WEBVTT 00:00:00.000 --> 00:00:05.440 I can't believe Android would do something like this! Back in my day, I was a teeny little baby boy. 00:00:05.440 --> 00:00:09.360 That's not related, but it is true. Some Android users are being blocked from 00:00:09.360 --> 00:00:13.680 side-loading apps and instead forced to download them from the Play Store. It's a weird move, 00:00:13.680 --> 00:00:17.520 since side-loading has been one of Android's main selling points, whereas the iPhone's selling 00:00:17.520 --> 00:00:22.320 point was not getting bullied over the color green. Android users who have side-loaded certain apps 00:00:22.320 --> 00:00:26.960 have received unsurpassable pop-ups, prompting them to install the app from the Play Store instead. 00:00:27.040 --> 00:00:31.120 There are some issues with enforcing this. For starters, some apps can't be downloaded from 00:00:31.120 --> 00:00:35.520 the Play Store in certain regions. On top of that, this affects users of alternate operating systems 00:00:35.520 --> 00:00:41.440 like the privacy and security-focused graphing OS. And on top of that, I'm uncomfortable with the 00:00:41.440 --> 00:00:47.040 idea of being told what to do by a f***ing rectangle! Not all restrictions are bad. For example, 00:00:47.040 --> 00:00:51.760 Android 15 will restrict sensitive permissions like being able to listen to notifications from 00:00:51.760 --> 00:00:56.000 side-loaded apps by default, which can prevent malicious apps from accessing your data. It's 00:00:56.000 --> 00:01:00.560 just weird to me that Android apps are being restricted at the same time that Apple is preparing 00:01:00.560 --> 00:01:06.800 to allow third-party app stores on EU iPads. I mean, what is it? Opposite day? OpenAI has released 00:01:06.800 --> 00:01:13.280 01? Allegedly, the company's first family to have reasoning skills. In their official announcement, 00:01:13.280 --> 00:01:18.080 OpenAI said 01 was trained to spend more time thinking about problems before responding, 00:01:18.080 --> 00:01:23.040 like a person would. I feel like that's kind of an inaccurate description of people based on, you 00:01:23.600 --> 00:01:28.640 know, all of history. Whatever the reason, thanks to OpenAI's overestimation of what a person is 00:01:28.640 --> 00:01:35.680 capable of, 01 is apparently far more accurate than their previous model, GPT-4O, at answering math, 00:01:35.680 --> 00:01:40.720 coding, and PhD-level science questions. Even more impressively, it nearly passed the 00:01:40.720 --> 00:01:45.520 I'm Not Your Buddy guy test, developed by totally real Canadian computer scientist, 00:01:45.520 --> 00:01:49.120 Terrence and Philip. Now, before you get all excited about using 01, you should know that 00:01:49.200 --> 00:01:54.960 there are some massive restrictions that nobody will be able to bypass. Not only is the model 00:01:54.960 --> 00:02:00.960 unable to browse the web or analyze files, but users are also limited to 30 or 50 messages a week, 00:02:00.960 --> 00:02:05.760 depending on the model version. On top of that, at least one of the models, the 01 preview, 00:02:05.760 --> 00:02:10.800 is up to four times more expensive than GPT-4O. On the plus side, it hasn't expressed a desire 00:02:10.800 --> 00:02:15.760 to bring about the downfall of mankind, at least not yet. The US government is allegedly asking 00:02:16.160 --> 00:02:20.640 NVIDIA and Apple to kindly use Intel's foundries. What? In case you weren't aware, 00:02:20.640 --> 00:02:26.160 Intel has been having a bit of a rough go. Apparently CEO Pat Gelsinger was so upset, 00:02:26.160 --> 00:02:30.880 he went and cried to mom. Or rather, United States Commerce Secretary Gina Ramondo. 00:02:30.880 --> 00:02:35.440 In a meeting with Ramondo, Gelsinger apparently expressed frustration with America's reliance on 00:02:35.440 --> 00:02:40.560 TSMC for chip production. Ramondo, like any support a parent would do, then set up private 00:02:40.560 --> 00:02:44.560 meetings with the shareholders of NVIDIA and Apple, hoping to convince the shareholders to 00:02:44.640 --> 00:02:49.520 convince the companies to give US foundries a shot. It's a bit interesting to see the US government 00:02:49.520 --> 00:02:53.680 indirectly help Intel in this way, specifically because they're currently withholding billions 00:02:53.680 --> 00:02:57.920 of dollars in funding from the company. Around $20 billion under the CHIPS Act, 00:02:57.920 --> 00:03:02.640 but that's because it's all dependent on Intel's promises to actively establish semiconductor 00:03:02.640 --> 00:03:06.960 facilities inside the US. It turns out that's kind of hard to do when your business is a little bit 00:03:06.960 --> 00:03:12.960 on fire, including your processors. It's a real fire sale. Meanwhile, TSMC's trial production 00:03:12.960 --> 00:03:17.600 at its Arizona fab last week produced yield rates on par with their Taiwanese facilities, 00:03:17.600 --> 00:03:21.040 probably making Intel's all-time low field just a little bit lower. 00:03:21.040 --> 00:03:26.320 And Pat Gelsinger's already pretty short. Is he? Oh, sorry, I'm just used to short people around all the time. 00:03:27.200 --> 00:03:31.120 Our quick bits, they're quicker than Vin Diesel when he's running late to a family reunion, 00:03:31.120 --> 00:03:36.240 because family. Speaking of family, the final specs of Intel's upcoming Arrow Lake desktop 00:03:36.240 --> 00:03:41.680 CPU family have been leaked by a Chinese outlet, Bench Life, merely a week after a UK retailer 00:03:41.760 --> 00:03:46.240 accidentally listed the chips on its site. I'm kind of starting to feel bad for Intel, 00:03:46.240 --> 00:03:50.400 I'm not. While the CPUs are expected to be officially announced on October 10th, 00:03:50.400 --> 00:03:54.480 Bench Life claims they won't be released until October 24th. Haha, better wait. 00:03:54.480 --> 00:03:58.960 But you should maybe consider buying one. For Pat Gelsinger's sake, if not anyone else, 00:03:58.960 --> 00:04:03.520 if enough sell, he might stop crying. But he definitely won't get taller. 00:04:03.520 --> 00:04:08.160 Several AI vendors such as Microsoft, OpenAI, and Anthropic have apparently committed to help 00:04:08.240 --> 00:04:12.080 combat non-consensual deep fakes and child sexual abuse material. 00:04:12.080 --> 00:04:16.720 Around the same time, Meta, Snap, and TikTok have formed an unholy alliance to stop the 00:04:16.720 --> 00:04:21.760 spread of self-harm content. Which is good, since those three companies' platforms are 00:04:21.760 --> 00:04:26.480 not exactly known for being good for people. The coalition hopes that they'll be able to share 00:04:26.480 --> 00:04:30.480 what problematic content they've found on their platform so that other members can easily find 00:04:30.480 --> 00:04:35.840 it on their own platforms. I never thought I'd see the day when Meta, Snap, and TikTok share 00:04:35.840 --> 00:04:39.840 data for good. I mean, they're not really doing it for good, they're just doing it because they 00:04:39.840 --> 00:04:43.360 have to, right? I mean, they did it voluntarily. Really? Are you sure? 00:04:43.360 --> 00:04:48.160 On a completely different note, Uber is adding autonomous taxis to its apps in Austin and Atlanta, 00:04:48.160 --> 00:04:52.240 thanks to a partnership with Waymo. So don't be too surprised if you hail a ride sometime in 00:04:52.240 --> 00:04:57.360 early 2025 and an empty car shows up. App users can apparently increase the likelihood of getting 00:04:57.360 --> 00:05:01.760 chauffeured around by a ghost by opting into autonomous rides, but it's not clear if that 00:05:01.760 --> 00:05:06.560 includes the option to be completely opted out. Can you request a Victorian ghost? 00:05:06.560 --> 00:05:10.720 While there have been concerns of robo taxis roaming the streets without human supervision, 00:05:10.720 --> 00:05:14.800 Waymo has been comparatively open about its crash statistics, which not only show that 00:05:14.800 --> 00:05:19.040 Waymo taxis get into severe crashes around a third as much as human drivers, 00:05:19.040 --> 00:05:24.080 but also that most of those crashes involve getting rear-ended by flesh-and-blood drivers on the road. 00:05:24.080 --> 00:05:26.880 It's because they're robots, isn't it? Yes. 00:05:27.840 --> 00:05:34.000 Yes. Google's notebook LM has a new AI feature called Audio Overview, which summarizes 00:05:34.000 --> 00:05:40.000 documents in the form of a lively podcast with banter between two AI hosts, not unlike the WAN Show. 00:05:40.000 --> 00:05:43.600 This is potentially a novel new way to take in otherwise boring information 00:05:43.600 --> 00:05:47.840 and a serious threat to the long-term employment of stoned white dudes with microphones in their 00:05:47.840 --> 00:05:52.640 basement, not like the WAN Show. Users can even download the file to take it on the go, 00:05:52.640 --> 00:05:56.320 though they should be conscious of the fact that LLMs, much like podcast hosts, 00:05:56.320 --> 00:06:00.960 are extremely prone to what we in the biz call just making shit up, like the WAN Show. 00:06:01.840 --> 00:06:06.880 And a collaboration between LEGO and McLaren has resulted in a fully functional replica of a McLaren 00:06:06.880 --> 00:06:14.320 P1 made out of 342,817 LEGO Technic pieces. Plus, of course, a metal frame, standard wheels, 00:06:14.320 --> 00:06:20.000 and a typical car battery. Look, imagination be damned, you can't make everything out of LEGO. 00:06:20.080 --> 00:06:25.520 But in 2018, LEGO did create a life-size Bugatti Chiron, but that model could only go 00:06:25.520 --> 00:06:32.240 12 miles per hour and had no steering. The LEGO P1, however, uses 768 electric 00:06:32.240 --> 00:06:37.680 LEGO motors to reach top speeds of 40 miles per hour. Plus, it can, you know, turn. 00:06:37.680 --> 00:06:42.880 And you'll get your turn to watch more tech news when you come back on Monday. Bye!