 Daily Tech News Show is made possible by its listeners, thanks to all of you, including Kevin, Paul Teeson, and Ali Sanjabi. Coming up on DTNS, is it bad that algorithms train on public images? What about when they're your public images? And Patrick Norton helps us understand the new chip war between AMD and Intel. This is the Daily Tech News for Tuesday, September 27th, 2022 in Los Angeles. I'm Tom Merritt. And from Studio Redwood, I'm Sarah Lane. Somewhere in St. Louis, I'm Patrick Norton. And I'm the show's producer, Roger Chang. We are gonna have a thumb war or a chip war, one of the two, possibly both. But let's start with a few tech things you should know. iRobot makes the Roomba line, which are robots that vacuum. iRobot also makes robots that mop and actually has one version that does both, but even the company says doesn't do it very well. Now, the company's announced the Roomba Combo J7 Plus. It offers proper two-in-one mopping and vacuuming in one unit. So to mop an arm lowers a pad to the floor and then lifts up and lays flush on the top of the robot for safe storage. It can distinguish carpeting and rugs from hardware or tile or linoleum, stuff that you want mopped, but you don't want vacuumed or vice versa. The Roomba Combo J7 Plus goes up for pre-order today for $1,099 and ships October 4th. Chinese augmented reality company NReal, that's the letter N, R-E-A-L, is launching its NReal air glasses in the U.S. Glasses were released in Asian markets last year and you can find them now on Amazon for $379. They are compatible with Android or if you buy a $59 dongle, $59 dongle also with iPhone. The air is intended to be a personal theater. So it's basically a big monitor. You would stream your TV service, your games. It's the equivalent of having a 130-inch screen three meters in front of your face or a 201-inch screen six meters away from your face. NReal air glasses project images in front of you without blocking the rest of your vision. So these are things where you can still see the room around you. NReal air glasses are the less expensive alternative to the $599 NReal light that launched in the U.S. last year. Bloomberg's Mark Gurman wrote in his Power On newsletter that it is unlikely Apple will do its usual October product announcement event next month. If that's true, Apple would reportedly still release new products but announce them by press release instead of with a big old fancy live stream where everybody looks like they're jumping through hoops at Apple's. Death Star. Gurman expects Apple to launch M2-powered 11-inch and 12.9-inch iPad Pros, new 14-inch and 16-inch M2 MacBook Pros, an M2 and M2 Pro Mac Minis, and possibly an Apple TV with an A14 chip. That's the one I want. Apple only held one autumn event in 2019 and in 2017. So not unprecedented, but slightly unusual. Yeah, yeah. I'll be happy if we don't have to sit there for 90 minutes listening to them. Just give me the press release. Tell me what it is. But you love the sitting. I do love sitting sometimes. Won't argue with that. There's another foldable phone on the market, folks. The Vivo X Fold Plus. It's an upgrade to the Vivo X Fold that was launched earlier this year. Has a faster and more stable Snapdragon 8 Plus Gen 1 processor. Ultrasonic fingerprint readers. A bigger battery, faster charging. The 6.53-inch OLED cover display has full HD Plus resolution and a 120Hz refresh rate. And then when you unfold it, you get an 8.3-inch screen with HDR10+. Also has a 16-megapixel selfie camera, both inside and outside. You can buy it in China for $9,999 yuan. That's about $1,415 U.S. And cheaper than the Samsung Galaxy Z Fold 4. But it doesn't have all the features that the Z Fold 4 has. And it will also not be available outside of China. Axios reports that on Friday, the U.S. Federal Communications Commission approved a long-delayed proposal to crack down on spam texts. Spam texts can include various tricks designed to steal money or personal information. And their volume now exceeds robocalls. If anything, they're quite a nuisance. The proposal passed on a 4-2-0 vote, now seeking comment from cell phone companies to explain how they plan to block text from numbers known to be illegal or fraudulent. Alright, we are dedicated here to making you the smartest in the room on the unfolding AI ethics text-to-image generation space. So I wanted to call attention to a motherboard article up today called AI is probably using your images and it's not easy to opt out. It does a good job of laying out what images are referred to in the data sets that are used to train algorithms, especially for text-to-image generators like stable diffusion, mid-journey, those kinds of things. I think for a lot of folks who listen to DTNS, it won't surprise you to hear that if an image is available publicly on the internet for anybody with a browser to look at, it can also be scraped and used in a data set. I know, I know. We should all should be clear though what this all means. Here's how it works for Laon, L-A-I-O-N, or the Large-Scale Artificial Intelligence Open Network, that's a nonprofit organization making large-scale machine learning models, data sets, and related code available to the general public. First of all, the image files themselves are not stored in the data set. Laon accesses the images in order to compute a similarity score between pictures and alt text, but it doesn't actually store the images. So the data set includes links, descriptions, and scores, but not those image files. Yeah, I think, even myself, when I think of this, I think of a big database of images, and that's not what it is. When an organization like Stable Diffusion or DeepMind or anybody that's using the Laon data set uses it, if it wants to access the images, it's going to have to redownload them from the address, from the URL. And if it needs more than links, descriptions, and scores, that's all on it. It's not in the data set. If the images are no longer at that link, then they won't get an image. So there's control on the part of the host of the image of whether that image is still available to the data set user. There are things on the internet that you may not realize are available, though. Motherboard notes a couple of examples of medical images being referenced in the data set. And in some cases, the patients and even the doctors that took the images were unaware that they had been posted publicly. Now, one possible reason for that could be the doctor used it in a journal with consent without really thinking about the fact that journals are posted online and sometimes images are posted and publicly available from those journals. Another possibility could be that there was a medical website that didn't properly protect images from being indexed by a crawler. Nothing necessarily even nefarious has to happen for that to be the case if it's just an improperly created robots.txt file or something. Yeah, so there's a bit of a brewing controversy over the ethics and legality of using publicly available data, even if you're a person who offered it up yourself. You didn't necessarily think about all of the uses that it might have way down the road. And as we've discussed before, it's arguably fair use since the photos aren't stored and the use made by the algorithm is clearly transformative. But the public perception may be that, well, that's unfair, since when I posted that picture on my blog, I didn't intend for it to be used to train an algorithm at all. That's just some photo of me being happy on a beach somewhere. Even further, if a doctor took a picture of somebody's rash, neither party may have expected it to be publicly available. And I'm using rash very, it could be a lot of things. But is there actual harm in these uses? And if so, what is that harm? And if there is harm, who's on the hook for this? Is it the websites? Is it the government, the algorithm makers, the algorithm users or some combo of all? Yeah, because again, I think it's worth reminding us because my head goes there. We're not talking about the picture of your rash being shown to everyone who uses the AI, right? It's a score created so that that algorithm knows what a rash looks like and it won't necessarily make yours when it outputs things. Keeping all of that in mind, though, I think it still rubs people the wrong way. Patrick, where's your head at on this? Oh, goodness. Part of me is like anything to make AI smarter is good. And part of me is like anything to make AI smarter is bad because, you know, I've seen the movies because I love science fiction. Mostly, I feel like this is one of those things where in 20 years, it's either going to be incredibly benign and it was incredibly important to advance, you know, AI and image recognition and processing, or there's going to be some horrible dog leg in the technology or the people using it or the events happening. And none of it's going to be particularly predictable at this point. And I also think it comes back to anything you don't want on the internet or anything you don't want people to find, you should not post on the internet. And... Yeah, and I think it's fair to say there's a separate issue of, well, wait, why did those medical images get out there? I don't think that happens as commonly as maybe this motherboard article might imply. I don't think that's the bigger problem. I think the bigger problem is, hey, I put these up on my blog because only my friends see my blog and now they're being indexed in a data set. Like that's creepy. But again, they're not in the data set. They're just being used to train the data set. And so they were public. So it's not a violation of your consent, but they also didn't ask, Sarah, how do you feel about that? I am not very up in arms about this. Do the rest of you, or have I ever mentioned that Twitter account, this person, does not exist? It's basically an algorithmic picture of a person who does not exist. And then they kind of have cheeky, like so-and-so lives in Vietnam and relationship status is undecided, you know, kind of thing. The whole idea is that you look at it and you go, it really looks like a human. That human does not exist. Somebody might look kind of like this person that was generated by AI in the world and they could kind of go, huh, it looks a little bit like me. But if something that is an image of any part of me or something that I've taken becomes a well-informed AI version of something else, I find it hard to get too upset about it. If I was really upset about it, boy would I have to take a lot of photos off the Internet that I have willingly put up over the years. And I don't really want to do that either. If it got to the point where I thought, you know, this could be really harmful to me, then I would. But I don't think we're there yet. I think there's probably not nearly as much harm as people are jumping to the conclusion because this is very transformative. It is quite unlikely unless you're famous and your image is already out there that these text image generators are going to generate you based on you being in the dataset. That said, I am also very much in favor of people being in control of their own data. So a standard that would put a robot's TXT style control on images to say if someone hasn't set the flag to yes, you can scrape it for AI, then don't scrape it, would be a good standard. I don't know that I would go so far as to say that should be the law, but just for ethics and for good human relations, I think that would be a good standard to put in place. I feel like somebody in the EU right now is starting to manifest the core of the idea you just described, right? Yeah, but as a law. First of all, we don't exist here, which is this person does not exist on Twitter. Sarah, you've just probably destroyed my afternoon because I'm going to be digging into this. I'm fascinated and horrified by this one. It's a good one. We don't exist here on Twitter. I am just fascinated, number one. And number two, it's really interesting because I think you may end up at some point where people will want to opt out or will refuse on political grounds or religious grounds or personal beliefs or whatever it is that they really don't want to be part of training AI, which is an interesting concept and maybe a big deal for reasons we cannot extrapolate at this time or may just be people being people because people do that. But it's very weird the idea that something's out there, looking at your images, looking at stuff you've created, learning from that, but then again. They're not looking at it. That's the thing, right? It's no different than the browser. The browser that goes to your web page and gets the images doing the exact same thing that this data set is doing, which is downloading it, processing it, and then discarding it. It's not keeping it. And there's no person looking at it. Yeah. I don't know. I think we've leapt into the metaphysical part of the conversation. But I am very curious to see where it goes. I'm very curious to see where it evolves because it seems like one of those things. I think, Sarah, now that either nobody's going to care or it will become a very big deal. And probably both. That's different points. Yeah. Seriously. Yeah. The next time I see my image on this person does not exist. Trust you, me. I will be upset. You'll be like, wait, I exist. Yeah. Hold on a second. This algorithm is getting weird. Or I just look like that many other people. Okay. So we're going to talk about Intel's Raptor Lake announcement in just a couple of minutes. But before we do, and because we have Patrick Norton here, let's get Patrick's take on the third party reviews of AMD's Ryzen Zen 4 CPUs. That's the recently announced Ryzen 7000 series, which launched on Monday. Flagship is the Ryzen 9 7950X at 5.7 gigahertz for $699. Competitive price, still kind of on the high end. The Verge was impressed by its performance though on creator apps. Less so by gaming use cases. The Verge also found some sleep issues with the new AM5 motherboard. Patrick, what have you gleaned from the reviews out there thus far? Oh my goodness. So it's been a really kind of crazy couple of months for hardware. We've had GPUs getting affordable. EVGA has walked away from the business. And right now Ryzen is dropping so many cores and they're running so much faster. Right? The 7950X 16 cores, 32 threads. The 7900X 12 cores, 2400 threads. Both of these are well over 5 gigahertz and max boost motor and their turbo mode. And that's crazy because that's, you know, this is a huge jump in clock performance, which is making them much more competitive with the Intel processors. They also dropped in like one megabyte of cash, L2 cash per core. And that alone can give a serious boost in performance. And architectural changes in Ryzen is adding something like 13% more instructions per clock cycle. So between that and the boost to the huge boost in AMD's clock speeds, you've got this tremendous performance boost. So benchmarks, most of the benchmarks out there are contesting them against Intel's Core i9-12900KS. That came out at $900. It's selling for $730. I love Puget Systems productivity benchmarks. It's PugetSystems.com. If you're rendering an Unreal engine, which is really all about the cores, the 7900X and the 7950X stomp the Core i9 flagship. Photoshop, there's big gains for the Ryzen chips, like as much as 25% to 30%, depending on which aspect of Photoshop you're benchmarking. So they're now in parity with that $900 Intel CPU. Premiere Pro, After Effects, DaVinci Resolve Studio. It's really all about the codecs you're using. And you should go to PugetSystems.com to learn more about that. Gaming is really all over the map, right? Linus Tech's tips got really deep. And they have a review that AMD is in trouble, right? That's the Ryzen 7000 full review. They did a tremendous amount of benchmarking. So the 7600X, AMD said the Ryzen 5 7600X, a $300 CPU, will outperform Intel's Core i9-12900K in gaming at less than half, or like less than half the price. They didn't quite beat the 1200K, but Linus Tech's tips found it within 3% overall, which is a big deal. It's impressive. Less impressive. The flagship 750X is maybe 5% faster than the 1200KS, right? So you have this relatively inexpensive processor that has this huge performance boost, but if you spend twice as much on a Ryzen processor, you're not getting twice as much gaming performance. So the high-end Ryzen 7000 CPUs, really competing with Intel's last-gen flagship, but in many cases, they're actually slower or a percentage point or two faster, just that little bit faster than the $420 Ryzen 7500X3D, which is a last-generation Ryzen processor, which also uses DDR4 and AM4 motherboards that are cheap, cheap, cheap. All the new 7000 series processors are using DDR5, which is like twice as expensive as DDR4. They also include integrated GPUs in case you can't find a GPU, which I think would have been more important four months ago than it is now. Technically, AMD says it's not for games, but it'll play older titles just fine for now. It's a good bump, but the big question is what's going to happen when the 13-gen Intel CPUs actually ship or given the hand of testers, and we see what Intel's 13th-gen performance is against the Ryzen 7000 series. Indeed. And in fact, Intel conveniently decided to announce details on that just a day after AMD shipped their new 7000 series chips. We're going to get to that in a second. Real quick, if you want to email us, you have a thought about something on the show, send us an email. Our email address is feedback at dailytechnewshow.com. Intel announced those details about the 13th-gen Core chips, aka Raptor Lake. The flagship is the Core i9-13900K with 24 cores. So eight Golden Cove performance cores, the P-Cores are now Golden Cove, and 16 Efficiency Cores. That's up to 5.8 gigahertz max turbo frequency. Eight more eCores than the 12900K, and Intel says the 13900K should be 15% faster than the 12900K at single threaded tasks and 41% faster on multi-threaded tasks, things like video encoding and 3D rendering. Now, Patrick, what did you get out of Intel's? I thought very efficient announcement of these new chips. You know, if I'm AMD, I'm going rock-roll. You know, it'll be very, very, very, very interesting to see what the real-world performance is in the Core i9-13900K, because if it is 41% faster at multi-threaded tasks, that is a huge deal and is going to be enormously difficult for AMD to compete with. There were rumors that there were going to be price jumps on these, but it seems like the majority of the Intel processors are priced similarly. I think the Core i5s get like a 10% or 15% bump. There's a lot going on here, right? Intel 7-Process, 3rd Gen Superfin Transistors. Intel did not make any comparisons to the AMD 7000 series. They did say they expect to hit 6 gigahertz in 2023. These are super power-sucking chips. I don't know how else to say it, right? You know, it's, I think, up to 253 watt max in turbo mode, which to me seems like a huge number. So if your power supply is a little sketchy, you might want to upgrade that. I'm also kind of crazy. I always find it kind of maddening. Intel has dropped 22 processors in the 13-Gen lineup. They are always dropping a tremendous amount. PCI-Gen 5 up to 16 lanes. DDR5 is 5600, DDR5 is 5200. Mothermore manufacturers will be able to offer DDR4. That's important given the expensive DDR5 right now. But I think it's going to be interesting to see what the performance looks like, especially on fairly heavy-duty apps like Premiere or Unreal rendering, because Intel does split those cores into performance cores or PCORs and efficiency cores or efficient cores. Not all cores are equal for monster, multi-threaded, incredibly taxing applications. I'm very, very curious to see how these work out. I'm also very curious to see if anybody does benchmarking on the 600-series motherboards, which will still support 13-Gen chips. The new Z790 motherboards, they're offering like eight more PCI-4.0 lanes, more USB 3.2, Gen2 by 2. So Intel was comparing a little bit between the 1300K, both beating and losing to AMD's Ryzen 7 5800X3D, which loses to the Ryzen 7000 series flagship chips in applications, in sort of high-end creative applications, but is not much faster in terms of gaming. So short answer. Man, there's a lot of processors. Man, there's a lot of prospects. And man, I'm very, very curious to see what the flagship performance looks like when you're looking at things like Premiere or Unreal or some of the other creative applications. And with gaming, I suspect there's going to be a pretty hefty bump in gaming performance from Intel, all of which is speculative. Now, we're getting boxed versions of these processors October 20th, $589 for the 1300K, $319 for the 1300K, the i5. They'll start appearing in systems too. Alienware's Aurora R15 was mentioned. They are keeping the prices the same, which is a bit of a surprise, except for the Core K series. The Core i3 1300K is the only chip to get a price increase to $319. And the speculation I read on Tom's hardware was that Intel is betting that AMD isn't as competitive head-to-head with that chip. And so it can make its cash there because they sell a lot of that mid-range chip. I'm curious to see this. I thought it was the Core i5 that were getting the bump, but I will trust your sources on this one. I don't know. Again, I'm very, very curious to see what this looks like because Intel has just gotten so aggressive and the performance. I'm really curious if they're delivering a 15% boost in single Core and what, 41% multi-threaded applications, I think the whole Ryzen Intel relationship is going to shift radically October 20th. Yeah. Do you think it does shift radically? Or do you think we see just sort of a nudge ahead? Right? Do you think we see AMD suffer in this? I feel like people get very hyped up and start to overstate things for small differences sometimes. These are more significant differences than I think we usually see, but how significant are they? Okay, so my first thought is that one, you're absolutely correct as a CPU enthusiast and a longtime Ryzen purchaser for the last few years. I am very curious. I don't have a lot of skin in the game, but I'm still very curious to see what performance looks like. As you pointed out earlier, Intel absolutely dropped this announcement just to be cruel, I think, to AMD's release. Oh, they would never. Oh, they would. I like Pat Gelsinger, but I wouldn't want to get into a knife fight with him. But the fact that they were like, hey, we're going to compare it to this CPU, which is actually this previous generation Ryzen CPU, which is actually outperforming or performing very, very close within like 3% or 5% of the two most expensive current generation Ryzen CPUs. I smell a lot of gamesmanship here. I smell a lot of internal information passed maybe around from partners for both companies. And off the top of my head, I'm guessing it's not going to be a significant jump. That said, 41% multi-threaded performance sounds like it could be significant in some of the applications where AMD currently has the lead, primarily because they have so many of what Intel would call performance cores, right? I wouldn't be surprised to see evaluations in three months' time that are, AMD is good for this, let's say gaming, Intel is good for this, let's say, you know, video rendering, and those are my guesses. But something along those lines, rather than one is better than the other, more of a differentiation, because it seems like that may be what Intel's going for here. Well, yeah, but right now, right, when you look at the performance analysis of the 7000 series chips, the productivity apps or the creativity apps or some of them, especially like Unreal Engine, because of all those cores, that's where AMD has this huge advantage. If there's a 41% boost in performance of the 12900K, that starts to really eat away and where AMD has some really strong, compelling reasons to buy them, at least in workstations. With gaming, I think it's much more of a wash at this point. And 15% for the same price, I'd be looking at Intel at that point if it's legit. And 41% of the same price, we'll see. I always find it interesting to see what numbers a company actually uses to generate their... It's this much faster, because there are lies, there are damn lies, and there are statistics, and then there are benchmarks. A couple other notes here. A new thread director baked directly into Windows 11 2H22 update that just came out is targeting the Raptor hybrid architecture, so that's worth knowing about. Also, Intel announced the Arc A770 GPU, 65% faster, they say, than the competition in ray tracing. That'll be available October 12th at $329. And Intel also announced an app called Unison for Windows on Intel machines. It offers the ability to get text calls and share files between a PC and either an Android or an iOS phone. There's support for VPNs, firewalls, and IT management. Microsoft's, your phone does some of that, but it only works with Android, and it doesn't do all of it. Unison will work on some 12th gen EVO PCs, from HP, Acer and Lenovo. That'll start this autumn, and then it'll come to the 13th gen EVO systems sometime next year. Patrick was like, huh. Huh. It's such an Apple thing for Intel to do. I also desperately wish it would go back to older generation Intel powered PCs, because it seems like if it actually works, it could be incredibly useful. Yeah, indeed. Alright, let's check out the mailbag. Alright, this one comes from Mike. I just had to mention the Australian data breach of the Telco Optus. It's been on the news here ever since it broke last week. I've had dozens of emails from institutions promising to be on the lookout for fraud and scams with Optus promising credit monitoring which brings back memories of the Equifax breach. Mike says the main twist today is that the threat actor has had a change of heart and has revoked the ransom and has promised to delete the data. I'm sure you've heard about it, but I just thought I'd give some info from somebody here on the ground. It is everywhere. News, politics, even my mom asked me about it. I'm still getting emails and notifications from my banks and crypto exchanges to be on the lookout for related scams. Seems all the fuss may have changed that threat actor's mind. Mike says I'm a long-time listener, supporter, and pacemaker recipient, and it hasn't been hacked yet. I'm glad to hear that your pacemakers have been hacked. And thanks for putting a flag on this for us. Bleeping Computer had a good article today about that person feeling the heat from law enforcement and saying I've deleted the only copy of it. I give up my ransom. I'm not going to do anything. Please don't come and arrest me. Good luck with that. It's been a terrible mistake. I'm just going to go back into the bushes. And it is one of those attacks where the stuff that was leaked out was emails, addresses, driver's license. They're going to have to reset some people's driver's licenses to give them new numbers. But no payment information. So this is the kind of stuff that will be used for phishing scams to try to trick you. So yeah, if you are in Australia and a customer of anything related to telcooptis, be aware of that. Thank you, Mike. Thank you, Mike. And thanks to everybody who writes in with really good intel for us. If you have some questions, comments, we'll just take all of it. Feedback at dailytechnewshow.com. We will also take Patrick Norton be on the show whenever he so desires. Patrick, it's been great as always. Let folks know where to keep up with your work. Oh my goodness. AVXL.com has kind of been on a temporary hiatus. I started a new job, a place called CTI Technologies or Conference Technologies. And if you need AV integrated solutions, well, that's an ad, so we'll stop right there. But AVXL is the place to find me and Robert Herron talking about audio and home theater and headphones and all that good stuff. Please look for it on your favorite podcast. Well, and congrats on the new gig as well. Thank you. Health insurance. Oh, man. It's good stuff. You don't want to not have health insurance. Take it from your girl over here. Special thanks to Brian LeCrette, who's one of our top lifetime supporters for DTNS. Our top lifetime supporters always get a round of applause. And today, Brian, it goes to you. Thank you for your support. You can do what Brian did and continue to support us at a high level for a long time and we're very appreciative of that. Or you can be brand new and just start supporting us now. And we'll be very appreciative of that and mention you on the show tomorrow, patreon.com. Speaking of patrons, stick around for our extended show, Good Day Internet or GDI, which rolls right in after DTNS wraps up. But if you want to catch DTNS live, you can. We record live Monday through Friday at 2,800 UTC. Find out more at dailytechnewshow.com slash live. We're back tomorrow with Scott Johnson joining us. Don't mess it. Talk to you soon.