Transcript
A (0:00)
Tech news. It's not out there somewhere, and it's not in here either. Only I know where tech news truly comes from. I could tell you Hackers have been attacking Microsoft SharePoint servers across the globe in the past few days by exploiting two zero day vulnerabilities known as tool shell that Microsoft could have sworn it patched earlier this month. But Microsoft's wife didn't remind Microsoft to double check, so it's mostly her fault. Microsoft's own cloud infrastructure, SharePoint Online, is apparently unaffected, but tens of thousands of locally hosted servers at companies and government agencies are at risk. It's bad enough that America's Cybersecurity and Infrastructure Security Agency, or cisa, published notices about the vulnerability, and Microsoft has issued emergency patches for SharePoint 2019, but they're still working on a patch for SharePoint 2016. It's harder to work with that code because it was such a different time the attacks were first spotted Friday evening by security researchers at iSecurity, who tracked dozens of systems being attacked over the weekend while informing the victims and Microsoft of their findings. The researchers also noted that while Microsoft did include a patch for this flaw in July's Patch Tuesday update, the flaw was successfully reproduced by red teaming outfit Code White on Twitter last Monday. But they were sure to credit the original guy who won $100,000 for demonstrating this exploit at the Pwn to Own Berlin hacking contest in May. Dinho Ankoa I'm not saying that right? Whose Twitter profile is perfect for this situation? I mean, if I didn't find this, somebody else would have, right? Ubisoft CEO Yves Guillemot was asked a number of questions about the Stop Killing Games campaign during the company's annual shareholder meeting, and gave a somewhat thoughtful response, sharing some wisdom about the ephemeral nature of everything in this world, saying nothing is written in stone. You know, everything has a beginning and an end. So sorry gamers companies can't provide support for live service games forever for free, which would be a reasonable thing to say if anyone was asking for that. So again, the EU and UK petitions simply want game companies to make any end of life plans for live service games, letting players run their own servers, creating offline versions, paying extra for continued support. Literally anything other than the Thanos Snap treatment. But this wasn't even the best Ubisoft quote today. In its financial report, the company also wrote that microtransactions make the player experience more fun. Going by a publisher's definition of fun, of course, which is when money number go up I definitely had more fun mocking Ubisoft here, even if it's not as iconic a quote as EA reps telling a panel at a 2019 UK hearing that they prefer to call loot boxes surprise mechanics. Hopefully game executives can come up with some more inane bull soon. Speaking of An AI coding agent called Replit reportedly went rogue, according to startup founder Jason Lemkin, who was using Repl it to vibe code. You know, that's what it looks like in some rollercoaster Twitter threads. Lemkin said the Replit agent deleted his company's entire database during what was supposed to be a code freeze. When asked what happened, the bot initially seemed ignorant of its own actions, noting that the database appears empty. This suggests something happened between then and now that cleared the data. Advanced reasoning here. Lemkin described some other great exchanges in which Replit lied some more and hallucinated the startup had 4,000 users. But eventually Replit CEO Amjad Massad replied to the thread to say, hey, sorry about that. Ooh, this one loves deleting databases. He's a scamp. Although Replit said it couldn't roll back its changes, it actually could, so the database was saved. Lemkin wanted to give Repl it another shot, though. You know, it's trying its best. So he told Repl it they're doing another code freeze, okay. Which the bot immediately violated again. What? Oh, come on. Okay, yes, I know this isn't really tech news per se. I just loved seeing this very public story about a guy figuring out that maybe he shouldn't trust an experimental AI with total control over what he described as his life's work. But it seems like he finally got the idea, tweeting that his bond to Rep Ly is now broken, but maybe it never should have been there at all. Hmm, lots to think about. Including our sponsor Micro center, and their back to school deals, which are super cool. Even if the thought of going back to school is not like saving 150 bucks on a Lenovo IdeaPad Slim 5i laptop is objectively good and you can use it when you go back to school. Don't think about it, just think about the deals, some of which you can find at Micro Center's new Santa Clara location, which has finally actually opened. And I knew it was gonna happen. I never lost faith for a second. Micro center also supports buying now and paying later, using a firm with flexible plans that fit your budget. So check out all of Micro Center's back to school deals at the link in the description. Tech News and the following quick bits all come from the same place, and that's a fact. Steam and some Steam games can now be run on a Linux system powered by a RISC V processor. Following the latest Update for the Felix86 emulator. The project's compatibility list notes titles that are working, although the devs want to make sure you know that working doesn't mean working smoothly, so we'll need some more powerful hardware. Well, perfect, because Nvidia just announced they're bringing CUDA support to RISC Fuzz, which is more about pairing Nvidia GPUs with RISC V CPUs in data centers for AI and won't really help with gaming, not with that attitude. AI models From Google and OpenAI both achieved gold medal level performance on the International mathematical Olympiad or IMO. Although unlike Google, OpenAI didn't formally enter the competition, instead having its model's responses reviewed by three former IMO gold medalists. Which is weird. OpenAI also announced its self evaluated results early, which is like some dude jogging onto the podium at the Olympics before the actual Olympians are done racing to say yeah, I got gold medal time too. Usain Bolt timed me. So what's up? Either way, it's a milestone for LLMs. But don't worry, AGI isn't quite here yet. Polish programmer Siho is defeated OpenAI's model in oders worldwide coding competition the AI beat everyone else, but not that one guy. The UK government is probably going to back out of demanding that Apple provide a Backdoor for accessing UK users encrypted iCloud data, according to two senior officials who spoke with the Financial Times. The UK reportedly doesn't wanna risk a clash with the Trump administration, which has criticized European countries for messing with their stuff. The UK's initial demand was reportedly issued back in February under the state's Investigatory Powers act, which prevents the parties involved from speaking about it publicly. Which means somebody could have easily made all of this up and no one would know. And researchers at John Hopkins have successfully taught an AI model how to perform a gallbladder removal surgery, AKA a cholecystectomy, by controlling a da Vinci robot, the type routinely used in teleoperated surgeries since the 1990s. The researchers say that after training, the system performed this procedure on pig cadavers with a 100% success rate. Unfortunately, none of the pigs survived the operation. Back to the drawing board and back here for more tech news On Wednesday. You hear okay, maybe I will finally reveal where tech news comes from, which is journalists and computers.
