 This 10th year of Daily Tech News show is made possible by you, the listener. Thanks to all of you, including John and Becky Johnston, Chris Benito, Steve Ayerderola and new patrons. Welcome on on in Daryl, Michael, Alex and EpoopMePants.com. On this episode of DTNS AI tools to help doctors and local news journalists and why threads is succeeding by imitating Twitter. But shouldn't that also work for Twitter? This is the Daily Tech News for Monday, July 10th, 2023 in Los Angeles. I'm Tom Merritt from lovely Cleveland in Ohio. I'm Richard Raffalino from deep in the heart of Texas. I'm Justin Robert Young and on the shows producer Roger Chang. Those of you who just get DTNS and not good day internet don't know that we we escaped destruction today. My my fix to audio hijack worked nice and we're nice streaming right along. Um, I propose we start with the quick hits. Are there any objections? Nay, nay. Open AI is making its code interpreter plug in available to all y'all. Chat GPT plus subscribers, that is. If you're not a chat GPT plus subscriber, you're not all or y'all. Open AI developed the plug in for its own internal use. It lets chat GPT write and run code in Python and access files and manipulate files that you approve it to access up to 100 megabytes. That lets chat GPT do things like generate charts, create maps, do data visualizations, graphics, interactive HTML pages, then a whole lot of other stuff appears to be especially useful for data scientists dealing with complex data sets. Well, now we know what Austrian activist Max Schrems will be up to in the near future. That's because the European Union announced it adopted a new transatlantic data adequacy agreement with the United States. You justice commissioner Dieter Raidner said the agreement will allow for personal data flows between the two on the basis of a stable and trusted arrangement that protects individuals and provides legal certainty to companies. Of course, that's what the prior two data sharing agreements also promise, but those were struck down in court over concerns that European data could fall under US surveillance powers. And since there's been no change to US surveillance law, that issue still persists. Max Schrems remains critical of the new agreement, saying he expects the issue to be before the Court of Justice for the European Union by the start of 2024. Yeah, we have no little more episode about this. And basically every time they try a little, OK, but if you're European, you get to go to an ombudsman. The European Court of Justice says yes, but the US can still spy on you until they change that. I don't know that this works. So we'll see. BMW Motorrad, the motorcycle division of the German automaker, announced the Connected Ride Smart Glasses. These provide a heads up display for motorcyclists showing current speed, speed limit on the road, bike gear and GPS instructions, all in a little heads up display. It's kind of cool. Smart glasses need to connect to the Motorrad Connected app. They give you 10 hours of battery life, more or less. Kind of look like regular sports on glasses. Maybe the arms are a little beefier than otherwise. And they cost you 690 euros. No word on a release date, though. Fox kind announced it with Drew from a joint semiconductor venture with the Indian conglomerate Vedanta. The two announced plans last year to build semiconductor and display plans in Gujarat, India. The Vendanta said it remains committed to the plans for the new facilities and has already lined up other partners for the project. And Twitch announced it's rolling out support for an ephemeral stories feature coming this October. These will show up in the following page on the mobile app and creators will be able to limit visibility to followers. The mobile app will also add a discovery feed in late 2023, featuring a mix of live and recorded content. All right, let's talk. Yes, indeed, Tom. Let's talk about the Wall Street Journal. And they had an article out with sources saying Google began testing a language model called Med Palm 2 with some health care companies, including the Mayo Clinic back in April. This model can be used to answer medical questions or perform tasks on health data, things like organizing or summarizing content. The model itself is based on Palm 2, which Google announced at its IO conference in May, and it's been rolling out to its barred chat bot. Google trained Med Palm on a curated set of medical expert demonstrations and sees it eventually being useful in places with more limited access to doctors. Kind of a big deal. So Justin, we're kind of in the early stages of seeing these more specialized large language models rolled out. Obviously, we have kind of the catch all ones like chat GPT. I'm curious, though, how soon do you think something like a medical chat bot is going to be normalized? It is currently normalized, at least in this household. I have been going to Dr. GPT for various different small ailments over the last few months and considering the price is right. And oftentimes I can ask follow up questions with a fairly high level of assurance that this is at least worthwhile to try or at least gauge a sense of general medical knowledge that I can then take to a medical professional. It has been extraordinarily helpful. The reality of the system like this is not about whether or not a large language model that is trained on medical text can be helpful. We already know that they can. What this is being trained for is so it can pass muster for any legal liability and make sure that it passes any kind of qualifications that would fit under the insurance that doctors operate under. Justin, when you ask Dr. Chad for medical advice, how do you know it's correct? Well, oftentimes I will also check it with other stuff that I find on the Internet. And these are little things for your cross referencing and using a common sense and these are things that basically I would love to ask somebody on the street if they told me they were a doctor. Sure. That's how I would. That's how I would. I think that's the important difference here, too, is what they're doing with MedPalm is not a public outreach. This is how can we help the doctor itself? So I think that is also part of this is and this doesn't in any way contradict your statement about medical liability because once you're involved in a doctor, you've got all kinds of other regulations, not just HIPAA involved. But I think what Google is trying to do here is say, let's make something that helps in the hospital setting, that helps the professionals, that helps the doctors, helps the nurses, not just something casual like what you could do with Chatchi. Well, yeah, but I think even then the solution would be effectively making Chatchi PT dumber in that it would be trained on less and not more. And so therefore, it would only be these medical texts and medical cases for which it would be trained on versus that plus everything else, which is where it is now. So what I do want to highlight here is that based on where we are, we have something that is medically efficacious. Chad GPT will never tell you that it is because they don't want those problems. And that's why it is at your own apparel that you go on and go through it. If you are just looking for AI to be a more effective summarization and search tool for a very limited trained dataset, I believe it can be extraordinarily effective. And that's what this is doing. So I do wonder, obviously, you know, this use case, this is a extraordinarily profitable opportunity for Google, Microsoft, whoever open AI to get into this market and provide these services like health care is like an internal industry. Right. But on the other, like I also, based on how you're describing, it's already you're already using it, Justin, I also see this as like a rush to we don't want the headline that Bard or chat GPT said, do medical treatment next, even though it disclaims liability for it, that will be the headline on some, you know, on, you know, New York Times or something like that. And so like the efforts to train these also is to get ahead of this is how people want to use this. So let's make it a good thing to use. But again, again, I think it's important to note MedPalm to not meant to be WebMD, which is how Justin's using chat GPT, right? And to that side, WebMD went through exactly the same stuff of like, don't rely on WebMD, WebMD had to put up a bunch of things saying, this is just general advice, consult it with your physician. All the things Justin is very wisely doing when he uses chat GPT. What MedPalm do to is doing is saying, I can help a doctor. And I think that's where it's interesting that it is as good or better than doctors in reasoning in consensus supported answers. Like this is what everybody agrees is true. Signs of incorrect comprehension are not there. It's not like, oh, you didn't understand. Yeah, but it still has more inaccuracies and irrelevancies in its answers than a doctor does. And that that's the gap it needs to make. But I would say in research on this, they are looking at consumer medical questions, right? These aren't doctor to doctor. This isn't purely. This is a doctor to a consumer, but it's not consumer to the to the language model. It's the doctor to the language. Yes, but I guess that is still a like an expert to layman like conversation. And I don't I mean, I think there's the possibility of. I don't want to speak for the research or what their intentions are, but it seems to me that that is also in what Google is planning for down the road. Now, again, that is that is not specifically what the researchers looking at. So I don't want to overstate that. And let's also, again, let me go back to the legal side of this, which I think is very, very important. We just saw a lawsuit that came in from Sarah Silverman to both META and Open AI, along with a few other people that basically challenging the training models that it was learned that it was trained on. I don't think this is a particularly efficacious lawsuit, but it is indicative of the fact that we are going to see a lot of this stuff in the next five years. We are entering into dare I say, Tom, patent war territory. The legal flurry we will see over these LL out the old hits. Yeah, you're talking to my language. Yeah, no, there's going to be already Google, even in this announcement, it's like we never get access to any customer information. If it's going to use any health customer information in the future, that'll be kept within the hospital. We'll never touch it. So yeah, that's that's going to be a big part of this in addition to the training data and what it's trained on. Well, let's talk about something that's infinitely more important than your health or welfare journalism and gadgets. Andrew Tarantola has an article up about NYU Media Labs, AI and local news initiative now in its second year. Most of the headlines you're going to see screaming out there are about how AI is threatening journalism. These are the folks trying to be like, yeah, but it could also help them. Let's let's make tools that would help journalists. Six research teams recently participated in a demo day in June. I'll give you the highlights of what they did. The Gram Media Group did a natural language model to kind of improve the comments on a local news article. So the bot would go in and make the first comment to try to set the tone of the conversation and encourage positive discussion. A team from Cornell tried to create an equivalent of co-pilot, which is for coders, but for journalists. So it would do things like automate transcription, organize information, create images and headlines, improve the SEO. It could even write a first draft in your style if you wanted it to. Third one, WNYC Public Radio developed speech to text for real-time captioning for live broadcasts. It could also do summarizes of a clip. So if you wanted to just have a text summary of a clip, Noble Media is working on ad tech to better reach smaller audiences for smaller publications by geography or demography. Bangla AI surfaced and translated relevant news stories in Bengali for New York's Bangladeshi community and then basically translates for journalists and writes their wire copy for them in Bengali. And finally, a team called Chequeado showed off Monitorio, which assists in fact checking Spanish language stories and monitors social media for trending misinfo that it could then identify and journalists could write to combat things that are out there that are poorly understood or misunderstood. Justin, thank goodness we're talking about journalism, of course. What do you think local news local news doesn't get enough attention? What do you think about these attempts to create tools to help local journalists? Oh, man, AI, journalism. Don't worry. Thankfully, I'm carrying water for Twitter later in the show. So I don't have to jam it into this particular story. Yeah, I will say this about AI in large language models in general, as we currently understand them, they are writers tools. They are creators tools and we've been led down a path to think of them as the end point for creation and not a helpful tool for creation. I think that these are all great examples of using these to automate a lot of the busy work that goes into making fully formed and robust output and for local news that are operating on smaller and smaller margins. I think this is incredibly important. Yeah, seeing some of this stuff, it's not just in the creator tools. Like obviously like the Cornell stuff with their their version of co-pilot, as I think like like what I would think of is like the the assistive element to it, but I also am interested in these tools in terms of like trying to increase total addressable market, not just to like wring out more productivity out of an individual journalist, certainly increasing productivity. Now, you know, if these tools could do that, that's a good thing. But like, you know, adding providing translation for for different languages community within it within a given locale, I think could be supremely helpful in terms of like, hey, we can get this news in front of more eyeballs that couldn't otherwise consume it, that kind of same kind of thing for the news outreach. That to me is stuff I never even would have thought of, but would seem to me again is like a different tack, not just, hey, I can I can get the drudge work out of here, but this helps kind of expand our audience in a way that I kind of never thought of same thing with the translation for a WNYC text stuff. Again, tons of people, you know, with with hearing impairments and to be able to put that out there in real time is kind of something I never even thought was possible. Yeah, I think when we are crafting policy and when I say we, I don't mean the three of us here, but you know, when we as humanity are crafting policy, we will serve if called absolutely. When we're crafting policy on how to limit the uses of these new tools, we have to keep in mind this side of it. One of the problems with local news is that it is often seen as something that you can't make money off of. It's too costly. You have to spend a lot on a local reporter and then you've limited your audience by making it local. So anything that can help that local reporter produce more, produce at a higher level of quality is going to make the math for supporting local news work out better. So these are important things in providing local news. And when I say local, I think our heads tend to go to, like, you know, our hometown, but it also means communities like Rich was talking about, both communities by language, communities by ethnicity, communities by accessibility and these kinds of tools can make all of those communities reachable in a way that is economically supportable that we don't have now. We are going to help get you to a community for Android. We have been helping out with a podcast that is entirely about Android from Android aficionados, Ron Richards and Juan Tweedow. They're going to bring you a new show Tuesday. If you're listening to this Tuesday morning today, if it's Monday, then I mean tomorrow, but it's a podcast devoted exclusively to Android news and information launching Tuesday, July 11th at 8 p.m. East Coast, 5 p.m. Pacific. If you want to watch it live, go to twitch.tv slash good day internet and watch it live. You can subscribe to it right now. Make sure you get it into your podcatcher of choice at www.androidfaithful.com. Thank you, Ron and when for for bringing this to us and bringing it to the audience that was missing it. We appreciate it. Once again, all about it. Yeah. Android faithful.com. All right, well, you may have heard about this whole threads thing. And Instagram head Adam Osseri confirmed that Meta's threads app surpassed 100 million accounts since launching on July 5th. For comparison, TikTok hit that number nine months after its global launch. Chat GPT took two months. So compared to those, it was really fast. Now, it seems like Meta itself had no idea that threads was going to take off like this. Obviously, the draw of bringing over your Instagram followers made the new platform appealing, gave it a sense of acceleration. Investor Turner Novak posited that while this helped, the bigger draw was the TikTok style algorithmic feed being available at launch so that content from people with few followers still had reach. Interestingly, a post by veteran product manager Eugene Way bemoaned, switching to a TikTok style algorithm on Twitter as the major reason why that platform was so ripe for disruption. And one of the constant complaints about threads from people who otherwise love it is that they can't just get a feed of the people they follow. It's this algorithmic feed. So, Justin, you got 100 million out of 2.35 billion monthly active Instagram users. Why are we seeing so much excitement over threads and why would an algorithmic feed be a boon for threads when it wasn't for Twitter? I'll take the first question first. Are we supposed to be impressed by 100 million? Yes, you are supposed to be. It sounds a little crazy, right? But how much is 100 million really when it's not 100 million people that had to create a new account, not 100 million people that had to enter in their email address, not 100 million people that had to pick a username, not 100 million people that had to go to their email and click on a link to make sure that they have had the right account. They just hit the buttons like most of us did that are part of Instagram that has a gigantic two billion feature. Now, if they continue this insane growth, then it's something that I will be incredibly impressed by. But as an initial push for a very fuzzy app that is currently very much in the zeitgeist of our media is yet another new version of what this company Metta, which is currently in flux up to the tune of changing their name a couple of years ago. I think that there's a lot of reasons why there were a lot of eyeballs on this. I think very deep down the list to get to your second question is the algorithmic feed. I don't think that that matters one wit, unless we are talking about the future of threads and giving people the option to have an only followers feed, because I do think that that is something that is important. I do not believe that having an only algorithmic feed will be a boon to them going forward. I will push back a little on the unimpressiveness of 100 million people because you're rich. Yon. How many services has Google spun up that are going to be killers of name app X of of zoom of productivity app of social networking app? Google Plus, you had your same Google account and you could just you already using Gmail that how many million Gmail users that they could just plug right into their no problem whatsoever. I don't remember them hitting, you know, 100 million accounts or even five percent of that user base, which is about where they're at. You know, give or take a couple of 100 million Europeans that aren't on threads yet or applicable to be on threads yet. You know, I you know, that is still a big number. Yes, meta has to be graded on a curve because they are a portion of the earth of is their user base, right, of humans on earth. So in terms of that, yes, being only the size of, you know, part of the United States, I guess, total population is less impressive than if it were chat GPT, which is coming at you had to create that whole new account. They had to make a use case for a whole new, you know, technology stack, right? You're not emulating something that people know. That is still deeply impressive to hit that number in that short of time. But I don't want to say that this is this is nothing. Admittedly, even most Siri, though, was saying this is all about retention, right? Like again, we had an easy social, you know, we were using existing social graph. We made the onboarding as quick as possible, like two clicks and you were on. So, you know, they're even saying we don't know how sticky this will be. But I don't want to completely yawn, Justin, at the 100 million figure. Well, that's fine, because I did it twice. Let's let's discount the 250 million monthly active users on Instagram in Europe. We'll just take it. We'll just take that out of out of the count while they wait to see how the Digital Markets Act is going to be applied to making people sign up with an Instagram account, which seems to be the big rub there that Instagram is trying to feel. You're talking about around a five percent conversion rate. In five days, in five days. Yeah, I mean, a lot of attention, a lot of attention that's put on this product in this space, it is a very famous space right now. And and you're seeing some independent analysis, not just from Cloudflare, but but from some other folks out there. I'm trying to find the name of the place similar web estimating a drop in Twitter traffic because of threats. So I think it's fair to say you had an unexpectedly large effect. You didn't see this with other similar types of efforts to do a micro blogging service. Yeah, the bigger question is, is it a fad? Is it a thing that everybody went to because they could easily sign up with Instagram and will they stick around? And that's where the algorithmic feed will make or break it. It doesn't matter that people are complaining about it. People complain about things they use and love all the time. The question is, is it going to be useful enough that they get stuck there and say, well, but I wouldn't want to stop using threads because it's giving me all these things I want. Well, and this is where meta needs to pump on the gas in terms of all the other hook into Instagram, because right now the only real hook into Instagram is bringing your account over. That's a huge one. Don't get me wrong. But like there's no cross DMs or even DMs in thread. It's very clunky. There's no search. Like it is such a bare bones app. It is shocking that they shipped it outside of in a tremendous market opportunity that Twitter presented them with. So like they also don't have all of the hooks in threads yet to really kind of encourage that sticking to see that, which I think is important to consider. If you had enough hooks and threads, you'd have a hooked rug. Yeah. Well, what are your guys favorite things about threads? People interactions. It's it's the interactions I'm getting that I wasn't getting unmasked on that I wasn't getting on Blue Sky. Yeah, I got I've had less interactions on I have had less interactions. I like that you can upload any photo aspect. I've actually had threads where I disagreed with people and we didn't hate each other and it was nice. I don't know that that is a feature of threads or just because it's new, I feel like maybe just because it's. Yeah, those are getting to know you prices. Mm hmm. All right, let's check out the mail bag. All right, well, a long time DTNS supporter Tim wanted to share his thoughts about our story on Friday's show about an emoji being accepted in a contract agreement or as a contract agreement he writes in. A couple of you use the word interpretation in your discussion about the emoji and you hit the nail on the head and the legal realm determining what someone meant with an emoji won't be all that different from what currently happens with normal language and just involve symbols that can be more open to interpretation. But he gives us context during the trial that I was recently part of a significant amount of time was spent with lawyers on each side trying to interpret language, their advantage, like how a comma should change the meaning of the statute. This case is sensational because it involves an emoji, but it's really no different than what currently goes on in these types of proceedings. Yeah, I've been on the legal side of this, too, depending on what judge you get, what that comma means at the end of a law or something like that. Definitely seen that for sure. So thank you, Tim. This is why laws are so complicated. When you read a law and you're like, why didn't they write it in plain language? Because if you write it in plain language, you leave room for different interpretations and then the courts have to step in and go, well, I feel like this comma implies that that and that's why you need judges. So yeah, it would be great if they said, no, a contract is only valuable if somebody signs it, but that's for a legislature to do. That's not for a judge to do. The judge just has to look at the law and say, OK, based on what the law said and based on a reasonable person would say, here's what I think happened. And that's what they were doing in that case. Thank you, Tim. That was a great email. Also, Nick wrote in and said, hey, DTS team, Tom asked us for some uses of blockchain that did not involve crypto and I have one. Our county wildlife department put together a system a few years ago to track hunting and fishing licenses using blockchain. It didn't have to be fast and the public and immutable attributes worked well for this. We don't need to carry a physical fishing or hunting license around with us. We can use an app to pull the license and its history. It's pretty cool. Yeah, it is. I kind of love that too, because it's not where people's heads go when they're like, well, blockchain, I'm sure there's another use somewhere in some industry and enterprise. And this guy's like, yeah, like catching a salmon. Yeah, exactly. The hunters and the fish are the catfish on board. So yeah, we're good to go. Well, something we also found to use for was Justin Robert Young. Justin, thank you for all of your yawns and for being on the show. Where could people find you online and what are you up to? Well, you can listen to the aforementioned episode that Tom said, but know a little more at know a little more. You can find us anywhere that you are on. You get your fine podcasts and don't worry, friends, we are in production on the next round of episodes as we speak. Yeah. Give us give us a month or so. It's going to be the mother of all seasons. I'm telling you, good, good stuff. Patrons, stick around for the extended show, Good Day Internet, where we'll be pouring some out for Evernote. Now, Evernote's not dead, but it's definitely pining for the fjords. Remember, you can also catch the show live Monday through Friday at 4 p.m. Eastern, twenty hundred UTC. And if you desire to find out more, there is a URL available daily tech new show dot com slash live. We'll back tomorrow talking to how to plan your PC up PC upgrade with Will Smith. See you there. This show is part of the Frog Pants Network. Get more at frogpants.com. Prime and Club hopes you have enjoyed this program.