Tech News Weekly Episode 300 Transcript
Please be advised this transcript is AI-generated and may not be word for word. Time codes refer to the approximate times in the ad-supported version of the show.
0:00:00 - Jason Howell
Coming up next on Tech News Weekly. It's me, jason Howell. Mikah Sargent is off today so he will not be sitting next to me. He'll be back next week, don't worry about it, but we've got a couple of great interviews for you. First, talking with Kyle Barr from Gizmodo. He talks all about the Books 3 data set and the lawsuit that it's involved with. Also, we talked with Joseph Cox. Now with 404 Media. They actually started a media company to talk about stories like the one that he discusses credit header data and the payload within, how easy it is to get access to that data and what people are doing with it, what criminals are doing with it. And then, finally, I have in my hot hands here the Samsung Galaxy Z Fold 5. I give you my full thoughts on this in a detailed review coming up next on Tech News Weekly. This is Tech News Weekly Episode 300, recorded Thursday, August 24, 2023, inside the Books3 AI Dataset.
0:01:09 - Leo Laporte
Listeners of this program get an ad-free version if they're members of Club Twit. $7 a month gives you ad-free versions of all of our shows, plus membership in the Club Twit Discord, a great clubhouse for Twit listeners. And finally the Twit Plus feed with shows like Stacey's Book Club, the Untitled Linux Show, the GizFiz and more. Go to twittv slash Club Twit. And thanks for your support.
0:01:36 - Jason Howell
Hello, welcome to Tech News Weekly, the show where we talk to people who are making and breaking the tech news. This week, it is just I, jason Howell, sitting at the table. That's why you don't see the full table and a Mikah Sargent over there. He is at the podcast movement, moving and shaking podcast moving and shaking. That's what he's doing there. So it is just me today, but I am thrilled because we've got some great interviews and we're going to dive right in.
Large language models. They need training data, right. When we talk about AI, we talk about all the training data that's required to go into them in order to train them to do all these magical things that we keep seeing them doing. Large sets of training data, in fact, very large. So what is the data and where is it actually coming from? And how does a rights holder actually know when their property has been used to train an AI system? Well, joining me to talk about the Books 3 dataset is Kyle Barr. Wrote about this for Gizmodo. Welcome, kyle. It's good to have you here. Great to be here, jason. Thank you. Yeah, this is an interesting story to kind of see how this is evolving over time. So you wrote about this. It's an excellent article. People should definitely go to Gizmodo and read your take on this. But first let's start with the AI. What does it have to do with this particular dataset, this Book 3 dataset's existence?
0:02:55 - Kyle Barr
So the AI is kind of large internet archive Well, that's how they would describe themselves. They want to collect a lot of data that might just go extinct otherwise. You know, just a lot of old books or material that most people can't find. But it's also what's called like sometimes referred to as a shadow library or sometimes referred to, as you know, a piracy archive. So a lot of data just gets put onto these, onto websites like this, and one of those datasets in this case was the library bibliothic. So that's another kind of piracy archive and that's actually related to Books 3. That's where Books 3 came from.
0:03:37 - Jason Howell
Okay, so the AI like is it kind of like the internet archive, but the internet archive, if the internet archive was more like pirated material, less just, you know, storing the contents of the web, that sort of thing.
0:03:51 - Kyle Barr
Essentially, it's one of those gray areas of piracy ethics. Right? Pirates don't necessarily, especially if you're pirating books. Their stated goal isn't necessarily to screw over rights holders. They're more interested in making sure people have access to the information, and that's ostensibly what the AI is kind of there for.
0:04:15 - Jason Howell
Okay, All right, so that's interesting. I had not heard of the AI prior to reading articles, so I find that very fascinating to know that that exists and what it's doing. Who created the Books 3 dataset Like? Do we have an idea of who's behind it? And is it easy because I mean, you know this is being referred to as like a dataset full of books that you know some of the rights holders didn't explicitly give their permission for to be in there Is it easy to identify some of the titles that were actually inside, and how has that done?
0:04:48 - Kyle Barr
Right. So the person who created Books 3, their name is Sean Presser. He's very open about his work with Books 3, because, again, it's about ethics for him and we'll probably get more into that later. But as far as how easy it is to identify, like I said, it comes from bibliotik, which, once you download the files, you have all the different books in there, so you can identify what's in there. The actual Books 3 data is a little bit more complicated because it's designed for training AI. Essentially, presser took bibliotik, got as many books as he could and then transformed those books whether if they were in PDF or some other file format into atxt format, which is how the AI is able to train on it, because it just needs words, right? It doesn't. It can't identify, you know, in a PDF document necessarily what words are there, but it can in atxt file.
0:05:42 - Jason Howell
Hmm, interesting, and I think I saw somewhere, like I was looking at, a little pieces of the data, like the titles of the books weren't necessarily transferred over. Maybe I'm just construed that, but it was just the content, so it's just like. So it's not. In other words, it's not like you get this data set and, yay, I have access to all of these, you know, these books that I can read. It really is formatted in a kind of an inconvenient way for humans, but a convenient way for AI to do what it does.
0:06:12 - Kyle Barr
Exactly, and that's the whole point. Right, this wasn't. This was designed from the get-go specifically to train AI and specifically train open source AI or AI projects that aren't open AI or Meta or Microsoft. Right, this was from the get-go, back in like 2020, when some of the stuff was happening. When GTP3 came out, presser was one of these people looking at how big GTP3 was and saying, well, what happens to the open source guys? We need something to compete. Let's make an open source data set. And that's where this comes in and also where it gets introduced to the pile, which is another 800 gigabyte data set that was used to train a lot of different AI.
0:06:58 - Jason Howell
Okay, okay and yeah, you wrote about Sean Presser. He had a really great point to share about the balance of power with AI tools. Talk a little bit about that, because I think he makes a really valid point as far as access to these tools and who controls them.
0:07:16 - Kyle Barr
Yeah. So it kind of goes back to that sense of there are small companies out there that want to make AI, but the big problem is the mass amounts of data that's required to train AI. So if, let's say, open AI is training their stuff on hundreds of gigabytes of files and all that has to be formatted correctly and it costs time and money to do that, like an open source company or somebody that's a startup has to start all from scratch, or they find these data points on the internet, which allows them to actually start from a place, instead of from ground zero, where they have no data, absolutely like no text, nothing to actually share with the AI to actually train it, and so that's kind of where he's coming from. If open AI can train their stuff on this mass amount of data, which a lot of it actually already comes from pirated sources, then why can't an open source company do it as well?
0:08:24 - Jason Howell
Hmm, yeah, interesting. Is it just books inside of this data set, or is there other material contained within there as well?
0:08:32 - Kyle Barr
So obsessively. It's supposed to be just books, though I can't verify how much of that data is in books.
0:08:39 - Jason Howell
You haven't gone through all 250,000 or whatever. Maybe much content is in there. Closed by page.
0:08:45 - Kyle Barr
Closed to 200,000. Yeah, no, I mean, if it comes from the bibliotech library, most of it is books. Yeah, and how many of those books are like currently on shelves now? Some of them, as, as evidenced by recent lawsuits and by the Danish rights holders group that took down the content from the original source.
0:09:08 - Jason Howell
Yeah, right, right Now, you mentioned recent lawsuits, which is a great kind of segue into probably you know a large reason why this is even a big deal. Books three was mentioned in the Sarah Silverman lawsuit. That is very recent as the repository used to train Meta's Lama AI. You also wrote that books one and books two kind of the predecessors to books three, as you can imagine from the titles there makes up 15% of GPT-3's training data, which just has me thinking like okay. So if these are data sets that are full of what most would look at and constitute as quotes, pirated material, like do these large technology companies that are actually using them, like they're obviously using them? Do they know? Or is it a matter of like, do they care? Or that the legal kind of line there is such gray area that they truly believe they can use and not get in trouble about it?
0:10:10 - Kyle Barr
Yeah, I mean for one. Books One and Books Two, to be exact, are open AIs resources for training their AI Got it.
So Books One can go back to the original GPT One and Two. That was a few gigabytes and back then you could mostly find their data was just like scraping up information from Reddit and Wikipedia, which is still contained in the data. Books One was probably based on Book Corpus, which is another set of data that also has copyrighted material in it. Books Two there's very little public information about it, but a lot of suspected that it also comes from a pirated resource, just open. Ai didn't say what's in it, and that's what is part of the problem. Right now is a lot of AI companies aren't saying what's in their training data. They just refer to it as this is the file that we use, but they don't say what's in the file because they don't want to get sued.
Meta trained theirs on Books Three, which was created as a response to Books One and Two from OpenAI. So more than likely these are AI researchers, these are internet people. They code for living. They know that this stuff comes from and it's not a secret either, like Sean Presser has been very open from the very beginning where it came from. It's a Google search will tell you where it came from. So Meta knows what's in it and they especially know because after they released Lama One, they released Lama Two.
Lama Two doesn't refer to Books Three or to the pile either. It just says we trained AI and this many gigabytes of data. They don't say what it is, but if it's a transition from the first one to the second one, more than likely it contains Books Three, though Meta won't tell you if it does, because they wouldn't just retrain their entire model on entirely new data. And it's consistent with what's Meta has done before. They have their audio generator, their auto generator AI models now and those they say were trained on open source audiobooks, though they didn't say where exactly they came from. And that's consistent with all these companies now open AI GPT-4 model, which has 45 terabytes of training data. They don't tell you a single bit of what's inside. That's all proprietary and until somebody eventually leaks it will only guess what's in it.
0:13:00 - Jason Howell
Yeah, that's so fascinating. That really seems to be kind of at the heart of where all this is heading right now is that it is the secret sauce that powers these systems, and they realize that Otherwise they'd be probably more forthcoming with what's actually in there, and if it's driving their system, they have little interest in revealing the secret sauce that they're working with. Yet you've got people like Sarah Silverman and, I imagine, other lawsuits that are really pressing this issue to determine whether they have rights to the data that is being fed and training these systems, and if at some point that balances out or transitions in a direction that requires them to reveal these things, it's going to be very interesting. I suppose the big question here is whether this does the inclusion of these things and what they're used to generate, constitutes fair use or not. I mean, does it seem like the Silverman case is becoming kind of the canary in the coal mine for how all this stuff is going to play out?
0:14:10 - Kyle Barr
Well, there's a lot of AI cases going through the courts right now. The Sarah Silverman one has got a lot of press because it's a major figure, but there's going to be a lot of these cases that might have contradictory results. One judge might say, yeah, this is fair use. And some law experts have said, yeah, this would constitute fair use because they're not reproducing the content, although language models and AIR generators have proven that they can just simply recreate the original content that was in the training data and that's actually included in the lawsuit that Getty Images brought against Dali and some of the other AIR generators. So there's a lot of questions right now and it could go either way. It could have multiple contradictory decisions in the future, in which case a higher court would have to decide. Now I'm not saying like, oh, we're definitely going to see a Supreme Court decision on this, but it's a very strong possibility unless everybody starts agreeing whether or not this is actually constitutes fair use.
0:15:20 - Jason Howell
Yeah, I have a feeling the former, probably, not the latter, everybody agreeing on anything at this stage. Yeah, we'll see Kyle. Thank you so much for taking time to talk to us about this Super fascinating stuff. Great story. Of course, kyle Barr can be found at gizmodocom. If people want to follow you online, if they want to find you online and follow your work, aside from going to gizmodo, where can they find you on the socials?
0:15:43 - Kyle Barr
Mostly Twitter. I'm not going to call it X. No one even asked me to call it X at Kyle Barr 5. And otherwise you can just find my work online or just search me on Google. I guess Eventually they're just going to like remove me from everything. So we know.
0:16:00 - Jason Howell
Well, thank you, kyle, appreciate your time. We'll talk to you soon. Thank you, take care, All right, more interviews coming up. Good headers carry a Huh mighty payload, apparently, for criminals looking to steal identities. In some cases it's as easy as a chat with a bot. Even this is just super fascinating how this is happening behind the scenes. Joining me to talk about this is Joseph Cox, who wrote about this for the new 404 media company. Welcome, joseph, big time congrats on the new venture. We can talk about a little bit about that more a little bit later. But first of all I just want to say congratulations, because that's a big deal.
0:16:39 - Joseph Cox
Thank you so much. I'm excited to be here doing I think this is like my first video interview as 404 media. It is indeed.
0:16:48 - Jason Howell
Well, you guys just launched.
So I'm super Exactly, yeah, super excited to get you on the week of, because I've just I've loved following you and your colleagues work over advice. We've had you on this show a number of times. That's part of the reason why we bring you on, because we love your work, so it's awesome to see that you're doing this. We'll talk about that in a second. Let's talk about the story. First of all, credit headers. Tell us a little bit about what a credit header is. What kind of data actually resides in there? That's, I guess, very valuable.
0:17:18 - Joseph Cox
Sure. So I guess, first of all, everybody's probably familiar with a credit report, right? You know, this is this pre sensitive document that Xperia and Equifax TransUnion has about how much money you borrowed, how much you owe, that sort of thing. The credit header is basically everything in that document which is, above all of that, financial information. So this can be your name, date of birth address, social security number as well, so it's a lot of personally identifiable information. Now, you know, there's one law that governs the credit report and that's a pretty sensitive one, but there's another law which governs credit header, and that is why the two are treated differently and why, as we've shown in our investigation, some unsavory people who got access to that personal information. Yeah, yeah.
0:18:09 - Jason Howell
So okay. So then the data ends up there through this credit reporting process. How does it actually get there in a way that I guess that people can get access to this? It's crazy that I mean I understand why that information would be there for something like a credit report. You know, this requires a lot of very important sensitive data, but then it just seems like it's accessible and that seems very wrong. How does it get there?
0:18:39 - Joseph Cox
Well, that's the crazy thing in that I think the vast majority of people don't necessarily realize that when they sign up to a credit card, their bank is giving this personal information over. I think people understand that oh yes, they need to know whether I owe somebody money or not, but they don't know that that information is transferred to these bureaus. And many don't know that the bureaus have created this complex supply chain of data where they then sell the credit header or transfer it to other companies. So below the credit bureaus, you may have data brokers who gather data from various places and then sell it onwards. Underneath them, you'll have a lot of companies and industries actually using this credit header data.
One we mentioned is like a company that helps universities contact their alumni. So you know, oh, you went to Harvard or wherever. I'm just making that up. We want to have money from you and one of my co-workers, emmanuel, he brought up the story of you know he moves house and then somehow his university is sent to military. Like how the hell do they find my new address so quickly? It's this is the credit header data. Beyond the alumni things as private investigators, there's law enforcement as well all of these industries. But crucially in this article is that there's another layer below that which people aren't very familiar with, and that's how criminals are now tapping into that supply chain data as well. There's like a very tall pyramid credit bureaus, liquor brokers, industries, criminals.
0:20:13 - Jason Howell
Yeah, yeah, indeed. So then how did you gain access to this data? Because for your, for the report that you wrote up, you actually were able to get access some of this data and it sounded like kind of the bot ecosystem was. Was that play in order to pull it off?
0:20:30 - Joseph Cox
Yeah, we always want to show sort of actually the consequences of this sort of thing and sort of what's possible. You know I've bought location data before when we found that was being sold to bounty hunters here, as you will do to. I log on to telegram. I find this bot that's used by criminals. I pay $15 in Bitcoin and I just enter the name of a target and the state I think they live in. This is with their consent. I press enter and a few seconds later it spits out a file and that contains all of that header information I was talking about and especially every single physical address this person had lived in since they were basically a child. You can follow them through across decades. You know, and we confirm the data is accurate. And you know you don't have to be very technically sophisticated to type a name and a state into telegram and press enter so you can imagine just sorts of people who'd be able to do this. You know you don't have to be some hacker. You could just be an opportunistic criminal.
0:21:36 - Jason Howell
Indeed, okay. So what kind of criminals are actually tapping into this data and what are they doing with it? Because I mean the image, the hero image of this article, you know, shows like ammunition and I mean this isn't just, like you know, light stuff. The consequences here are pretty, pretty heavy. So what kind of criminals are out there using the stuff?
0:21:57 - Joseph Cox
Yeah, exactly Over the past few years, especially the past couple of years, there's been a dramatic rise in physical violence inside the cybercrime underground, especially like the cybercrime underground of young people. So it may be people who sim swap to steal people's cryptocurrency, it may be people who hack into Gmail accounts, that sort of thing, but along that, there's also this now violence as a service. And if you want to throw a brick through somebody's house, someone on Telegram will go do that for you. If you want somebody to be beaten up, somebody will do that. If you want someone shot and we've seen evidence of that or kidnapped, somebody on Telegram will do that for you. And in these communities, in these Telegram groups, is exactly where these bots are being advertised. It's these sorts of people that are using this incredibly trivially to use data supply chain.
0:22:59 - Jason Howell
Wow, yeah, that's crazy stuff and I have continually a lot of respect for you and your colleagues because I wouldn't feel comfortable going into the Telegram where this stuff is just being traded. And yet you guys do that in order to research this stuff. So a lot of respect going away. Your article makes it sound like this data, which is very consequential stuff, is pretty easily made available to other companies. As more and more we kind of learn this about the data the data markets out there is that so much of our data is traded so freely. Same here. How is that access controlled? What kind of control mechanisms are in place to protect that data, even though it looks like it's not doing a whole lot of good?
0:23:49 - Joseph Cox
Yeah, so when you're a company and you gain access to credit-headed data, you basically have to have a permissible use. I think that's the term used in the relevant legislation, and it's similar to how DMV records can be accessed by private investigators, that sort of thing, and I'll use PIs as the main example. But they go to a website and they say, yes, I have a legitimate use to access this data and that's basically it.
It's a pinky promise that you have a real reason to use it and you're going to use it properly, which is, of course, one of the ways we found at least one criminal gained access was simply by posing as a PI. As a former law enforcement officer, they said, yeah, that's me, and they provided fake credentials and they got access to some data. So it is a very, very weakly enforced sorry pinky promise. Now there may be changes around this. The Consumer Financial Protection Bureau almost in parallel to us, saying it's going to crack down on the sale or the distribution of credit header data, but when I contacted them saying what do you think about violent criminals getting access to this by posing as PIs, they gave me a statement about stopping AI getting access to this data, or something which is not to berate the Bureau too much, but it is to say that I think this is an overlooked issue, even by the very agency that's tasked with regulating it.
0:25:15 - Jason Howell
Yeah, interesting stuff. Okay, do people have any sort of control or ability to put some protections in a place so that this doesn't happen to them? I mean, it kind of sounds like no. It kind of sounds like, if you're filling out credit reports, kiss that information goodbye and just allow it to go where it goes, which is unfortunate.
0:25:40 - Joseph Cox
Yeah, it is, I'm sorry to say, exceptionally difficult. I mean, you could do one approach, sort of tackling at the bottom of that pyramid, and you could try to go to every single company that you think may have your credit header. You are never going to find all of those companies. Some of them won't respond to you, some of them will just tell you to get lost as well, and it's like an unknown unknown. You don't know how many companies to do. So you try to flip it around, maybe.
I don't want to give data to the credit bureaus in the first place. Obviously, that's very difficult because the vast majority of American adults use a credit card. What I would say is that, even though it may not be guaranteed to work because if the data has already been extracted from the credit bureaus, you could try to put a credit freeze onto your reports, and that stops third parties companies being able to pull the report from that moment onwards, which is not a complete guarantee that your data won't be out there floating somewhere, but it will stop collection from that point onwards at least. Hopefully, yeah, hopefully.
0:26:42 - Jason Howell
Cross your fingers. Well, interesting stuff. I know you guys are going to be writing about all sorts of stories like this, because it is very reminiscent of the stories that I was following when you were at Vice. So tell us a little bit about 404 Media. I'm super excited for you guys. I think you've got a slam dunk with the approach and perspective that you bring to this new business. So tell us a little bit about it.
0:27:09 - Joseph Cox
Yeah, totally so. As some people may know, vice had financial issues and a bankruptcy. I won't tell that whole story, but basically, when it was revealed, our executives were also making salaries of hundreds upon hundreds of thousands of dollars and receiving bonuses at the same time as laying off well over 100 of our coworkers. That was the end for me and I decided to quit, and we wanted to do a subscriber driven media outlet. That is a very simple concept. At the end of the day, you pay us to do journalism, we will do good journalism. It really, really is as simple as that, and we want people to subscribe annually or monthly. You'll get access to podcasts, to events, to articles. Everything is free to access for at least this week and then maybe next week or maybe sometime after that, we will experiment with a paywall.
We want to provide value to our subscribers but, crucially, any articles gathered by Freedom of Information Act requests which we do constantly, like one today about the DHS those will always be free because we want to publish those in the public interest, but they cost money. The FBI asks for hundreds or thousands of dollars to release these documents. So if that sounds appealing to anybody, please go to 404mediaco. Sign up monthly or yearly, and I can't stress enough that your money directly goes to supporting us. So first it runs the website. Second of all, it gives us a living wage so we don't die and we can continue to do the journalism, and then eventually, we would of course like to grow in a sustainable, responsible way.
0:28:56 - Jason Howell
Excellent. Well, I applaud what you're doing and I'm super excited to follow along and see kind of where you go from here. It's a really exciting moment for you guys and, yeah, a lot of respect for your team. So thank you for hopping on and telling us a little bit about this, Sharon. What's going on at 404media? Joseph Cox, of course 404mediaco. And if people want to follow you online, where can they find you, Joseph?
0:29:25 - Joseph Cox
404mediaco is the best place, but if you want to get me on Twitter, it's Joseph F Cox, and then 404media is at 404co on.
0:29:34 - Jason Howell
Twitter as well. Excellent. I wonder if, like for the coming months, anytime someone brings up their Twitter presence, it's still going to be Twitter and X will never be mentioned. It seems that most people who have been there a very long time just refuse to say X, myself included. I can't bring myself to do it, joseph, thank you again. Really a pleasure talking with you and we will see you very soon, take care. All right, so I am going to share a review with you all.
I got a device in my hands about a week and a half ago from the folks at AT&T. It is on the AT&T network. I have the Samsung Galaxy Z Fold 5. And you know it's a I've got to say this is a pretty solid, solid feeling device. So I thought I'd give a share of my thoughts. I got to kind of wipe off the screen a little bit because it's a little fingerprinting, but thought I'd give you my review of the Z Fold 5. And you can kind of, you know, figure out for yourself whether you want to drop $1,800 or more on this device, because it is a pricey piece of hardware here. So you can see here and, like I said, this is on the AT&T network, just FYI, I did get this as a review unit from AT&T, so I'm going to be sending this back to them here in a few weeks, but while I have had it I've had a lot of fun with it, because it really is. You know, it feels futuristic in the sense that it is a device that folds and everything.
They've made some changes to the design. I think the notable change because there aren't an insane amount of changes between this device and the Z Fold 4. But I think the most notable change that you see if you're using your eyes to see what's going on, is that they changed the hinge system. If you'd looked at any of the previous Z Folds, this would be a teardrop shape and basically what that means is that over here on the edge, counter to the hinge, it would be touching the way it is. But over here you'd actually be able to see a little light through, kind of like open up slightly. And I don't know if that was always kind of like a challenge with the hinge that they weren't able to figure out, or I think, you know, sometimes they've spun it as well. You know that actually allows us to prevent like dust from, you know, to give some sort of passageway for dust, because if it snaps all the way closed then like dust and maybe even like a pebble or some tiny little you know something, can get in there and damage the screen. Apparently they've gotten over that because they've made the change and it is now like this. Other foldables have done that too, so it so when essentially what that means is that lays flat completely when it is closed. I mean it doesn't actually lay flat if you've got it on that side because you got the camera bumps that prevent it from doing that, but it does mean that the device just feels a little bit more polished, I'd say, as a result. But I mean the device itself is, you know, a really nice feeling.
Device is a little heavy, which is still the case. It kind of always has been still thicker than a normal phone. When you're talking about kind of thickness, you know this is my Pixel 7 Pro and I mean there we go. It is definitely thicker, even with a case on the 7 Pro it's thicker than what the case is. But they have thinned it out slightly. They've trimmed a mirror two millimeters off when it's folded the way it is right now. So that is something I suppose Still pretty heavy, like I said, kind of like a brick. You know, if you're carrying around this around in your pocket, you definitely feel it more than you do some other smartphones, you know, like the thinner smartphones or maybe even just a little bit smaller. It does lack dust resistance. It's resistant to water immersion but still not resistant to dust. So there's that. But I think all through, you know, altogether it really is very similar to last year's model and you know some minor improvements to the design, like the fact that it can fold entirely flat now, and I mean they've never had a problem with this folding or unfolding flat as well. Like some of the other phones, I think the Pixel Fold has a hard time folding, unfolding entirely flat. This one doesn't have that problem.
But I would say that at this stage you know we're where are we about five years into the Z Fold's existence, if I got my math right and it's a really great recipe. Samsung, you know, continues to show, I think, almost flex, the fact that their foldable game is about as polished and excels anybody else in the industry right now, purely for the fact that they've been doing it about as long as anybody in the industry and they've got the resources to really throw at it and make it, you know, a really impressive device and to do all make all the changes that people are asking for. At the same time I kind of feel like, okay, we've seen this, where do you go from here? And you know, between the z-fold 4 and the z-fold 5, this kind of marginal changes, I almost feel like you know, five years in Samsung's, at the point to where they kind of you know they got to kind of capture the imagination again. We're kind of getting used to their like. To speak for myself, I'm kind of getting used to what we're seeing here. How can it read? How can Samsung reignite the design to give it a little bit extra boost, or does it continue to just Iterate the way it is? I kind of think it's probably gonna be the latter versus the former.
But on the outside the already very smudged outside, depending on the angle you've got this display. The outer display, as you can see, is very, very narrow, and some of the other devices like when I was at Google I owe I got to Get my hands on the pixel fold and it has kind of like a wider, almost booklike Um dimension to it. This one really ops for the tall, slender, candy bar phone. When it's folded, the pixel fold kind of comes out a little bit more. So it kind of feels like when you open it up it, you know, things are a little bit wider. It feels more like a book, and what that also gives it is that the front of the device feels more like a regular phone. Right here You've got this really kind of narrow view. You got to find the fingerprint sensor. Sometimes I have a hard time with a fingerprint sensor on the side, as you can see. It's right there, but it has a really narrow view.
So if you're doing things, you know, like typing the touch points on this can be really hard to hit. But you know, is that the end of the world? Not really. When you've got a foldable because you can open it up, boom, your touch points get a lot larger. It, as you can see, the keyboard kind of splits out so that each finger or each thumb on the opposing sides of the displays, Um, have their own zone for the keyboard to type on to. And I think you can actually fold that together if you don't like that. But but anyways, you have the ability to kind of make things bigger and not just bigger but a lot bigger. Right, this? This interior display, 7.6 inch display looks great.
Yes, you're going to get those fold, uh, those fold marks as you see here. I mean, I think every review you ever see of a z fold or a z flip is going to point out these fold marks. The reality is, when you're using these phones for any extended period of time my experience has been you really start to tune out that fold, like it like In a review. When you look at this it looks incredibly imperfect, but in daily use you just start to not care. Um, because you know again, because of what you get out of it, this larger landscape view, um, the ability to like, collapse things down and, uh, here, let's see here. So I've got, you know, the, the camera app and all my controls are down here and if I take some pictures, I've got a little previews down here that I can go ahead and throw back up on the screen and get my you know other other images down on the carousel down there. You get the ability to do those types of things because of the way that this folds. Um, and so If that crease down there, if you're willing to get over the crease, you get a lot of trade off, for that is, um Is kind of my thought there. Um, so not a big deal, and I've got a whole lot more to talk about with this phone before we go any further. Let's take a quick break.
Performance wise. You've got the snapdragon 8 gen 2 Uh, 12 gigs of ram, uh, peak performance. I don't feel like I need to go into that very much. I mean, this is samsung, you know, really going for it when it comes to performance in their flagships. Samsung always does good with their flagship performance, in my opinion, especially when they're leaning on the snapdragon, the latest snapdragon. Um, so you know, you get your fast gaming. You get your fast Maneuvering through the ui, which I feel like is kind of table stakes at this point. You should definitely have that on your smartphone. Um, you get fast uh photo performance. You know lots of Taking pictures in challenging environments. I was able to kind of do that on a quick Basis without waiting or things kind of lagging and trying to catch up that sort of stuff.
Um, software wise, as you can see Uh down here, you get this little task bar, which is a really nice Bonus. In previous years. You haven't had that. Maybe you had that on the z fold for Uh, because I know, you know, samsung and google have kind of been working Alongside each other to really well. Well, google has been working on android to make it more Tablet friendly, and then samsung benefits from that with the fold series, right, because this really is like a, a tablet size display. So you get, uh, you get this, this task bar down below, I can launch a, a chrome instance.
Um, if I go To my, you know my multitasking view, I can tap on that, open it in split screen view. Now I've got google search, or I've got chrome on one side, I've got my app drawer on the other, or, excuse me. Then I could maybe launch, you know, the play store and Do a search, or or whatever. I can go through it. Essentially, this allows you to really kind of customize what you have. I could even let's see here, I could even drag my camera up to this bottom portion and, boop, I've now got three different kind of environments Running fully resizable. Um, I could drag Information from one screen to the other, if I have, you know, if I'm writing an email and on the other side is a browser, I can actually tap and drag Um content into that email. That's really nice, including images, by the way. So it really is kind of becoming this like very manipulative, um drag and drop environment, which is really sweet. And then you can also, um, save App pairs, so I could star this and add the app pair to my home screen, and that allows me Eh, where is it going to go? It goes right there. So if I like that particular app pair, I can save that and boop, it goes right back to it. So, depending on your workflow, for a very particular thing, you can set that up, have it stored and then always kind of Pull from it, which is just really nice.
I think that's really the, the big story for the fold and and the why is it's ability to, yes, go from a smartphone you know size device into the tablet layout. But what are the benefits that you get from doing that? You get this, this like broadened ability to see things larger, to Manipulate and and work between apps in a way that is maybe more difficult when you're just dealing, you know, with a single screen on a candy bar phone. Uh, it's not as, not as enjoyable, let's just say, but it really is on the z fold 5. And then, camera wise, um, I actually it's probably better to show some of the images that I that I linked to and, uh, you know, for a trigger warning, you're gonna see a lot of my new puppy, bronson, uh, because I just love him to death and he's very, uh, very photo friendly.
But, um, I was really happy with it, with the pictures that I got of this, and not surprised at all at that Samsung, at this point, is really dialed in their camera hardware, um, especially for this price point $1,800 you better have that's with a wide angle taken really close. That's why it's stretched out. But, um, a lot of clarity in sunlight, um, that one's maybe a little washed out. I don't know if it was the, the lighting or if I was kind of moving when I snapped the shot. That's, uh, some amazing Mediterranean food that we had, uh, over the weekend and, uh, you know, turned into a great, uh, great photo.
A couple of photos from a concert that we went to at the um, um in in Berkeley and kind of zoomed in too. So this has a 3x optical zoom. So, you know, I was able to get some reasonable shots. I would love to have more reach than a 3x, but Nonetheless, this was darkness and, you know, lit up on the stage and I think if you go to the next one, you'll see this was, uh, my morning jacket and actually if you, if I don't know, if you go back to that one and kind of zoom in A little bit, if it'll allow you to zoom in, um, you start to see a little bit more of the detail that it's actually able to get from pretty far away. I mean the, the singer on the right side you can see. You know, we were very far away and you could still get some really nice clarity and some reach With the camera on this device.
So Night shots look excellent. Um, you know, when I make a weird face, uh, my nose looks weird, that's me, that's not the camera, don't worry. Um, and uh, I think go again. This was taken this morning actually, at like, I mean, the sun was not even up, like this makes it look like it was really bright out, but it was almost darkness, and so I was trying to work with the night mode and I thought it did a pretty darn good job. The light that's shining in on the deck is coming from our kitchen window, but you know, that's that's a little ways away, and then all the light in the sky, that's just the sun that's threatening to come into the sky. I was pretty impressed with that as a night shot. So overall, I think the camera system on here is good and it should be good that that's about the same time. It's just not in the direct sun, but yeah, so I was pretty happy with the camera performance.
On the Z Fold 5. I do wish that I had the pixel fold to like do a direct comparison with. In a perfect world, that's what this would have been. Here's the pixel fold, here's the Z Fold 5. How do they compare directly? Because I'm super curious and I do know that I'm going to be getting a pixel fold for review sometime soon. Google tells me I don't know exactly when that's going to be, but but again, just kind of going back to the Z Fold 5, it feels like Samsung continues to refine the, the formula, and I don't think that they're wrong to do so. I think that it's really working for them and this is, you know, a pretty impressive device as a result of it.
However, it's $1,800. It's a very costly device. So you know it definitely falls into that category of like. You got to ask yourself, like do is an $1,800 device worth it for what I get here? And I mean there are a lot of people who are very happy with the Z Fold 5. I'm not sure if I was spending my money on a new phone. I'm not sure that I would spend $1,800 on a Z Fold 5, not because it's not a good phone, but just because, like, that's just a lot of money for me to drop on a smartphone. But I know people out there do value it and I do appreciate and like what Samsung is doing here. I just kind of think they need to. They need to on the next version of this device. They need to kind of show what's new. And you know I get why they're iterating. They really want to be seen as the de facto standard when it comes to foldables. I think that's working for them and I think they're there. Now, what else can you do? What is the next secret sauce that you can inject into this to keep, keep the excitement going? So, so $1,800. That's it. If you want to get one of these, just $1,800. So there you go. That's my review of the Samsung Z Fold 5. Big thanks to AT&T for sending me this smartphone for review. All right, we've reached the end of this episode.
Tech news weekly. We do this show every Thursday. Usually I do it with Mikah Sargent sitting next to me. Thankfully, he will be back next week so we can toss back and forth between a couple of interviews, couple stories of the week. That's what we do around here.
twit.tv/tnw to subscribe, which is super important. Please go there and subscribe so you can get our audio and video formats whichever way you prefer, without having to hunt for them. If you want to get all of our shows ad free, we've got something for you to check out. It is called Club Twit and it is $7 per month. That gets you every show with no ads. That gets you exclusive twit plus content. It's a feed that you only get if you're inside of the club.
We do tons of shows. Hands on Mac: Hands on Mac: Hands on Windows: I just did a fireside chat with Dan Patterson, one of my good friends in the podcast and news industry, a couple of nights ago and I was really, really happy with that chat. You can find that in the club we're doing all sorts of really cool stuff. Ant Pruit, of course, is managing a lot of it. There's a photo walk, stacey's book club, an untitled Linux show, untitled AI show, which I promise is going to have a title very soon because I'm making that a priority at this point. All that for $7 a month. You also get access to our Discord, which is where you watch a lot of this content, the extra content.
twit.tv/clubtwit. If you want to check that out, join and support me in my pursuit of what this AI show is going to be. It's actually turning into a lot of fun playing around with ideas and concepts. I'm going to be doing that again today. If you're a member of the club, you will see it. twit.tv/clubtwit. You can find me at Jason Howell on Twitter, not on X, unless you're a maniac and you want to call it X. I'm mastered on Twitter.com. Just look for my name and a social network and you will probably find me. Big thanks to Ant, to John, to John, to Burke behind the scenes, to everybody here who help us do this show each and every week, and thanks to you for watching and listening. We'll see you next time on Tech News Weekly. Bye, everybody.
0:48:21 - Scott Wilkinson
Hey there, scott Wilkinson here. In case you hadn't heard, Home Theater Geeks is back. Each week I bring you the latest audio video news, tips and tricks to get the most out of your AV system, product reviews and more. You can enjoy Home Theater Geeks only if you're a member of Club Twit, which costs seven bucks a month, or you can subscribe to Home Theater Geeks by itself for only $2.99 a month. I hope you'll join me for a weekly dose of Home Theater Geekatude.
