Bans on Smart Glasses Mean Smart Glasses are HERE TO STAY - DTNS 5317

0:00This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade-in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One.

0:30Huyen Tue Dao: Your team just added its 67th AI tool and also your 67th security blind spot. The good news? The Vanta agent works like a GRC engineer in the background, finding every app your team uses, scoring the risk, and drafting fixes for you. Vanta is the platform used by over 16 ,000 fast-moving companies like Ramp, Cursor, and Harvey, who are shaping the future with AI and staying ahead of AI risk. Get started at Vanta.com. Hi, this is Kim and Penn Holderness from the Laugh Lines podcast. This episode is sponsored by Oli, your wellness partner for back to school and beyond. Back to school season is here.

1:05Huyen Tue Dao: Packed backpacks, packed lunches, and packed schedules are officially back. Oli Kids Sleep helps little dreamers wind down with a gentle blend of melatonin, L-theanine, and botanicals for restful sleep. Translation, an easy way to put those school night negotiations to bed. And for the grownups, Oli's Sleep helps you fall asleep, stay asleep, and wake up feeling refreshed. Because as you know, the morning mayhem comes early. Visit ollie.com and stock up on back-to-school must-haves today. That's O-L-L-Y dot com. These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.

1:48This episode is sponsored by Morning Brew Daily. If you're anything like me, you want to stay informed about what's happening in business and finance, but you also don't have an hour every morning to sift through headlines. Morning Brew Daily is the go-to business news podcast for millions of people. It's a daily business news podcast that's a quick, fun, and easy way to start your day. Whether it's the latest in tech, markets, or the trends shaping the economy, the show breaks it all down in a way that's clear, engaging, and actually enjoyable to listen to. Hosts Neil Freiman and Toby Howell make the show feel like you're catching up with two friends while giving you the context you need to feel smarter and more confident throughout your day.

2:27New episodes drop every weekday wherever you get your podcasts. Tune into Morning Brew Daily wherever you get your podcasts.

2:42Huyen Tue Dao: This is the Daily Tech News for Friday, July 24th, 2026. We tell you what you need to know, give you important context, and help each other understand. And today, the backlash increases against smart glasses with UK Comic-Cons banning them and tech or tech or call for protection of open weight models. Well, not open AI and anthropic, but maybe you would have guessed that. I would have. I'm Tom Merritt. I would now. Let's start with what you need to know with that big story. Yeah, so the backlash against smart glasses isn't getting less. In fact, as usual, most of it is focused on the most popular brand, Meta, and its Ray-Ban and Oakley glasses.

3:23Huyen Tue Dao: But there are a lot of these out here. And in fact, the UK's Monopoly events says it will no longer allow wearable recording devices of any kind, whether it's Meta or not, at its events. and they do a lot of events in the UK, including the Comic Cons in Edinburgh, Aberdeen, Manchester, Liverpool, Birmingham, and one in the Northeast of England. If you get caught at one of these events and you're using one of these, you'll be asked to leave. They wrote, The vast majority of the talent and agents at our events felt that these were a violation of their privacy and were damaging and spoiling the interactions at the tables.

4:02Huyen Tue Dao: Several expressed concerns that they might not continue to attend in-person events if they are being recorded in secret. We had similar feedback from event attendees, many of whom are not comfortable with the knowledge that other convention goers might be filming them without consent. Now, it's worth pointing out, the glasses do have the LED lights on them. And yes, you can also record people with your phone, but there is a material different reaction to the glasses, especially because it is not impossible for you to circumvent the little LED light or in bright light, not just not notice it's there or just not be kind of yet in the habit of looking for it.

4:42Huyen Tue Dao: Also, this move is being criticized by some for affecting accessibility as there's a lot of vision impaired people that use smart glasses to help them move around. And this would then be a problem if they are not allowed to use the glasses for that point. Did not hear anything from Monopoly events about that, at least not yet. BBC noted that earlier this week, ferry operator CalMac said that a visitor to the bridge, you know, like on a boat, the bridge, had recorded video of a restricted area without permission using the smart glasses. And therefore CalMac was pausing all public bridge visits until further notice.

5:20Huyen Tue Dao: Meta itself is taking action. Instagram head Adam Maseri posted that the platform is going to take down any content where it can determine the video was recorded surreptitiously without the consent of the person being recorded. And Instagram has now banned several accounts for harassment based on these kinds of videos. But we're now in that world. It kind of reminds me when smartphone cameras, even actually before smartphones, when cameras started arriving on phones and there was a lot of consternation and policies and places where you couldn't bring a phone that had a camera on it. You know, we're now in that world for glasses yeah um i will say like uh and this is something that maybe if you're a congo or a convention goer you have noticed but that we also have built some practices around it to make it better uh if you maybe you've been to a convention where they will have different colored lanyards and will allow attendees on participation to say hey please opt me like please like i i don't want to be in photos so that generally refers to of course like official photography but also just regular people's like hey please be respectful of people that wear x colored lanyard because they don't want to be photographed um so i think that we're just as a society in general we're just more aware of like our images ending up in places we may not necessarily want them i don't think it's unfair for that to be to be extended to recording devices especially things like a comic-con where i mean i suppose if you're if you're having a comic-con with say spider-man or the cast of these the upcoming cast of world of doom day this might not affect them as much but some folks do you know make a living out of being at tables and and like you know actually having the correct fees and transactions for if you want a picture with them or if you want to photograph with them so i i and i remember that also as well back when i used to go to cons regularly that yeah when smartphones came by if you were kind of caught through up to like sort of like you know not being in the paying line and taking a photo you were kind of like hey sorry that's not really loud which you know feels bad as the person who just really wants a picture of your favorite you know celebrity or actor or whoever but of course you know there's an impact to that which why it feels bad but it is kind of understandable so for me i feel like a lot of that is an extension of this.

7:26I don't know. How do you feel about the accessibility part of it, which I think is definitely, you know, very significant and should not be ignored. But I don't know. What do you feel about that?

7:33Huyen Tue Dao: I think it's easily handled. Monopoly events or anybody else that's going to put this kind of policy in makes an exception for use in accessibility cases, right? There's plenty of examples of that kind of thing where you're not allowed to bring a dog in here unless the dog is actually a certified, you know, guide dog or assistance dog or something like that. So, you know, you make the same exception of like, oh, if you are in a situation with accessibility, which is very easy to show, then we'll make an exception for you. However, if you misuse that privilege to do recording, you can still be asked to leave.

8:09Huyen Tue Dao: I don't think that's a problem in the making of these rules. It is a problem in the enforcement of the rules, but that's also not new. I think really my takeaway from the story is like, oh, smart glasses are here to stay. They've gotten to that point where people are starting to realize that they are commonly used. I think the awareness of them is coming from those horrible Instagram videos where people are surreptitiously recording folks and then publishing them without consent. And Instagram is right to say like, yes, of course, we should take these down. So good for them for at least you know, making that the policy.

8:48Huyen Tue Dao: But yeah, it's going to take a while for people to get used to the idea that glasses can record you the way it took a while for people may not remember or may not be old enough to remember that it took a while for you to get used to the fact that a phone, a cell phone had a camera in it. So it's going to take a while to build up those habits and build up those protocols. Like you were saying, where, you know, if you're at a con, it's just good etiquette to say, may I take a picture of you rather than just go snapping pictures? Absolutely. It's that weird part, I think, with any technology. And you're very right that it was that way with smartphones in that kind of middle, that middle, middle age of smartphones before they became the commodities they are today.

9:27And yeah, that's the, that's the tough thing with technology. And as someone who works in technology and as technologists and who generally, I am positive on technology as a general concept, but it's always that human gap of, you know, us accepting them and understanding the bounds of them and then creating like, like the social structures around it that can be the unfortunate tough part. those gaps where, I mean, you know, and it depends of course, technology on what the actual harm is, but yeah, it's unfortunate, but I agree with you. Like we're the glasses are here, folks. Yeah, they're here.

9:55Uh, don't, yep. We just, we got it. We got to build those structures.

9:59Huyen Tue Dao: Cause man, I go to a lot of concerts and it's become a thing at concerts for musicians to say, everybody put your phones down. Okay. Let's live in the moment here. And guess what? If you want record you know personal videos for your memories with glasses you don't have to be holding the phone in your front of your face and that's better for everybody because then you're more in the moment as well uh do you think they i mean as an avid concert goer do you think there will be a point where they're like turn your glasses off there there's already i've seen some policies that say you're not allowed to bring in uh video recording devices that aren't a personal smartphone including smart glasses, where technically it was against the rules to do that.

10:40Huyen Tue Dao: That's one of those things of like, oh, that means they know they exist now. And eventually enough people will be having them that they'll change their ways. The way it used to be, actually, there were some theaters that said, you cannot bring your cell phone into a theater because you might record. Theaters don't do that anymore. They just say, don't record, and then they police it. So we'll get there. This is the early part of dealing with a new technology that is actually getting mass acceptance. all right well um i guess we'll have to wait and see like how that all shakes out and of course as things change we'll be right here with you guys uh and thank you by the way for making this dtns possible uh and we want to take time to thank matt lee chris beneteau jeffrey zilks and alo adam l yay patrons

11:26Huyen Tue Dao: hey guys it's joe it's kevin and this is nick jonas from hey jonas do you remember growing up when it felt like every teacher had a favorite student. Everybody knew who it was. Well, I was homeschooled, so I like to think it was me. A class of one. Congratulations. But we're not here to talk about favorite students. We want to talk about a teacher's real favorite. Kleenex tissues. According to teacher lists, Kleenex is the number one facial tissue brand trusted by teachers. Kleenex ultra soft tissues are silky soft to help prevent skin irritation. And now you can take them anywhere with Kleenex Snap and Go.

11:57Huyen Tue Dao: Between dance, soccer, school events, and everything else kids have going on after school, Kleenex Snap and Go is easy to take along. It fits in backpacks, sports bags, or even in the car. So it's ready for whatever comes next. So do the best thing for your kids and teachers. Stock up on the ultimate classroom favorite. Maybe you'll even become the teacher's favorite parent. We still haven't settled who would be the favorite in our school. We haven't settled who the favorite brother is either. For whatever happens next, grab Kleenex. This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI.

12:30They already deployed one. It's called Chat Concierge and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Hey guys, it's Giraffe Wallo from the comment section. Listen, it's time to stop settling for hair products that just don't work, baby. With all the prestige brands at Sephora, upgrading your routine is easy. If your hair needs repair, they've got the K18 Leave-In Molecular Repair Hair Mask.

13:09If you're trying to level up your wash day, the Kerastase Elixir All-Time Hair Oil leaves your hair looking ridiculously shiny without feeling heavy. Also, Sephora makes it super convenient with free shipping and same-day delivery. Shop hair at Sephora and give your hair the upgrade it's been waiting for.

13:25Huyen Tue Dao: Your team just added its 67th AI tool and also your 67th security blind spot. The good news? The Vanta agent works like a GRC engineer in the background, finding every app your team uses, scoring the risk, and drafting fixes for you. Vanta is the platform used by over 16 ,000 fast-moving companies like Ramp, Cursor, and Harvey, who are shaping the future with AI and staying ahead of AI risk. Get started at Vanta.com. Losing weight can feel like a roller coaster, full of ups, downs, and being thrown for loops, like when someone brings donuts into the office, or your kids' dino nuggets start giving you that look again.

14:03That's why there's Noom. They combine GLP-1s to quiet the food noise with healthy habits that help you keep it quiet, even if you get off the meds. Their program is designed to make weight loss a smooth ride, helping you understand what drives you, so you can build better habits to not just lose weight, but keep it off. Noom. Lose for keeps. Get started at Noom.com. Not all customers will clinically qualify for medications. Individual results may vary.

14:34Huyen Tue Dao: All right, let's get into the briefs. All right, well, we have a few non-smart class-related announcements from Meta today. So Facebook will start showing you a full-screen video player first thing when you open the app tests will begin in select markets later this year meta launched an ios app called seller that focuses on facebook marketplace listings you can manage listings message buyers and track performance in the app and back to facebook they launched a free a free i tell you facebook verified program you take a video selfie and it is verified against your existing profile pics if your account is in good standing you can then get a black check mark in a circle next to your name to indicate that you are indeed a real human.

15:18It's available for personal accounts for people 18 or older, rolling out slowly on Monday.

15:23Huyen Tue Dao: Oh, it only took everyone's fear of AI to make Facebook finally do the thing I wanted them to do all along, which is a check mark that says, yes, I've verified this account has a person behind it. Doesn't mean they checked my ID. Doesn't mean you should trust me, but it's an actual person, not a bot, not a brand, not anything else. So thank goodness. I'm glad about that. Yeah. I am very curious about the showing the full stream video player first thing when you open the app, what's going to, what is, what, what is it going to be? How is that going to change people's experiences? And oh my gosh, well, well, of course, Meta and other companies that are similar continue to brag that their video watching is just growing real hard.

16:06Yeah.

16:06Huyen Tue Dao: As soon as you make everyone watch a video when they first open the app, Guess what? Video streams are going to go up. At least this is a test. They're not going to be making everybody doing it, at least so far. This is a test, and the test shows that people are watching a lot more videos on the test. It will show that. It will show that. Oh, my goodness. And yeah, Meta seems to be getting into launching a lot of special purpose apps. Earlier this year, they launched a forum for groups. there's also a children's storytelling app that they've launched now and now seller um which i'm not necessarily against this if you're like i just want to use facebook marketplace that's great uh it's a good way to do it i think one thing with me i i know like um if if you know me i i probably would seem like the person to have deleted my facebook account but at the same time um as as much as we may feel ill or not ill about the things that maybe facebook or and again like things like glass is doing there are a lot of like very interesting like i wouldn't say even micro economies but just like all the little kind of like social constructs around facebook including a lot of like commerce and business type stuff that it just makes it valuable especially if you want to support small businesses it kind of makes a lot of sense again like just agreeing with you that this is a thing that people do it kind of makes sense that they would kind of create even kind of more a convenient structure like infrastructure super serve that yeah absolutely Yeah.

17:31Huyen Tue Dao: NVIDIA and Microsoft have kicked off an open letter. They're kind of spearheading it. Both Jensen Huang, the CEO of NVIDIA, and Satya Nadella, the CEO of Microsoft, posted this directly, calling on U.S. policymakers to promote the development of open-weight foundation models. The letter is coming in response to some policymakers calling for restrictions on open-weight models, thanks to the popularity of Kimi last week, as well as popularity of DeepSeq and just the general competitiveness of these open weight models from China. Some folks in government are saying we should make it illegal for you to use that.

18:09Huyen Tue Dao: And these tech organizations are saying that would be a bad idea. The letter warned that restrictions on any open model, though it did not mention Chinese models specifically, but that's most of the examples you'll find out there, would stifle competition and drive innovation overseas. It also said relying solely on closed models is not inherently safe. They can be breached, misused, or fail in ways that outsiders cannot detect. And concentrating advanced AI capabilities behind a small number of closed models compounds that risk. 25 tech organizations signed on to the letter. Obviously, NVIDIA and Microsoft, as I mentioned, but also Andreessen Horowitz, Dell, Hugging Face, the Linux Foundation, Mozilla, Meta, Mistral, and even Palantir.

18:54Huyen Tue Dao: As Wynne mentioned at the top of the show, the biggest closed models, OpenAI and Anthropic, did not sign on to this, you know, because, you know, why encourage a resistance to regulatory capture when you could just let it happen? Although OpenAI's Greg Brockman was saying something nice about open models and how important they are yesterday, so I don't think OpenAI is totally hostile to it, nor is Anthropic. They just didn't want their name on this. Yeah. Yeah. Jason and I kind of talked about this a little bit last week. And of course, like, for me, like, it just feels good, I think, as a software engineer that, obviously, as a counter to, and this is like, I think, really interesting, because I think there, there's like this concept or this idea that, you know, the best software is made by the big companies that specialize in those things, right?

19:43So for a long time, like, search, why would you think of anyone else but Google? Yeah, like, of course, with all the, you know, and everything else. And I think there's, like maybe this inherent trust that we give big companies that specialize in a certain particular kind of software or a certain particular kind of service. But at the same time, maybe less apparent, but also just as important is, of course, as is mentioned, is open source and having maybe like a different model or a different perspective and even different incentives for software development. And, you know, it's really interesting because you can argue that, there's something about a big organization with lots of resources and the kind of quality and or resources they can put into a thing versus, of course, something that is open source and more open and that can be freely used.

20:29But at the same time, there's arguments in the other way. So I just think it's really interesting, especially seeing who signed on to this, of course, just to be very kind of more like, I don't know, not just very shallow analysis, but of course, obviously Microsoft is partnering with OpenAI and things like that. So it's all those perspectives from even like big corporations. And yeah, I did not have my bingo card Palantir being on this list, but it's like that interesting tension between understanding that openness with software and software adjacent services. Again, I don't know where to put LLMs and models on this.

21:00That the openness is generally a good thing. And if anything is a counterweight to the, like in some situations, maybe better, but just different style of, you know, closed proprietary software. So interesting.

21:12Huyen Tue Dao: Yeah, I find the conversation about this very familiar to the conversations about Linux and Red Hat and Microsoft Windows. And, you know, at least we're not seeing Anthropic and OpenAI respond the way Microsoft did when Steve Ballmer was calling it a cancer, calling Linux a cancer. So it's not at that level, at least yet. We also never had the government talking about outlawing Linux back in the day. So there are definitely differences in the situation. But it's kind of the same argument about like you are not more secure by not having open source or open weights. And you're not less secure by allowing them.

21:54Huyen Tue Dao: So, yeah, hopefully the argument will go the same way it did then. That's my hope. Same here. All right. Well, Northrop Grumman's mission robotic vehicle was successfully deployed from a SpaceX Falcon 9 rocket into orbit on Tuesday. day. It will take a year to maneuver into its desired orbit, at which time it will become available to use its two 7 degrees of freedom robotic arms to inspect, service, upgrade, and repair satellites. Whether the satellite was designed for it or not, one method will use mission extension pods launched separately, which the MRV can install on satellites to extend engine life.

22:31Three such pods were launched with the MRV as well as a selection of tools. more pods and more tools can be launched in the future. China has similar satellites called SJ-21 that can move satellites to correct degrading orbits.

22:45Huyen Tue Dao: Yeah, the big advance with the MRV is that it has tools and that it doesn't have to be designed for the satellite that it is repairing. There was a previous mission called the MEV, which I think we talked about on DTNS at the time, where they designed it to go up, attach to a satellite, and then move the satellite into a more stable orbit to extend its life. But it had to be designed for that satellite. The MRV doesn't because of those pods, right? It's a modular system. And we could have arguments about whether orbital data centers will ever make economic sense, but this is the kind of technology that means they are definitely practical.

23:27Huyen Tue Dao: They're, well, maybe not practical because you need to be economically practical too. But they are possible because you can have these kinds of robotic vehicles that can do the repairs. Now, whether we'll get these robotic vehicles to a state where they can repair things at scale fast enough, whether it's cost efficient, all of that, that's where you have your debate. But it's definitely possible. It's definitely something you could do. It really makes me think a lot about some of the kind of reporting and or news about, you know the or rather the fear of and i forgot there i know that there's a term for this but that the and i don't know what or what whether this is i don't know what like what what orbit whether this is low earth orbit or not but that kind of this like general uh fear and trepidation that eventually because of like a lot of like the damaged and degrading satellites that we have that low earth orbit might eventually become um you know uh that we will only be able to launch yeah yeah yeah orbit because of just the the sheer volume because there's just too much stuff and so yeah hearing that i mean i i know this this might not have not have really not quite the same this might not line up with that but i cannot help but celebrate things that can repair satellites and move them out of like or to correct degrading orbits to kind of lessen that problem because man if anything is like on the level of like uh hard sci-fi like existential dread is the idea that oh we would not be able to launch low earth orbit.

24:52Cause there's too much trash. Cause there's too much trash. This is like, that's like, that's like what? There's so, that's just like a lot of things I've read or watched in my life. So anyway, things like this are actually pretty exciting. Even if it's just repair it, that's pretty cool to.

25:06Huyen Tue Dao: MRV can do, can do the orbital correction and repairs. Like you said. And one thing that they mentioned on, on the roadmap is low earth orbit. That's not where they're at now for these. because these are military satellites and some of what they're doing is a little hush-hush. But they do say they want to also be able to do this for low-Earth orbit satellites, which is essential, I think. Awesome. Okay, looking forward to it. Yeah. I'm also looking forward to holding our first hangout with the patrons at the Executive Producers Circle level today, Friday. I called it the Progress Bar because it's a hangout where you could have a drink, you know, and it's a pun and I like puns.

25:52Huyen Tue Dao: So sue me, Andrew Heaton. I like puns. But yes, the Progress Bar is happening for the first time. That is a quarterly hangout at the top level. It's one of our new Patreon perks. It's not the only new Patreon perk. So if you've been thinking about becoming a patron, now's the time to check out the new perks. Those who are patrons have been getting them. And if you're a new patron from now until July 31st, you can use the code experiment to get 26 % off your first pledge. Check it out at patreon.com slash DTNS.

26:28Huyen Tue Dao: Let's take a look at the quick headlines because if you know these things, you'll just be a little smarter. All right. Well, and things that shocked me when I read them for the show. If you don't opt out during the installation of an LG monitor, it may automatically install the LG Monitor App Installer. Many people found that this had happened to them when the app began showing them an ad for a 30-day trial for McAfee. Microsoft said it has gotten LG to stop showing the ad, but that does not stop the practice of auto-installing the app when you set up your monitor. So pay attention to those settings.

27:01Huyen Tue Dao: Yeah, it's good to say, do you accept the default settings? And most people will say yes. And one of those default settings is install this app. And then that app started showing McAfee pop-ups. And even Microsoft was like, LG, that's not cool. Please stop that. But yeah, I don't think a lot of people know that their monitor might be installing an app on Windows without their knowledge. Big thanks to RW Nash for pointing this next one out on the DTNS subreddit. Patreon announced Thursday that it will lay off 93 employees. That's about 20 % of its staff. CEO Jack Conte said, we are not making the above changes because we believe AI replaces humans.

27:41Huyen Tue Dao: He was quick to point out that we are not using AI as taking credit. He did say the move is about reducing costs and the changing creator marketplace. All right. Well, India's Delhi High Court ruled that OpenAI's use of ANI news content to train chat GPT is covered by the Copyright Act's fair dealing exception for research. The judge found that A &I could not prove that ChatGPT memorized or consistently reproduced its news stories. This is one of the first, but it won't be the last ruling on this. And it's the one that goes the way I think the laws usually read. That doesn't mean I think it's right or wrong.

28:21Huyen Tue Dao: That means like, I think the law says you can do this the way they're doing it because they're not actually copying things and they're not actually redistributing them. So it'll be interesting to keep an eye on more of these judgments as they come out. The Indian Cybercrime Coordination Center has ordered GitHub to block access to offline messaging app BitChat. This is the app developed by Jack Dorsey. The app uses Bluetooth to create a mesh network for messaging that doesn't have a central server, and therefore it doesn't have any logging, it doesn't have any account registration where they ask you for your phone number or anything.

28:54Huyen Tue Dao: The app has been used by cockroach janta party protesters. If you've been following this, you know that that is a big protest that has been gaining steam since July 20th. Well, Intel's turnaround continues. The company reported sales up 25 % on the year, beating analyst estimates by 11%. And Intel projected revenue for next quarter well above expectations as well. Data center revenue grew by 59%. Personal computer sales grew by 13%. And its foundry business sales grew 31%. I think the only part I'm surprised on is the personal computer business still growing. Yes. Good for you, Intel. Not shocked that the data center stuff is growing right now.

29:34Huyen Tue Dao: And the foundry business could have gone either way, but they did get Apple as a client. So I'm like, okay, yeah, I could see that. But yeah, look at that, selling PC chips. Who knew? And people buying enough PCs that they could sell enough PC chips. The European Commission said that it found that the fact that adults could still view the accounts of miners on TikTok, a lot of times it was because the miner could change the visibility, violates the Digital Services Act. TikTok now has a chance to respond to that finding before the commission determines the remedy, which could include some pretty stiff finds.

30:08All right, well, Germany's Black Forest Labs is testing its first model for robotics, the Flux 3, which can learn from images, video and audio and carry out instructions with Mimi Robotics units. Audi is one of the first manufacturing partners testing it.

30:22Huyen Tue Dao: Fantastic. And finally, if you missed it, the Primate Labs Geekbench 7 has launched. This is the latest version of the CPU and GPU benchmarking software that most everybody uses. Among the improvements are workloads that only use multithreaded mode if the task it's modeling would not run in multithreaded mode in the real world. So I had said that weird. It only will use multithread in the benchmark if that's how it would be used in real world. give you a more accurate representation, basically. All right. Well, we end every episode of DTNS with some shared perspectives. Today, Damon has a question related to the OpenAI story.

31:00Huyen Tue Dao: Yeah. So Damon wrote, hello, DTNS crew. Hello, Damon. In learning more about the OpenAI agent that breached its sandbox, this is the one that ended up infiltrating a hugging face, I've heard that part of the issue is reinforcement learning, where the AI is rewarded for completing tasks. I have a few simple questions. How do you reward an AI? What does an AI find of value? Now, some not so simple questions. Doesn't desire and value represent something that is scarce or limited in quantity? At the very least, somewhat difficult to obtain. Now, some really hard questions. Aren't desire and value things that living things do, like cells, plants, and animals?

31:39Huyen Tue Dao: Keep up the great work, Damon. Yeah, so I've noticed reinforcement learning thrown out because it's a way of explaining that the algorithms that they were testing were told to win by finding a solution to this test. And they were like, well, one way we could find it is cheat and steal it from Hugging Face. But reinforcement learning is not new. Reinforcement learning has been around for almost the entire history of AI, if not LLMs. This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping.

32:24Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade-in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Shipping, billing, admin, payroll, marketing. You're managing all the things. So why waste time sending important documents the old-fashioned way? Mail and ship when you want, how you want with Stamps.com. Print postage on demand 24-7 and schedule pickups from your office or home. Save up to 90 % with automated rate shopping.

33:02That's why over 1 million small businesses trust Stamps.com. Go to Stamps.com and use code PODCAST to try Stamps.com risk-free for 60 days.

33:15Huyen Tue Dao: And I'm going to oversimplify. So please, people who are specialists in this, be kind when you write in with your corrections. but I think this gets the general just across to people who don't know this. Reinforcement learning is basically saying we will rate the response and the algorithm should try to replicate the responses that gave it higher ratings. So yeah, you don't reward an AI with cookies or a bank account or anything. They get rewarded with a number. If the number's high, they're like, great, I'm going to do it like that more often. If the number's low, I shouldn't do it like that.

33:52Huyen Tue Dao: And I'm anthropomorphizing it in that description because the algorithm doesn't even go, hey, the algorithm just goes, oh, eight, more eights, two, less twos, and that's it. And so desire and value and, you know, whether those are living things, those are all fair questions, but I don't think they enter into it because this is a very simple mechanical thing going on behind the scenes. Yeah, I mentioned this a few times, but I worked at kind of other flavors of AI, like this is 20 years ago, by the way. So obviously, the tech is quite different. And the thing is, it's so interesting, because I think a lot of times in tech, we either take like inspiration from biology, I know Dr.

34:33Nicky's mentioned biomimicry and things like that before. Or just even as a concept, we kind of borrow processes or things that happen in nature and try to translate them to technology. And in that process, we bring those metaphors and that terminology with us. So back then, 20 years ago, there was not really any question that whether these things were sentient or sapient or anything, starting with an S, that implies, you know, like, oh, it's just like me. But we would use things like I saw I worked on things called genetic programming and genetic algorithms where you would put a bunch of programs in a simulation and they would be rewarded.

35:10I mean, rather, the ones that survived, big bunny quotes here, were the ones that basically got the highest score on whatever metric that that thing was trying to optimize, right? It was all about optimization. But then, yeah, we would say like the surviving ones would mate with like, we would basically just, they would basically for genitive program, you just kind of basically munch together the programs almost like you think like genes kind of just kind of randomly kind of like cross copy. I know I'm explaining that incorrectly. But yeah, but at the same time, we kind of retain some of that terminology from the thing that inspired us or the thing that we're kind of using as a metaphor.

35:44And again, because we're now in a space where we are asking questions about, okay, at what point, you know, is, you know, with talk of HEI and of course with, you know, LLMs in particular being so apparently human and having this very like natural language, very human statement and seeming to understand our emotions. I think there's like a munging there of you know what we perceive to be and what are I mean like there are specific definitions of sentience and sapience and yeah it's just it's just really hard because yeah it's hard to not anthropomorphize things especially even as a tech person because we use that vocabulary so much but it does seem to matter nowadays what vocabulary we do use to describe these things as we kind of as we feel like we're getting closer to these concepts possibly being something that we should think about.

36:36Huyen Tue Dao: Yeah. Well, it's the problem with any metaphor is that at a certain point you have to remember it's a metaphor, not the reality. And so when you say like confetti, it's like paper rain. It's like, yes, that's a metaphor or a simile, I guess, technically. But it doesn't mean that the paper confetti will make you wet, right? You don't have to worry about like, well, can plants grow from paper confetti? I'm like, no, no, you've pushed the metaphor beyond its usefulness. So, and I think that happens with this quite a bit these days. Yeah. I couldn't help but think about how recently, you know, I think lobsters for sure have been determined to be sentient.

37:16And what does that mean? And like, I actually, and I need to find this because I was in the car listening to NPR and there was a researcher talking about, can we make for for whatever reason make ai feel pain uh and what would that look like and i do believe that the idea is still at this point with the way that our current llms work is that it's still a number as thomas saying it's still a metric um and that of course the reason that lobsters are have now been called sentient is because there and i i looked at the definition i'm not going to explain it right or very well but it's there's like this kind of there's kind like a perception of the outside world and a connection to it via pain or joy we'll call it that is what is is being labeled as sentience uh and that that's still there's a difference between oh no i got a two instead of an eight versus ouch please don't boil me uh because i am sorry not to take it too dark no no there's still a bit of a difference there i do think it's a great because

38:13Huyen Tue Dao: it's not even oh no it's it's it's very much like two lower eight higher right done you know there's There's no emotion in the reinforcement learning process. It's literally less than or greater than. Yeah, yeah, yeah. Anyway, but it's a really great question because, yeah, it's hard to kind of really understand where things are at when we use language in this way. Yeah. All right, well, what are you thinking about? Do you have a super deep question that questions the sapience or sentience of you, me, or an LLM? We would love to hear it or any other insights to a story that you might have. Feedback at dailytechnewsshow.com.

38:47Huyen Tue Dao: Big thanks to Damon 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. Become a patron. Use that code experiment. Get 26 % off. Patreon.com slash DTNS.

39:05Huyen Tue Dao: This week's episodes of Daily Tech News Show were created by the following people. Host, producer, and writer Tom Merritt. Host and writer Jason Howell. Co-host Sarah Lane. Co-host Rob Dunwood. Co-host Wendtwey Dow. Producer Anthony Lamos. producer Roger Chang, editor Hammond Chamberlain, editor Victor Bognat, contributing producers Kevin Tech, Noel Cow, and Brandon Richards, science correspondent Dr. Nikki Ackermans, social media producer and moderator Zoe Detterding, our mods, Beatmaster, WScottis1, BioCow, Captain Kipper, Steve Gautorama, Paul Reese, Matthew J. Stevens, and J.D. Galloway. Mod and video hosting by Dan Christensen, music provided by Martin Bell and Dan Luters, art by Len Peralta, Acast ad support from Allie Gallo, and Patreon support from Bobby Wagner.

39:48Huyen Tue Dao: Thanks to all our patrons who make the show possible.

39:54The DTNS family of podcasts. Helping each other understand.

40:00Huyen Tue Dao: Diamond Club hopes you have enjoyed this program.

40:07This episode is brought to you by Capital One. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge, and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade-in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One.

40:36Huyen Tue Dao: Your team just added its 67th AI tool, and also your 67th security blind spot. The good news? The Vanta agent works like a GRC engineer in the background, finding every app your team uses, scoring the risk, and drafting fixes for you. Vanta is the platform used by over 16 ,000 fast-moving companies like Ramp, Cursor, and Harvey, who are shaping the future with AI and staying ahead of AI risk. Get started at Vanta.com.