Siri Gets Smart, While Grok Stays Itself - DTNS 5190

0:28This, oh, this old thing? Bloomberg's Mark Gurman reports that his sources indicate Apple will upgrade Siri this year to make it more conversational, make it kind of feature equivalent with chatbots. So Claude, ChatGPT, Gemini. Everybody is reporting this as Siri will become a chatbot. Well, Siri's kind of already a chatbot. It's just not very good at it yet, so it's going to get a lot better at it. It will probably be most similar to Gemini since Google's Gemini models will power much of it. The new version of Siri is being tested under the codename Campos, which made me immediately want to reach out to Dan Campos, host of Noticias de Tecnologie Express, and find out how he feels about that.

1:11Gurman mentioned features like summarized search responses, something that's been reported as world knowledge answers. Also, the ability to create content right from Siri, generate images right from Siri, adjust settings, analyze uploaded files. but you would do it within an app that you're using. It's going to be integrated into core apps like Mail, Music, Podcasts, TV, even Xcode and Photos. The company has tested a standalone Siri app, much like a ChatGPT app or a Gemini app, but German sources say they don't intend to release that to the public. Craig Federighi, who is running the show now, still thinks it's better to just put Siri into the thing you're using than to make it its own thing.

1:54You will see this update come with iOS 27, iPadOS 27, codename Rave, and macOS 27, codename Fizz, which we are expecting, as usual, to hear about at WWDC this summer. Meanwhile, you are going to get an update to Siri before that. If you've been wondering where all those great features mentioned in 2024 have gone, they are coming sooner. Now, these are going to be powered by a Gemini model, but they're less chatbot-like. So it's analyzing on-screen content. If you're looking at something, you can ask Siri about what you're looking at. The ability to let it use your personal data on device to make better responses.

2:32Those are expected to arrive in iOS 26.4 in the next few months. And one last interesting thing to note about this, Apple is designing Compos, that's the chatbot-like model, that the underlying model can be switched out. So there's an Apple framework around it, and then it just calls into Gemini to do the thinking. That gives Apple the flexibility to switch away from Google Gemini at some point if it so chooses or potentially give you the choice to decide which model you want to power it. The information also attributes the focus in direction on Siri to Craig Federighi taking over direct control of this project.

3:10Now, as always, I must start with the disclosure. I am that person and hopefully not the only one who does not use Siri from the time I got an iPhone 2 now. So you turned it off. I turned it all the way off. It doesn't even get to look at my apps. Yeah, well, because it's a use case thing. It's like, do I need this? No. Do I want to deal with it? No. And funny enough, I have to blame Nintendo for a lot of this because back in the whole brain age discussion back on the DS and from other Canadians. Again, this is back when Siri launched. So I am working on old prejudice here. It did not like Canadian accents.

3:52It was more of a fight. That's not a problem anymore. A lot of people complained that it wasn't very useful. And you're like, it was even less useful because I couldn't understand what I was saying. Yeah. It's like, I don't need to fight with my phone. I fight with enough tech. Yeah. Yeah. So it's improved on that. but you haven't yet seen anything that would make you turn it on. Does this seem like it might change that? I've never felt the urge to create something using this kind of thing. I've got friends who are into vibe coding now and people who've never done anything like this. Like if they can just talk to Siri to make something that they need, they will absolutely be all over this and be thrilled and then feel justified in the very expensive iPhone that they have.

4:36Yeah. What about the settings stuff? That seems like a more practical thing that might be more appealing if you can just tell it to do something and you don't have to dig into the settings app to do it. That would have saved me a lot of time. I had finally upgraded my iPhone and was very angry at the look of it. There's no reason for that UX change to have made me as angry as it did. But if I could have told Siri like, hey, listen, I want dark mode on everything but this. I want to get rid of all of these rounded corners and I want to reduce the the the motion and have it do all that for me instead of me having to look up that I did this all correctly in the right places for it to affect all of the apps.

5:16I might have been tempted to turn that on. Well, I am dubbing this the cutter threshold. if and when Compost finally launches, if it can convince you that, you know what, maybe I do want to turn that on, that it will have reached the cutter threshold. And I think that is a measure of success for Siri. Yeah, well, we'll have to see because now I've already got my phone how I like it again. Will it be able to convince you? A couple other things here too. The information reporting that Apple is developing a wearable pin described as a thin, flat circular disc with an aluminum and glass shell. very similar to the way the AirTag looks, frankly.

5:53It would have two cameras, three microphones, a speaker, and a button, which is a lot to put in an AirTag-sized device. Charging strip on the back. This would not be a single-use battery situation. And it sounds very similar to product designs rumored to be under development by OpenAI and former Apple designer Johnny Ive over there. And then the information's Wayne Ma also reporting that Apple will release a Home Hub smart display this spring. So we've been hearing about the one with a swiveling robotic base. Mark Gurman has been talking about that as well. And Ma seems to think like that one's about ready to hit the shelves.

6:27So that would be interesting to see as well. Yeah. Is that something you would use? Like AirTags, I understood the use case for that immediately. The glasses, I also pretty much understood. This, is this something you want in your life? This is something I want to see if I can find a use for. Humane made that AI pin and I signed up. and then it turned out it was awful and I canceled the order before they shipped it. I got my money back. Oh, you got out in time. You know, but all of the attempts at these have been falling short in functionality. If Apple can make it so it works, so I'm like able to just talk to it, you know, at the press of a button and the button, it very Star Trek like, you know, tap it and ask things and it actually responds.

7:08then I will be interested to see what needs that I didn't know needed to be filled. Could this fill? Because I feel like I felt the same way about the Amazon Echo. Everybody was like, why would you want that? And then I was like, I don't know, but I get the sense there might be something I'd use it for. And I found out it's great for just turning your smart home into a voice activated smart home. Right. And so I still use an Amazon Echo for exactly that and occasionally asking at the weather and getting flash briefings and stuff like that. I wonder if this will be the same way. Yeah. Cause like I fully understood for people who can wear the, um, the earbuds and the incredible, uh, advances in what you can do with the AirPods and stuff.

7:49So I'm thinking like, what is this going to do that you can't already do in this, but I'm guessing the camera stuff is going to have hopefully some unique uses. Being able to see around you would be a big differentiator. You're right. And, and just not having to having some people don't want to wear earbuds, right? Yeah, like me. Having it do that instead might be part of it. But it's going to be a, it seems very trite to say, it's going to be a big year for Apple. But these are all things that we've been hearing about for a long time that look like we're actually going to ship this year. CTNS is made possible by you, the listener.

8:22Thanks to Brandon Brooks, Johnny Hernandez, and high-tech Oki.

8:29All right, let's get right into the briefs. We have two adventurous ways to use Windows for you today, and I'm pretty excited to talk about this. The Nex phone from Nex Computer, that's N-E-X, is a$549, 6.58-inch phone with an eight-core Qualcomm chip, 12 gigs of RAM and 256 gigs of storage set to ship in Q3. And we are telling you this because it can triple boot into Android, Linux, and Windows 11. And it works partly because it runs Qualcomm's DragonWing SoC optimized for IoT. The idea is to plug it into the included USB-C dock to use with an external screen, keyboard, and mouse when using the desktop operating systems.

9:15Nex also sells the NextDoc 360, which gives you a laptop frame for the phone. You can reserve one now for$199 and pay the rest later. And if you're a little more old school like me, Windows hacker Bob Pony has prepared images for Windows 7 and Windows Vista that include all the security patches through January 2026, as Vista just issued its last security update through Premium Assurance. Yeah, this is great. I love that Bob Pony did this. And if you are out there asking, why would you need that? You don't. You're not the person this is for. This is for the person who is going, that sounds cool.

9:52I want that ISO. I want to play around with Windows 7 again and make sure I have all the security patches. is or there are some actual practical uses in certain enterprises where people are still running on Vista and Windows 7 for legitimate reasons. This could make things easier for them as well. As for that triple-boot phone, I mean, I guess, yes, you could do this with any phone technically if you have the skills, but having it by design out of the box able to do this is a big advantage. 549 bucks, not that bad. I think that Dragon Wing sock is a little underpowered for a lot of uses. You're not going to be doing more than, you know, maybe a little word processing and email checking.

10:31But if you're looking for a Chromebook style thing, you could just carry around a phone and then when you need it, plug it into a keyboard mouse or maybe carry around that laptop frame. Although that seems to defeat the purpose to carry that around as well. But yeah, I love this idea of having my Android phone that could also be a Chromebook style Windows machine or Linux machine when I need it. Pretty cool. Yeah, I was shocked at the price. I was like, that is almost reasonable. considering costs for some low-end laptops these days. And I agree on the enterprise thing because that was my first thought.

11:03I was like, oh, for some financial software that only runs on these older versions. Yeah, all for the Windows 7 and Windows Vista stuff. Yeah, absolutely. I wouldn't want to call this future-proof for them, but it would definitely make a difference. Yeah, and tying these two together, long-term security patches is something that you get with the Qualcomm Dragonwing system on a chip. I think those updates go through like 2030 or 2032 even. Wall Street Journal reports that consulting firm Section surveyed 5 ,000 workers and found a difference between executive and non-management employees over how much time they perceived large language models save them.

11:41This explains everything is what many of you are saying right now. Around two-thirds of non-management workers said that these kinds of tools save them anywhere from no time to two hours a week, whereas 40 % of executives said the tools save them more than eight hours a week. And there's even more executives that were saying they'd save them up to eight hours a week. Most people use LLM for search-style tasks or generating drafts. Fewer use it for complex tasks like data analysis or generating code. So the Wall Street Journal article is saying executives are pushing LLMs because they think they're saving them a lot of time, but they're not saving their workers' time.

12:20Another study presented by PricewaterhouseCoopers at the World Economic Forum in Davos says that 12 % of 4 ,500 chief executive surveys said new LLM tools gave cost and revenue benefits. Just 12 % said, yeah, we're seeing the actual benefits. So 40 % think they saved them eight hours and could save them money, but only 12 % are seeing them save money. And more than half said they have seen no significant financial benefit from using them yet. And yet, the Institute for Student Employers in the UK reports that entry-level recruiting fell 8 % on the last academic year as employers are trying to save money by using generative tools.

12:59New York Federal Reserve data in the U.S. for last August shows that unemployment among recent college grads age 22 to 27 is 5.8 percent compared to 4.1 percent for all workers. A lot of that has to do with the economy and uncertainty and other things. But there are a lot of companies hoping they can save money because the executive can save eight hours a week. And the workers are saying we're still trying to figure out what this actually saves us time. I'm not trying to argue that it doesn't save time by presenting these numbers, but it's clearly a disconnect between management and the workforce.

13:32And a sentence I don't usually say, in fairness to executives, they get, yeah, I don't know, 8 ,000 % more emails than I do in a day. And I do understand why, because like not all of that is necessarily going to need their attention and them being able to sort that is a benefit. Yeah. I don't know what else they do with their hours other than manage that stuff, but apparently they do need time to do other work. So this makes a difference. It probably is really good at speeding up executives and making them more productive. And I think it will get better at lots of things. And I think people will start to learn what it's really good at and save them more than two hours a week.

14:09But we're not there yet. And companies pushing people to use it without being able to tell them what they should use it for is not going to work. And that's what I think we're seeing. And like for my bias, a lot of my friends are coders and they're being forced to use it. Yeah. And they have found that their best use case is using it to start something. And then they go back and fix it because like they've spent so many hours on prompts where they realize like I could have done this faster. But getting it to start something that they then build off of, they find like, all right, it's not the worst.

14:39Yeah. And that's the thing. It is helpful at some stuff. It's giving it a bad name to force people to use it for things that's not good at it. Yeah, it kind of builds resentment when you get forced to do anything. And a couple of important notes for password manager users. LastPass issued a warning that malicious actors are sending phishing emails disguised as LastPass maintenance notifications, saying that users need to create an encrypted backup of LastPass vaults in an attempt to steal master passwords. LastPass is not, to clarify, is not asking customers to backup their vaults. If you get the email and click on the Create Backup Now button embedded in it, you'll be redirected to a phishing site.

15:22Meanwhile, 1Password introduced a tool to protect users of its password manager that alerts you before you manually paste in a password on a site that does not match the site the password is meant for. It's an extra pause to stop you from getting phished. You can activate the new protection in the 1Password browser extension settings under the Notification section. So, yeah, the first one is not a thing LastPass did. It's LastPass saying, hey, watch out. Somebody's trying to trick you. Don't get tricked. And the second one is just preventative maintenance saying we will stop you from auto completing if the site doesn't match.

15:57But if you do it yourself, there wasn't anything we could do until now. So I think that's kind of cool. Yeah. And then you get sites like Sony where whenever I'm trying to log into PlayStation, LastPass always warns me. It's like, hey, like, but this isn't the right thing. And it's like, it is the right thing. Sony just keeps changing what site you go through. Yeah, you either need to teach your password manager that there's more than one version or you can't ignore it, but you probably don't want to get in the habit of ignoring it too often. But pauses are good. So always double check things.

16:26Exactly. Well, if you have feedback about anything that gets brought up on the show, there's lots of ways to get in touch with us, including social networks. We are at DTNS show at DTNS show on X, Instagram threads, blue sky and Mastodon. If you're looking for us on TikTok or YouTube, you can find us at Daily Tech News Show.

16:46Finally, some quick headlines that are just good to know. And, uh, you know, knowing this stuff might make it look a little smarter in the future. Ubisoft announced Thursday that it plans to close studios in Halifax, Nova Scotia and Stockholm and end development on six games, including Prince of Persia Sands of Time Remake. Jeff Bezos, but not Amazon's Blue Origin, announced Wednesday that it will begin launching the first of 5 ,408 internet satellites in Q4 2027. So we've got a while. Samsung shared updates to its good lock feature this week, and sharp-eyed viewers noticed a reference to privacy display in one screenshot.

17:27a rumored feature that would use hardware to make your normal screen act like it has a privacy protection sheet over it. Yeah, those are the things where you can't see if you're looking from the side. That's a cool feature. Looks like it might actually be coming. South Korea beat Europe to the punch on being the first government to enact a comprehensive set of laws governing AI, because they're now in place, whereas the EU's AI Act is taking effect in phases through 2027. Ars Technica reports on an open source plugin for Cloud Code called Humanizer that uses Wikipedia's list of 24 language and formatting patterns that indicate generated text to stop a model from writing like AI.

18:08Yeah, they were like, it was so nice of Wikipedia to identify all the quirks. And now we can ask our model not to do them. A Lenovo product manager posted on Weibo that the 2026 ThinkBook 14 Plus and ThinkBook 16 Plus will support Lenovo's modular Magic Bay system. And that Pogo pin powered Magic Bay system itself will support third party accessories this year. Trying to get more people to make something for that ecosystem. The Wall Street Journal has a good reason why deep learning model developers are relying more and more on old versions of Pokemon, like Pokemon Blue, which I maintain is the best version, to evaluate how well their models reason.

18:50According to the article, it can track a model's reasoning, decision making and progress toward a goal over a long period, which is closer analogy to the types of tasks users are aiming AI at today. That is a good read. Also a good read, Bloomberg Businessweek has an excellent rundown on the actual progress being made by Google, Microsoft, and NVIDIA in creating more accurate weather forecasting models. And I know at least one of you, Bart, doesn't like Bloomberg Businessweek. I just want to say it was not written by Jordan Robertson or Michael Reilly, who wrote the never retracted Chinese hardware backstory for Businessweek in 2018.

19:27Your mileage may vary on whether you want to read this, but it is an excellent rundown on the progress being made on weather models. And Behrens has a good review of how the AI landscape has changed in the years since DeepSeek shook it up. Turns out U.S. companies still lead in chips, model quality, and sales. Those are the essentials for today. Let's dive a little deeper. We've been covering the many investigations into Grok AI, and it's used to create non-consensual and possibly illegal images. Andy Beach tells us why Grok is just the first and not last we'll see of this kind of story. Andy, extra special thanks for joining us for this topic.

20:06Hi, Tom. So yeah, nobody really wants to talk about people making CSAM with AI tools. Certainly don't want to talk about non-consensual sexual imagery made with these tools. And we've been covering this on DTNS. So if you haven't heard earlier coverage about the details, I'll refer you to that. But I thought it would be interesting to kind of pull back a little bit and talk about this, not just from what's happening specifically with Grok AI and XAI, but why this is not a problem that goes away if you fix it here. Yeah, absolutely. I think this is the first time we're having this in a massive way, but it's far from the last.

20:50And obviously, even just the Grok story is going to probably linger for quite a while, but it is likely to tee up both an increase in sort of following the existing consumer protection laws that are out there and maybe even causes governments to rethink globally some of the AI laws that they haven't pushed forward yet and start to put some new policies in place. So I think this is the beginning. Yeah. And we've seen efforts in the United States like the Defiance Act start to get a little more momentum again. We've obviously seen investigations in the EU and the UK. What is what, in your opinion, is the best approach to this sort of thing?

21:40Because I think in some corners, especially if you don't use these tools as much, it seems pretty cut and dry. Like you just don't let it do it. But it's not as easy to implement the way these tools work to say, just turn off all the bad stuff. Why is that? Right. Well, again, I think so we've talked in the past about how rapidly AI progresses. And when things move as fast as it is, it's easy for there to be blind spots because you maybe haven't spent as much time with the code. You haven't spent as much time interacting or pressure testing all of the edge cases. And so people will find a flaw and then the Internet is very good at exploiting those flaws to then do things.

22:31So I'm positive. I will say I'm willing to bet Grok.ai certainly never planned for it, but they also didn't put the right protections in place. And because they move so rapidly, as kind of all AI is right now, then it created this gap in what was in production that got exploited and became very public. And I think that is I know what I've heard a lot of the EU rulings and policy pieces are around slowing down, not not to hold back companies, but to give them the breathing room to make sure that what they're doing is both accurate and safe in the long term, because all of the problems don't appear the first time you go look at it.

23:19You have to sit with it. So so I think we're going to we're going to hear more about this, I think, very rapidly. Obviously, the UK is already looking into this particular story, but I think EU will drive new policies. And they've been at the forefront of putting a lot of the new ideas around AI policies out there. But I think in the states, before we even hear it from the federal government, I think a lot of what California is doing and New York is doing to apply the existing consumer laws to the current cases will drive potentially state-based policies that we see, which hopefully will then push up to a federal level at some point.

24:00Now, I think it's also appealing to think, hey, look at Sora. Sora doesn't have this problem. Look at Veo. Veo doesn't have this problem. Why couldn't – if everybody just does what they've been doing, then we don't have this problem. Is that fair? it it's fair but in in a way you know this this is almost grok's marketing coming back to bite it it's sort of uh early on like to make itself it sort of called itself like the ai bad boy basically it's like we we don't have those those silly guardrails that that everybody's complaining about because in the early days there there were certainly some complaints about about those and And part of what they did was loosen up some of those guardrails.

24:50And obviously this is a case where it went too far and now they're going to have to live with consequences around it. Well, and that I think is useful to remember if you take the personalities out of it. Sora, if you don't use it, you don't realize just how impossible it is to do some very benign things because not just – Sometimes even with yourself, I believe. Yeah, exactly. Right. some of it is for intellectual property reasons, not even non-consensual images or anything like we're talking about here, where the guardrails are very tight to make sure nothing gets through that would be objectionable.

25:28And even with those tight guardrails, like you said earlier, people figure out like, yeah, but if I do this, I can trick it into doing the thing it's not supposed to do. So there's always that aspect of it, but it also means that the product is less useful than it could be for legitimate reasons. And I think that's the thing people get hung up with like, well, what are those legitimate reasons? Are they really that legitimate? Are they really that important? And then it becomes a question of like, well, how much do you think these tools are important? But if you think these tools will be useful in the future, you want to get that line as close to everything that's legitimate and useful should be allowed so we can play with it and find out what it is.

26:04And everything that is illegitimate should not be allowed. And where they're setting the line is different, right? Soros set the line a little conservatively, and Grok did not. And it is paying the price for that. Well, so there's two pieces there. I mean, because we'll talk purely about sort of open AI and image and video. And yeah, they were a little more conservative, particularly around trademark pieces and sort of what we think of as just community safety, we'll say. And yeah, just for example, I asked Sora once to make a video of our editor, Hammond Chamberlain, as if he was cake. And it told me that that was a violation of making video about self-harm.

26:49And I'm like, what kind of video did Sora make? I don't want that. I'm glad it didn't show it to me. But it was very aggressively saying, you know, like, we're not even going to allow any hint of that. And I couldn't even imagine what it must have generated that caused that. Whereas on Grok, it probably would have given me something incredibly disturbing and just spit it out. Right. And here's what we also have to remember, though, is that like what is OK is a cultural decision humans make. And it changes year to year, decade to decade. You know, if we had set, if we had AI back in the 1980s, and we would have had a very different set of guidelines for what was okay usage then versus what is okay now.

27:33And what is okay right now will be different than is what is okay in two years. So it is not like they get to set the line once. It will be a continually moving line and it will be us as a society and a culture who partially push back from time to time on this is too far. Now it's not far enough. And so it will change. And what I was going to say earlier, too, was that, you know, it's not like open AI gets out of all of this. Like, yes, they certainly weren't creating images. They did a good job there. They've had similar issues, though, around, you know, people involved in self-harm after talking to chat GPT.

28:13And they were very quick to try and correct that and lock it down and put guardrails so that the system now de-escalates conversations that seem to be going in that direction. And that was them directly having to respond to not just bad press, but bad use cases that they hadn't realized that they had allowed. And in some cases, they've actually been able to go back and pinpoint it to a very specific release that they had. so that it was like a particular model that was put out in April of last year, I believe, that caused some of those problems. So I think we see this both in text and in images.

28:53This is just an example of images with Grok, but it goes back to that larger question of there's sort of a health and safety community element that goes into this. And this is where we need public policy and governance to come in to help weigh in. And then we need companies to be reactive about how they implement those policies and then react to it because there's always going to be somebody who finds the corner case and a way to exploit it. And how you deal with that exploitation when it comes out impacts greatly and reflects on your company and your product. Well, Andy, thank you for walking us through these kinds of issues on the regular.

29:34And of course, if you want to get more of Andy's thoughts, they should go subscribe to Engines of Change. Where do they get it? Enginesofchange.ai is the URL. And I talk a lot about media technology and AI and where the two intersect. Thanks, Andy. Thanks, Tom. I always appreciate Andy being able to step back and give us that 30 ,000 foot view, you know. We end every episode of DTNS with some shared perspectives. And today, Josh relates his current experience with Siri. Ah, yes. So Josh is going to be very excited at the news that Siri will be upgraded. Uh, Josh writes, as I heard Rob talk about the Siri upgrade earlier this week, I wanted to share my exchange with Siri just last evening, which just shows why it needs a change as soon as possible.

30:18This evening, I walked up to my front door, which has a Yale home kit lock with both hands full. I said, Hey Siri, open the door. Siri said, how can I help with the front door? Siri, can you please open the front door? Siri said, would you like me to look up how to open the front door? I'll need access to JetGPT for that. Uh, no, just unlock the front door. Uh, okay. I'll look for locksmiths in the area. Ouch. Oh my gosh. Poor Josh. Josh, I'm, I'm so sorry, but also this is just a textbook example of like why I never bothered with Siri. And Josh, thank you. Thank you. Thank you for sharing with, sharing your pain with us.

30:57I hope it gave you a good laugh as much as it did us. What are you thinking about? Do you have some insight into a story or just a great story about a story? Please share it with us at feedback at dailytechnewsshow.com. Yes, big thanks to Andy Beach and Josh for contributing to today's show. Thank you for being along for Daily Tech News Show. You're the folks who keep us in business. Subscribe to our sub stack or become a patron at patreon.com slash DTNS.

31:30The DTNS family of podcasts. helping each other understand. Diamond Club hopes you have enjoyed this program.