WEBVTT 00:00:00.000 --> 00:00:05.840 If you find TechLink to be an enrichment in your life, I invite you to write a kind little comment 00:00:05.840 --> 00:00:11.680 below. When I'm all snuggled up in bed tonight, I'll read some of them to my cat. Then he'll have to 00:00:11.680 --> 00:00:18.560 respect me. In the wake of an Oxford study that suggested barely anyone uses AI tools on a daily 00:00:18.560 --> 00:00:23.920 basis, three major AI companies have rolled out feature updates for their chatbots in an attempt 00:00:23.920 --> 00:00:29.600 to make them useful for someone other than coders and tech bros left over from the crypto bubble. 00:00:29.680 --> 00:00:35.280 OpenAI's ostensibly less evil cousin, Anthropic, announced tool use for Claude, 00:00:35.280 --> 00:00:41.520 allowing it to perform tasks and manipulate data. AI search engine Perplexity announced pages, 00:00:41.520 --> 00:00:47.360 which will create research reports on any topic that can be customized, shared, or presumably 00:00:47.360 --> 00:00:53.680 handed into your teacher. And OpenAI just gave free chat GPT users the ability to upload images 00:00:54.480 --> 00:01:00.320 and files and use custom GPTs. Those features were previously paywalled, which, as this fan points 00:01:00.320 --> 00:01:15.600 out, obviously points towards something special coming for paying users. OpenAI also just announced 00:01:15.600 --> 00:01:23.200 chat GPT EDU, which basically gives schools an education discount to convince them to use it, 00:01:23.200 --> 00:01:28.400 and it comes on the heels of a report claiming the company finally signed that rumored Apple deal 00:01:28.400 --> 00:01:33.840 to put their AI tech in the next iPhone. That same report claims Sam Altman has discussed 00:01:33.840 --> 00:01:40.960 transforming OpenAI from a wacky capped profit controlled by a non-profit into either a B 00:01:40.960 --> 00:01:46.720 corporation, allowed to care about something other than benefiting shareholders, or a regular old 00:01:46.720 --> 00:01:52.160 for-profit company. But it's only because that might be necessary to achieve a utopia designed by 00:01:52.160 --> 00:01:56.640 Sam's machine god and populated by sentient robots after they've killed us all off, okay? 00:01:58.000 --> 00:02:03.040 Trust him, it's worth it. TikTok is working on a modified version of its recommendation algorithm 00:02:03.040 --> 00:02:09.520 that would not share any backend with its Chinese counterpart, Duyen, and thus potentially save 00:02:09.520 --> 00:02:15.120 the app from being banned in the US. That's all according to insider sources who spoke to Reuters, 00:02:15.120 --> 00:02:21.680 but not according to TikTok, who immediately tweeted that the report is misleading and inaccurate. 00:02:21.680 --> 00:02:27.680 But seemed to be referring specifically to the claim that a US-centric algorithm 00:02:27.680 --> 00:02:33.920 would enable ByteDance to sell TikTok. They didn't explicitly deny they're working on such an 00:02:33.920 --> 00:02:38.880 algorithm. In fact, Crystal Hu, the report's co-author, says that when she reached out for 00:02:38.880 --> 00:02:45.520 comment on the story, TikTok told her the info was correct. To be fair, they probably didn't 00:02:45.520 --> 00:02:49.680 read the whole thing before getting bored and swiping up. But even if TikTok survives in the 00:02:49.680 --> 00:02:54.800 US, it may have to deal with upcoming bills like this one from New York's governor that could 00:02:54.800 --> 00:03:00.240 ban smartphones in schools. And then how will kids ignore their history class to watch another 00:03:00.240 --> 00:03:12.720 kid tell them that Helen Keller wasn't real? On TikTok? Yeah, Google has confirmed that this 00:03:12.720 --> 00:03:17.680 week's massive document leak revealing key details about how its search algorithm works 00:03:17.680 --> 00:03:23.120 is real after taking two days to cover their ears and pretend everything's fine as required by law. 00:03:23.120 --> 00:03:28.080 The tech giant didn't get into specifics. They only cautioned against making inaccurate assumptions 00:03:28.080 --> 00:03:33.600 about search based on out-of-context, outdated, or incomplete information, suggesting that the 00:03:33.600 --> 00:03:38.480 public trust Googles demonstrably inaccurate public statements instead. Google also took a 00:03:38.480 --> 00:03:43.840 moment to put up a blog post explaining why Google's search's new AI overviews told people it was 00:03:43.840 --> 00:03:49.440 okay to put glue in pizza sauce and eat rocks last week. Step one, it's good for you. It gave 00:03:49.440 --> 00:03:55.280 those answers simply because prior to those screenshots going viral, practically no one 00:03:55.280 --> 00:04:00.800 asked Google how many rocks they should eat. How would Google's search know that it wasn't okay 00:04:00.800 --> 00:04:06.720 to eat one little rock if it had never thought about it before? Google says AI overviews cited 00:04:06.720 --> 00:04:12.560 a satirical article recommending eating rocks because of a data void issue. There's just hardly 00:04:12.560 --> 00:04:17.360 any info on the web about whether eating rocks is okay, formatted as an answer to the question of 00:04:17.360 --> 00:04:26.400 how many rocks one should eat. We can all chuckle at this understandable error. It's quick bits time. 00:04:28.080 --> 00:04:32.800 Why haven't you written a comment yet? Spotify has announced that they will be giving refunds to 00:04:32.800 --> 00:04:38.240 customers who bought car thing. After the company announced they would be remotely bricking all 00:04:38.240 --> 00:04:44.880 car things, they initially offered purchasers nothing. Some who complained were offered a 00:04:44.880 --> 00:04:49.440 few months of Spotify premium, but wouldn't you know it? The company suddenly started offering 00:04:49.440 --> 00:04:55.680 refunds not long after a class action lawsuit was filed accusing Spotify of misleading customers by 00:04:55.680 --> 00:05:00.480 selling a product that would quickly be rendered unusable. But are they doing anything about all 00:05:00.480 --> 00:05:06.960 the e-waste they're creating? Has the ocean or any of the fish hired a lawyer? I didn't think so. 00:05:07.040 --> 00:05:12.160 Say you're a fish. Prove it. Google Cloud has released a public explanation for how it wound 00:05:12.160 --> 00:05:17.760 up accidentally deleting a customer account two weeks ago. Namely, somebody on Google's end 00:05:17.760 --> 00:05:23.840 accidentally left a parameter blank. Now that might not sound like an issue worthy of an entire 00:05:23.840 --> 00:05:29.440 blog post from a multi-billion dollar company, but the customer in question was Unisuper, 00:05:29.440 --> 00:05:36.240 an Australian pension fund charged with managing 135 billion Australian dollar-y-dos. 00:05:37.120 --> 00:05:41.760 On the bright side, the accounts have been successfully restored. On the less bright side, 00:05:41.760 --> 00:05:46.800 apparently your life savings can be wiped out in an instant if a dude named Robert Droptable 00:05:46.800 --> 00:05:55.520 decides to sign up for a 401k. The internet archive, the organization that runs the Wayback 00:05:55.520 --> 00:06:00.880 Machine, was hit with a several-day-long DDoS attack involving tens of thousands of fake 00:06:00.880 --> 00:06:05.360 information requests per second, according to the Archives Director of Library Services. 00:06:05.360 --> 00:06:10.640 While the exact motivation for the attack is unknown, a group called SN Blackmetta has taken 00:06:10.640 --> 00:06:15.600 credit for the assault by posting what appears to be the profile picture of an edgy 14-year-old 00:06:15.600 --> 00:06:20.400 from the early 2000s. Who could possibly understand what's going on in his dark mind? 00:06:20.480 --> 00:06:27.920 AI company The Simulation has announced ShowRunner, an app that wants to be the Netflix of AI, 00:06:27.920 --> 00:06:34.480 with AI-generated original series. The TV shows are pretty creative. There's one about super 00:06:34.480 --> 00:06:40.560 intelligent AI devices, a family drama about two siblings that create an AI version of their dead 00:06:40.560 --> 00:06:45.920 mom. There's even a horror anime about a future where humans choose to physically augment themselves 00:06:45.920 --> 00:06:52.480 with AI technology. Even worse, they want watchers of these shows to generate their own episodes. 00:06:52.480 --> 00:06:57.760 Yes, because the thing that will surely make AI-generated TV shows better is fan fiction. 00:06:57.760 --> 00:07:02.560 And a creator on Chinese video site Billy Billy has gone viral on social media, 00:07:02.560 --> 00:07:07.600 thanks to an incredibly intricate custom tank simulator for World of Tanks, 00:07:07.600 --> 00:07:13.760 using soda bottles for shells, MetaQuest 2 as a periscope, a Logitech driving wheel, 00:07:13.760 --> 00:07:19.840 and some kind of crank situation to control the main gun. It looks awesome. And frankly, 00:07:19.840 --> 00:07:24.080 if we could take all the resources currently being used to generate AI slop for the internet 00:07:24.080 --> 00:07:28.960 and build more stuff like this, I wouldn't have to fish for nice comments to read to my cat to be 00:07:28.960 --> 00:07:34.720 happy. I was joking. I don't have a cat. What I actually require to be happy is that you come back 00:07:34.720 --> 00:07:39.760 on Monday for more tech news, okay? Or I will cry. No pressure.