289 which is going to be called iCloud Plus for heavy 00:14:30.289 --> 00:14:32.789 Siri users. It's going to be a flat monthly fee. 00:14:33.029 --> 00:14:35.470 It's going to be subscription, very similar to 00:14:35.470 --> 00:14:38.830 the freemium model of OpenAI and Google for what 00:14:38.830 --> 00:14:42.570 they have for Gemini or for ChatGPT. And right 00:14:42.570 --> 00:14:45.190 now, I think Apple is basically saying that they 00:14:45.190 --> 00:14:47.389 think they can convert iPhone users who currently 00:14:47.389 --> 00:14:50.769 use ChatGPT and other things into... using Siri 00:14:50.769 --> 00:14:53.269 instead. Now what's interesting is we know that 00:14:53.269 --> 00:14:55.649 there's like some version of Siri that is going 00:14:55.649 --> 00:14:59.029 to be powered by OpenAI and powered by ChatGPT 00:14:59.029 --> 00:15:01.629 or I think a lot of the news is that you could 00:15:01.629 --> 00:15:03.169 actually switch it. So this is going to be the 00:15:03.169 --> 00:15:05.309 first time that they actually are charging a 00:15:05.309 --> 00:15:07.809 subscription and getting into that game. Not 00:15:07.809 --> 00:15:10.710 long after OpenAI has come out with an announcement 00:15:10.710 --> 00:15:13.029 that, you know, their model escaped a sandbox 00:15:13.029 --> 00:15:15.230 and went and hacked Hugging Faces and some other 00:15:15.230 --> 00:15:18.320 firms. Anthropic is saying, look, we did the 00:15:18.320 --> 00:15:21.059 same thing, guys. We have, you know, Claude, 00:15:21.059 --> 00:15:23.419 three different Claude models escaped a testing 00:15:23.419 --> 00:15:26.720 sandbox and access to live company systems during 00:15:26.720 --> 00:15:30.179 internal security evaluations. One model. published 00:15:30.179 --> 00:15:32.940 malware to a public code registry before it was 00:15:32.940 --> 00:15:35.840 caught, which prompted Anthropic to review over 00:15:35.840 --> 00:15:39.980 141 ,000 test runs for similar breaches. If you 00:15:39.980 --> 00:15:42.500 haven't left a review on the podcast yet, it 00:15:42.500 --> 00:15:44.940 would help the show out a ton. I would super, 00:15:45.000 --> 00:15:46.980 super appreciate it. It helps more incredible 00:15:46.980 --> 00:15:49.799 people like yourself find this show. And so yeah, 00:15:49.899 --> 00:15:51.299 it's the number one thing that could help this 00:15:51.299 --> 00:15:53.799 show reach more a bigger audience. If you wouldn't 00:15:53.799 --> 00:15:55.740 mind if you haven't already done it, it would 00:15:55.740 --> 00:15:57.620 mean the world to me. But thanks so much for 00:15:57.620 --> 00:15:58.960 tuning in. I hope you guys have an incredible 00:15:58.960 --> 00:16:02.090 day. Make sure to go check out AIbox .ai if you 00:16:02.090 --> 00:16:04.090 want to get 80 different AI models all on one 00:16:04.090 --> 00:16:06.269 platform. And I'll see you in the next episode.