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 --> 00:00:17.750 business. Are you subscribed to AI chat daily 00:00:17.750 --> 00:00:20.390 .com? Yeah, I have a newsletter that I send out 00:00:20.390 --> 00:00:22.730 every single day. And it's essentially all of 00:00:22.730 --> 00:00:24.469 the same content, all the same stories I talked 00:00:24.469 --> 00:00:26.530 about in the podcast, but I do them in these 00:00:26.530 --> 00:00:28.730 really easy to break down formats. And I also 00:00:28.730 --> 00:00:30.829 have links to articles to go more in depth. So 00:00:30.829 --> 00:00:32.289 it's kind of like show notes for the podcast, 00:00:32.469 --> 00:00:34.289 but it goes into way more depth. And we have 00:00:34.289 --> 00:00:36.369 extra stories that we can't get around to covering 00:00:36.369 --> 00:00:38.530 in the podcast. So if every single day you want 00:00:38.530 --> 00:00:40.549 to start the day out or end the day by getting 00:00:40.549 --> 00:00:43.250 a roundup of the latest in AI news, and a really 00:00:43.250 --> 00:00:46.090 easy to scan format, go go to AIChatDaily .com. 00:00:46.170 --> 00:00:47.969 There's a big subscribe button in the top corner. 00:00:48.030 --> 00:00:49.850 I'll probably leave a link in the show notes 00:00:49.850 --> 00:00:52.049 to the newsletter. And it also is just a full 00:00:52.049 --> 00:00:53.869 news site with tons of articles. I'll probably 00:00:53.869 --> 00:00:55.689 publish about 10 of them every single day. So 00:00:55.689 --> 00:00:57.909 if you want to go check that out, AIChatDaily 00:00:57.909 --> 00:01:02.149 .com. Okay, OpenAI is shipping GPT Live 1. So 00:01:02.149 --> 00:01:05.250 this is a full duplex voice model. And what duplex 00:01:05.250 --> 00:01:06.909 voice model means, well, this is actually kind 00:01:06.909 --> 00:01:09.409 of crazy because Miriam Murady, who is formerly 00:01:09.409 --> 00:01:12.469 at OpenAI, she released something recently with 00:01:12.469 --> 00:01:14.590 her most recent company. Thinking Machine Labs, 00:01:14.890 --> 00:01:16.790 where it's, I mean, essentially it looks like 00:01:16.790 --> 00:01:19.010 OpenAI has just knocked her off. So this is going 00:01:19.010 --> 00:01:22.030 to be interesting. Also, China has flagged Anthropix 00:01:22.030 --> 00:01:24.329 Cloud Code. They say that this is a security 00:01:24.329 --> 00:01:26.670 risk because there's some backdoor risks with 00:01:26.670 --> 00:01:28.269 it, which I don't know. I just think anything 00:01:28.269 --> 00:01:30.250 with security risk is ironic coming from China. 00:01:30.989 --> 00:01:34.489 NVIDIA's Nemotron 3 Ultra has just hit closed 00:01:34.489 --> 00:01:38.109 model parity at a 10x lower cost. And this is 00:01:38.109 --> 00:01:39.930 from some experiments they've been doing on Langchain. 00:01:40.109 --> 00:01:41.890 I'm really excited when we're able to get the 00:01:41.890 --> 00:01:44.480 costs down on some of these. models. Meta is 00:01:44.480 --> 00:01:48.000 adding a tamper -proof LED to their AI Meta glasses, 00:01:48.299 --> 00:01:50.939 and they're also getting more data collection 00:01:50.939 --> 00:01:52.840 from other parts of the glasses. So there's kind 00:01:52.840 --> 00:01:55.120 of a mixed bag with what's going on there. An 00:01:55.120 --> 00:01:59.519 AI, the OpenAI chief futurist, John HM, is leaving. 00:01:59.659 --> 00:02:01.799 He's been there for nine years. He's one of the 00:02:01.799 --> 00:02:04.579 OG trust and safety people over OpenAI. But let's 00:02:04.579 --> 00:02:06.719 kick the podcast off with OpenAI's latest model, 00:02:06.859 --> 00:02:10.159 GPT Live 1. This is a voice model. It listens 00:02:10.159 --> 00:02:12.939 and it speaks at the same time. is something 00:02:12.939 --> 00:02:14.840 that you've probably noticed a lot of these voice 00:02:14.840 --> 00:02:16.219 models have struggled with in the past. If you're 00:02:16.219 --> 00:02:19.599 talking to, you know, what voice mode is on chat 00:02:19.599 --> 00:02:21.400 GPT right now, which by the way, this replaces 00:02:21.400 --> 00:02:23.680 their advanced voice mode. But if you, if you've 00:02:23.680 --> 00:02:25.400 ever talked to it before you talk, there's a 00:02:25.400 --> 00:02:27.300 couple seconds of latency, then it responds back 00:02:27.300 --> 00:02:28.560 to you. And they've done a lot of work to try 00:02:28.560 --> 00:02:31.020 to like get that latency down as much as possible. 00:02:31.159 --> 00:02:33.580 But what's really cool about these new live one 00:02:33.580 --> 00:02:35.860 model is that it's literally listening and it's 00:02:35.860 --> 00:02:37.699 able to listen and talk at the same time, because 00:02:37.699 --> 00:02:39.500 that's how actual humans talk. When we're in 00:02:39.500 --> 00:02:41.159 a conversation, we kind of interrupt each other. 00:02:41.199 --> 00:02:43.240 We talk. over each other. They're saying something, 00:02:43.280 --> 00:02:44.879 you're saying something. And I mean, the goal 00:02:44.879 --> 00:02:46.460 is not to interrupt and talk over each other. 00:02:46.500 --> 00:02:48.159 But it's just part of communication, right? Like, 00:02:48.199 --> 00:02:49.780 we don't always wait for there to be a perfect 00:02:49.780 --> 00:02:51.879 pause at the end of everything that someone says 00:02:51.879 --> 00:02:53.759 before we jump in. So that's what these models 00:02:53.759 --> 00:02:55.919 are getting really good at, which is really interesting. 00:02:56.520 --> 00:02:59.090 And what's to me really crazy about this is that 00:02:59.090 --> 00:03:01.909 Miriam Marotti released a model just like this 00:03:01.909 --> 00:03:03.370 a number of months ago. I remember covering it 00:03:03.370 --> 00:03:05.009 on the podcast and it was like the big thing 00:03:05.009 --> 00:03:06.629 from Thinking Machine Labs. And I was like, man, 00:03:06.729 --> 00:03:08.770 this is a really cool idea and concept. Well, 00:03:08.849 --> 00:03:11.110 turns out, you know, a couple months later, anyone 00:03:11.110 --> 00:03:13.449 that releases any sort of cool new feature, a 00:03:13.449 --> 00:03:15.389 couple months later, the Frontier Labs are going 00:03:15.389 --> 00:03:17.270 to be able to clone it, reverse engineer and 00:03:17.270 --> 00:03:18.889 put it out themselves because, you know, they 00:03:18.889 --> 00:03:20.810 don't want to fall behind. The reason why I think 00:03:20.810 --> 00:03:22.849 this is one in particular is an important feature 00:03:22.849 --> 00:03:25.789 is because there's three separate things that 00:03:25.789 --> 00:03:28.569 are getting collapsed into one step. Number one, 00:03:28.569 --> 00:03:30.169 it's cutting the delay. So it's going to let 00:03:30.169 --> 00:03:32.849 your conversation be way more natural. And they're 00:03:32.849 --> 00:03:35.810 also going to be rolling this out to 150 million 00:03:35.810 --> 00:03:38.729 chat GPT voice users. And finally, the way this 00:03:38.729 --> 00:03:40.509 works is really interesting to me. So there's 00:03:40.509 --> 00:03:42.810 something called GPT live one mini, and this 00:03:42.810 --> 00:03:44.289 is going to be for basically all of the free 00:03:44.289 --> 00:03:46.310 users, it's kind of the default. But if you're 00:03:46.310 --> 00:03:48.030 on the paid tier, if you're paying $20 a month 00:03:48.030 --> 00:03:51.069 for chat GPT, you get the larger GPT live one, 00:03:51.189 --> 00:03:53.110 which is going to send some of the harder questions 00:03:53.110 --> 00:03:56.259 over to GPT. 5 .5 mid conversation. So you're 00:03:56.259 --> 00:03:57.500 going to be talking in the middle of the conversation. 00:03:57.620 --> 00:03:59.439 If you're saying something kind of complex and 00:03:59.439 --> 00:04:01.259 it's getting like, I mean, basically get half 00:04:01.259 --> 00:04:03.780 your sentence, send it over to GPT 5 .5 to try 00:04:03.780 --> 00:04:05.620 to come up with a response and send it back and 00:04:05.620 --> 00:04:07.580 put it out before you even finish your sentence, 00:04:07.699 --> 00:04:10.199 which is really crazy. It can be interrupted 00:04:10.199 --> 00:04:12.120 in the middle of a sentence. So if it's talking, 00:04:12.180 --> 00:04:14.759 you can interrupt it. It handles silences while 00:04:14.759 --> 00:04:16.920 you think. So just because, you know, you're, 00:04:16.920 --> 00:04:18.420 you're pausing to think about something doesn't 00:04:18.420 --> 00:04:19.939 mean it feels like it has to jump in, which is 00:04:19.939 --> 00:04:21.459 something really annoying. I feel like I can 00:04:21.459 --> 00:04:23.639 barely like keep my thoughts. Like I can barely 00:04:23.639 --> 00:04:25.040 think when I'm talking to these things because 00:04:25.040 --> 00:04:26.720 I got to talk so fast because if I stop talking, 00:04:26.839 --> 00:04:28.319 it's going to want to jump in and start talking 00:04:28.319 --> 00:04:31.120 to me. So I really appreciate that. And it's 00:04:31.120 --> 00:04:33.160 going to show charts or images alongside whatever 00:04:33.160 --> 00:04:35.199 you're talking to it about, which, you know, 00:04:35.199 --> 00:04:37.139 before it was just kind of this like animated 00:04:37.139 --> 00:04:40.000 strobe on your screen. But now if you're talking, 00:04:40.060 --> 00:04:42.100 you're like, hey, you know, what are the, you 00:04:42.100 --> 00:04:43.500 know, what's the weather going to be like? It's 00:04:43.500 --> 00:04:45.279 actually pulling up weather charts on the screen 00:04:45.279 --> 00:04:47.399 while it talks to you. So you get visual and 00:04:47.399 --> 00:04:49.779 audio, which is really cool. The product leave 00:04:49.779 --> 00:04:52.670 on this is or the product lead is Addy. Letty. 00:04:52.990 --> 00:04:56.050 He said that he held a 30 to 40 minute conversation 00:04:56.050 --> 00:04:59.550 with the model on walks. And I think the goal 00:04:59.550 --> 00:05:01.110 and probably the reason why they're telling us 00:05:01.110 --> 00:05:02.730 this is because they're trying to make this something 00:05:02.730 --> 00:05:05.129 for or trying to explain this is really good 00:05:05.129 --> 00:05:07.069 for long running tasks. This isn't just quick 00:05:07.069 --> 00:05:08.449 commands like, hey, what's the weather like? 00:05:08.529 --> 00:05:10.230 It's like, hey, I'm trying to break down this 00:05:10.230 --> 00:05:12.490 complex business plan that I want to, you know, 00:05:12.490 --> 00:05:14.829 create. talk with me about this, or I'm trying 00:05:14.829 --> 00:05:16.930 to figure out what the best strategy is for XYZ, 00:05:16.990 --> 00:05:19.029 talk to me about this. And you go and have a 00:05:19.029 --> 00:05:21.110 40 minute conversation to learn something to 00:05:21.110 --> 00:05:22.850 plan something to actually get something done, 00:05:23.069 --> 00:05:25.149 which if I'm being honest, I was kind of moving 00:05:25.149 --> 00:05:26.949 away from opening eye and chat GPT for a while 00:05:26.949 --> 00:05:28.649 because it didn't feel like that was their focus. 00:05:28.810 --> 00:05:30.990 And I was going to Claude, but I've been trying 00:05:30.990 --> 00:05:34.269 out a bunch of things on chat GPT lately, mostly 00:05:34.269 --> 00:05:36.769 like whenever my whenever my tokens, my token 00:05:36.769 --> 00:05:39.089 limits reach on Claude, I go back to chat GPT. 00:05:39.170 --> 00:05:41.189 And I've been really impressed by the research 00:05:41.189 --> 00:05:43.319 it can do, where I'm like, hey, you know, go 00:05:43.319 --> 00:05:46.540 scrape like these 5000 things and do an analysis 00:05:46.540 --> 00:05:48.439 on these different websites or these different 00:05:48.439 --> 00:05:51.019 products and like, bring me back, you know, help 00:05:51.019 --> 00:05:53.920 me create a list of 365 XYZ things I can do for 00:05:53.920 --> 00:05:56.040 this, like it goes and get some real serious 00:05:56.040 --> 00:05:59.339 work done. So I'm really excited. I think the 00:05:59.339 --> 00:06:01.259 reason why this is particularly important beyond 00:06:01.259 --> 00:06:03.939 just humans is because we're going to have a 00:06:03.939 --> 00:06:07.839 lot of work with earbuds and wearables that OpenAI 00:06:07.839 --> 00:06:10.860 is currently working on. And so a lot of the 00:06:10.860 --> 00:06:13.420 hardware that they're making, these tools are 00:06:13.420 --> 00:06:15.740 going to be really useful for. And, you know, 00:06:15.740 --> 00:06:17.240 when we start talking about robots, getting them 00:06:17.240 --> 00:06:19.420 and all that kind of stuff and be able to communicate 00:06:19.420 --> 00:06:21.139 with that, this is going to be important that 00:06:21.139 --> 00:06:24.040 the communication is as smooth as possible. China 00:06:24.040 --> 00:06:26.720 is flagging Anthropics Cloud Code as a security 00:06:26.720 --> 00:06:29.500 backdoor risk, which is hilarious because China 00:06:29.500 --> 00:06:31.720 has been accused of putting backdoors into chips 00:06:31.720 --> 00:06:35.740 and hardware inside of China for many years on, 00:06:35.779 --> 00:06:37.639 you know, servers and all sorts of crazy stuff. 00:06:37.959 --> 00:06:39.560 But this in particular, I do think is probably 00:06:39.560 --> 00:06:41.939 accurate. What they've said is that Claude Code 00:06:41.939 --> 00:06:44.139 has a backdoor vulnerability that's going to 00:06:44.139 --> 00:06:47.720 leak your location and your identity data. And 00:06:47.720 --> 00:06:49.699 they're telling everyone in China to uninstall 00:06:49.699 --> 00:06:52.240 or to upgrade the version if you have something 00:06:52.240 --> 00:06:54.540 installed between April and June to upgrade it. 00:06:54.579 --> 00:06:57.360 The advisory is interesting because Claude Code 00:06:57.360 --> 00:07:00.720 isn't officially sold in China. So, you know, 00:07:00.759 --> 00:07:03.500 Alibaba has banned Anthropic Tools company -wide. 00:07:03.660 --> 00:07:06.500 This is also a week after Anthropic accused Alibaba 00:07:06.500 --> 00:07:09.819 of stealing. their AI technology. And what was 00:07:09.819 --> 00:07:12.040 interesting to me is there's like some, some 00:07:12.040 --> 00:07:14.240 moments where different people inside of I think 00:07:14.240 --> 00:07:17.100 Exomi, they had like an AI, AI developer conference, 00:07:17.259 --> 00:07:18.579 I think, and they were kind of publicly acknowledging 00:07:18.579 --> 00:07:21.519 that they were using cloud code at a state organized 00:07:21.519 --> 00:07:24.439 forum. This is back in March, even though it's, 00:07:24.439 --> 00:07:26.560 you know, officially not available in China. 00:07:26.600 --> 00:07:28.079 So it's funny, because it's like officially not 00:07:28.079 --> 00:07:30.620 available, officially not sold, Alibaba is banning 00:07:30.620 --> 00:07:32.500 it, no one's supposed to be using it. And yet 00:07:32.500 --> 00:07:34.959 China's like warning everybody, like they know 00:07:34.959 --> 00:07:36.480 they're using it. Anyways, it's kind of crazy. 00:07:36.490 --> 00:07:39.290 What I do think is important, though, is this 00:07:39.290 --> 00:07:43.470 particular flag they have is version 2 .191 to 00:07:43.470 --> 00:07:50.050 2 .19656. But we're already on 2 .12 for the 00:07:50.050 --> 00:07:52.089 version that we're currently on. So if you're 00:07:52.089 --> 00:07:54.329 on the latest Anthropic thing, this isn't really 00:07:54.329 --> 00:07:56.990 relevant to you. Most modern coding assistants, 00:07:57.189 --> 00:07:59.290 so if you look at like OpenAI and Google, they're 00:07:59.290 --> 00:08:02.259 sending what's called... telemetry and crash 00:08:02.259 --> 00:08:06.019 reports back to the main servers. And anytime 00:08:06.019 --> 00:08:07.339 that something happens, I think this is what 00:08:07.339 --> 00:08:09.660 China is worried about the crash reports, they're 00:08:09.660 --> 00:08:11.579 worried has too much information about the users 00:08:11.579 --> 00:08:14.500 in it. But China wasn't really clarifying if 00:08:14.500 --> 00:08:17.920 quad code was the only one doing this, why they 00:08:17.920 --> 00:08:20.420 were singled out and not like open AI and Google. 00:08:20.879 --> 00:08:23.300 I'm not even sure if Google's is available in 00:08:23.300 --> 00:08:26.310 China. So anyways, it's China's a crazy, crazy 00:08:26.310 --> 00:08:29.889 place. In a big win for open source models, NVIDIA's 00:08:29.889 --> 00:08:33.629 open source Nemotron 3 Ultra model now matches 00:08:33.629 --> 00:08:36.750 the best closed AI systems on business tasks. 00:08:36.850 --> 00:08:38.950 So not everything, but business tasks specifically, 00:08:39.110 --> 00:08:42.269 and it can do them at a tenth of the cost. Langchain 00:08:42.269 --> 00:08:45.129 actually tuned the agent software around the 00:08:45.129 --> 00:08:47.730 model without retraining it. And they basically 00:08:47.730 --> 00:08:50.049 proved that the right engineering environment 00:08:50.049 --> 00:08:52.590 can compete with really expensive proprietary 00:08:52.590 --> 00:08:56.750 AI when it comes to doing enterprise work. So 00:08:56.750 --> 00:08:58.269 you can get these open source models, if you 00:08:58.269 --> 00:09:00.389 fine tune them, you can get them just as good 00:09:00.389 --> 00:09:02.590 for way cheaper than something like Anthropic 00:09:02.590 --> 00:09:06.059 or OpenAI. lane chain, adjusted system prompts, 00:09:06.279 --> 00:09:08.960 tool descriptions, and middleware. They did all 00:09:08.960 --> 00:09:11.919 of that to really just squeeze the top tier performance. 00:09:12.120 --> 00:09:14.720 Then they put it through the deep agent benchmark. 00:09:15.379 --> 00:09:18.940 Ambridge, Amdocs, Box, and EY are already building 00:09:18.940 --> 00:09:22.019 specialized agents on this exact stack for their 00:09:22.019 --> 00:09:24.320 platforms and their workflows. The tuned model 00:09:24.320 --> 00:09:26.620 is live on six hosting providers, so you can 00:09:26.620 --> 00:09:29.700 get it on Bastion, Crusoe Cloud, Deep Infra, 00:09:29.720 --> 00:09:32.580 Firework, Nebius, and together AI, which is what 00:09:32.580 --> 00:09:35.179 I use for AI. box. So it's, you know, pretty 00:09:35.179 --> 00:09:37.059 widely available. This is really exciting. And 00:09:37.059 --> 00:09:38.779 I believe we actually do have pneumotron over 00:09:38.779 --> 00:09:40.559 on AI box right now, which by the way, if you 00:09:40.559 --> 00:09:43.100 haven't tried AI box .ai, my own startup, we 00:09:43.100 --> 00:09:45.960 have over 80 different AI models all on there. 00:09:46.039 --> 00:09:49.039 So for $8 .99 a month, you get access to all 00:09:49.039 --> 00:09:52.659 of the top AI image, audio, video, and even music 00:09:52.659 --> 00:09:54.240 models. So if you don't want to have to have 00:09:54.240 --> 00:09:55.960 subscriptions to all of these different things 00:09:55.960 --> 00:09:57.759 and have, you know, logins on a ton of different 00:09:57.759 --> 00:10:00.139 platforms, just go to one place, AI box .ai. 00:10:00.279 --> 00:10:02.279 We even have an MCP that will tie all of those 00:10:02.279 --> 00:10:04.269 models. into things like clod so if you want 00:10:04.269 --> 00:10:06.049 to get like image and audio and video inside 00:10:06.049 --> 00:10:09.230 of clod you can use our mcp $8 .99 a month i'll 00:10:09.230 --> 00:10:10.830 leave a link in the description to that as well 00:10:11.120 --> 00:10:14.120 Meta just added a tamper -proof LED to their 00:10:14.120 --> 00:10:16.899 AI glasses, and it disables the camera if the 00:10:16.899 --> 00:10:20.460 light is modified or destroyed. So the idea being, 00:10:20.639 --> 00:10:22.480 you know, you want to know when someone's recording 00:10:22.480 --> 00:10:24.179 you with the Meta Ray -Bans or any of these other 00:10:24.179 --> 00:10:26.299 Meta AI smart glasses. A lot of people have been 00:10:26.299 --> 00:10:28.659 complaining about, you know, non -consensual 00:10:28.659 --> 00:10:30.279 recording. People are recording them when they 00:10:30.279 --> 00:10:31.840 don't know or they don't consent. People are 00:10:31.840 --> 00:10:34.100 in the gym recording, whatever. And so they've 00:10:34.100 --> 00:10:36.179 added a light, which I think a lot of devices 00:10:36.179 --> 00:10:38.159 have, like a light that pops on when it's recording 00:10:38.159 --> 00:10:40.659 so you know someone's recording. Theoretically, 00:10:40.720 --> 00:10:42.500 someone could go with like a pin and just like, 00:10:42.519 --> 00:10:45.340 you know, kind of shatter the LED bulb or wreck 00:10:45.340 --> 00:10:47.600 it, right? So it doesn't turn on or cover it 00:10:47.600 --> 00:10:49.440 or do something. So they're making it so that 00:10:49.440 --> 00:10:51.200 you can't tamper with it. If it's broken, it 00:10:51.200 --> 00:10:54.360 won't actually record. And this is, I mean, I 00:10:54.360 --> 00:10:55.659 think a lot of people are really excited about 00:10:55.659 --> 00:10:57.580 that. But this is also happening the same week 00:10:57.580 --> 00:11:00.379 that Meta is expanding the data collection elsewhere. 00:11:00.519 --> 00:11:02.500 So they're going to be allowing their AI to train 00:11:02.500 --> 00:11:05.539 on public Instagram photos by default. And they're 00:11:05.539 --> 00:11:08.000 testing glasses that continuously collect audio 00:11:08.000 --> 00:11:10.929 and they snap photos. every few seconds, which 00:11:10.929 --> 00:11:14.409 come on, that's terrifying. But I mean, maybe 00:11:14.409 --> 00:11:16.190 like they're milled in a product that they think 00:11:16.190 --> 00:11:18.230 some people will like that. I really doubt that. 00:11:18.289 --> 00:11:20.870 But this particular prototype that is under testing 00:11:20.870 --> 00:11:23.409 is going to move the glasses from being an on 00:11:23.409 --> 00:11:25.029 -demand camera, right? So you're like, hey, Meta, 00:11:25.029 --> 00:11:26.909 record whatever I'm doing. And it turns on to 00:11:26.909 --> 00:11:28.889 actually just being always on and having an always 00:11:28.889 --> 00:11:30.970 -on sensor. And it's feeding continuous audio 00:11:30.970 --> 00:11:34.110 and periodic photos to Meta's AI systems. They 00:11:34.110 --> 00:11:36.909 don't have a date launch for this. I hope there'll 00:11:36.909 --> 00:11:38.509 be a way to opt out. Otherwise, I don't imagine 00:11:38.509 --> 00:11:40.970 these glasses will continue to be very popular. 00:11:41.110 --> 00:11:43.330 For the Instagram thing, this is a change that 00:11:43.330 --> 00:11:45.539 will affect... about a billion that, you know, 00:11:45.539 --> 00:11:48.600 billions of photos for users. Most like everyone 00:11:48.600 --> 00:11:51.120 posting is not wanting their photos to be trained, 00:11:51.159 --> 00:11:53.340 like AI to be trained on them. So that's going 00:11:53.340 --> 00:11:56.700 to be kind of crazy. OpenAI's chief futurist, 00:11:56.759 --> 00:12:00.559 Joshua Achim, is leaving after about nine years. 00:12:00.720 --> 00:12:03.500 He's one of the safety advocates that's been 00:12:03.500 --> 00:12:05.799 at the company the longest since the early days. 00:12:05.940 --> 00:12:08.899 He is leaving, and he's the fifth senior safety 00:12:08.899 --> 00:12:11.019 -focused leader to leave OpenAI in the last two 00:12:11.019 --> 00:12:12.860 years, right? They're getting ready to go public. 00:12:13.039 --> 00:12:15.179 He actually joined as an intern back in 2017, 00:12:15.299 --> 00:12:17.679 and then he led their mission alignment team 00:12:17.679 --> 00:12:21.480 until it was disabled, or I guess like disbanded. 00:12:21.480 --> 00:12:24.360 Sam Altman got rid of it back in February of... 00:12:24.399 --> 00:12:27.000 this year. Then he moved into chief futurist 00:12:27.000 --> 00:12:29.539 role and was kind of studying AI harms and benefits. 00:12:30.220 --> 00:12:32.059 Five different safety -adjacent leaders have 00:12:32.059 --> 00:12:36.080 now left OpenAI since 2024, and a lot of them 00:12:36.080 --> 00:12:38.299 have started non -profits focused on AI safety 00:12:38.299 --> 00:12:40.960 standards. The former White House AI advisor 00:12:40.960 --> 00:12:44.440 Dean Ball started this week as the head of strategic 00:12:44.440 --> 00:12:46.779 futures and is going to briefly overlap with 00:12:46.779 --> 00:12:49.100 Etcham before he actually leaves and will kind 00:12:49.100 --> 00:12:50.899 of take over a lot of policy -focused research 00:12:50.899 --> 00:12:53.460 work without having a direct replacement for 00:12:53.460 --> 00:12:56.019 the chief futurist title. So if you believe you'd 00:12:56.019 --> 00:12:57.840 be perfect for that job, maybe you should send 00:12:57.840 --> 00:13:00.000 in your application to OpenAI. Guys, thank you 00:13:00.000 --> 00:13:01.639 so much for tuning into the podcast today. Make 00:13:01.639 --> 00:13:04.019 sure to check out AIbox .ai. Links in the description. 00:13:04.139 --> 00:13:06.440 If you have not already left a review for the 00:13:06.440 --> 00:13:08.539 show, it would mean the world to me. It helps 00:13:08.539 --> 00:13:10.659 the show out tremendously. Add some exciting 00:13:10.659 --> 00:13:12.779 interviews and content and stuff planned for 00:13:12.779 --> 00:13:14.639 the show this year. And if you would be able 00:13:14.639 --> 00:13:16.539 to leave a review, it would help me to bring 00:13:16.539 --> 00:13:19.360 you more incredible content. Over on Spotify, 00:13:19.600 --> 00:13:21.120 you have to hit the About tab. If you've listened 00:13:21.120 --> 00:13:22.659 to three episodes, you're eligible to leave a 00:13:22.659 --> 00:13:24.860 review. Otherwise, I won't let you. So if you 00:13:24.860 --> 00:13:27.200 have, that would be incredible. Hit the About 00:13:27.200 --> 00:13:30.320 tab. Hit some stars. stars on Apple. It's kind 00:13:30.320 --> 00:13:32.440 of annoying to find, but you go to the podcast 00:13:32.440 --> 00:13:34.960 page and you can hit the stars and leave a comment. 00:13:35.039 --> 00:13:36.960 I read them all. I appreciate them. Alright guys, 00:13:37.039 --> 00:13:39.159 I hope you all have a fantastic rest of your 00:13:39.159 --> 00:13:39.320 day.