 Daily Tech News show is made possible by its listeners, thanks to all of you including Miranda Janell, Justin Zellers and Pepper Geesey. Coming up on DTNS SoundCloud bets on algorithms to help you find better music, vehicle data could make us all safer, and do we focus too much on bias in AI? This is the Daily Tech News for Wednesday. May the fourth be with you. 2022 in Los Angeles, I'm Tom Merritt. And from Studio Redwood, I'm Sarah Lane. In Salt Lake City, I'm Scott Johnson. And I'm the show's producer, Roger Chang. We are going to talk about a hot button issue. It's not Twitter. Actually, we will talk about Twitter. But we're going to talk about bias in AI. But first, a few tech things you should know. Samsung's memory manufacturing is arguably more successful than its phones. The company just announced Universal Flash Storage 4.0, which can achieve 23.2 gigabits per second per lane. 30 would be better, but 23 is still pretty good, which is double what UFS 3.1 could do. This helps memory keep up with increasing throughput that comes from faster internet service like 5G. It's also 46% more power efficient than 3.1, so you get faster storage and also better battery life. And the 1 terabyte UFS 4.0 module is a hair smaller than the 512 gigabyte UFS 3.1 module. So you might say, sounds great. When do I get it? Samsung will begin mass production in Q3 of this year, meaning it could start showing up in phones by the end of the year and certainly in early 2023. The company also introduced pro-endurance microSD cards aimed at dashcams that can record continuously for 16 years. The previous max was five years without stopping, so quite a jump. No kidding. Here's my 16 years of dashcam footage. Please review it. Let's clear out all the Elon Musk buying Twitter stories right now. This shouldn't take long, so you don't even have to skip ahead too far if you're trying to avoid it. The chair of the UK Parliament's Digital, Culture, Media and Sport Committee, Julian Knight, has invited Musk to testify about his plans for Twitter. I guess he won't go, but he got invited. Musk has been talking to private investors who might go in on the deal with him, including existing shareholders that might be able to maintain their stake. He also says that he might take Twitter public again three years after he takes it private, which he plans to do on the close of the deal. A lot of companies have done that. Dell has done that, and he suggested he might charge for government and commercial use. There you go. That's it. That's the tweet. AMD plans to release a new gaming laptop CPU line code named Dragonrange designed for high power machines with a 55 watt TDP and up. There are 20 millimeter thick laptops that probably need to be plugged in most of the time. It's also AMD's shot at capturing the top of the laptop gaming market, arguably already dominating the midrange and below. AMD also announced a lower power Phoenix line of chips for thinner machines at 35 to 45 watts of power. Both will be part of the Ryzen 7000 series based on the Zen 4 architecture and will arrive in 2023. AMD will give more details on them before then, so the company sets. In fact, AMD said that next quarter it will start reporting gaming as its own financial segment. So expect more details on all of this at the June Analyst Day. All right. We have continually covered the U.S. dispute between mobile carriers and the Federal Aviation Administration and airlines over the deployment of C-Band 5G wireless. We have a whole episode of Know a Little More about it if you want to dig in. The second version is that while C-Band wireless does not operate in the same bandwidth as airplane altimeters, which are very important for landing safely, especially in fog, some older altimeters don't properly filter out all the C-Band transmission. And the fight has been over whose responsibility that is. Should they replace their old altimeters or should the carriers modify how they broadcast around airports? Right now, carriers are biting by restrictions near airports, but that goes away July 2022. Reuters reports that the FAA is meeting Wednesday with telecom and airline officials to push to retrofit and eventually replace altimeters that might face interference from C-Band 5G. The U.S. nonprofit National Bureau of Economic Research studied usage of CHIVO, that's El Salvador's national Bitcoin wallet. Since it launched last September, around half of Salvadorans have downloaded the wallet, which came with a $30 incentive, you know, just to get you started. Of those who downloaded it though, 61% have stopped using it. Remaining users tend to be unbanked people who use it to hold and transfer dollars. Yeah, so maybe not the glowing success some people would have liked. But there you go. That's the update. All right. We got some car news, Scott. Well, I had an experience last week where I got to get an up close and personal experience with a Rivian truck in the famous gear tunnel. And one of the things I learned about it that I didn't know is how much data this thing collects. It's always taking pictures, always taking video, always recording data. So these, we know this is already happening. They're out on the road. They're collecting data maybe more than ever before. And now we have a brand new effort. What is it? Yeah, GM. GM is partnering with a company called InRix to collect data from vehicles for U.S. transportation officials to analyze the data would be formatted for the transportation department's Safe Streets and Roads for All program, and would include pedestrian and construction worker data. So it's trying to prevent pedestrian and construction worker deaths. It could also be used by local governments for road safety projects, and also for Vision Zero, an effort to reduce fatalities and serious injuries on the road. I don't, it doesn't feel like it's the first time I've heard of this though. It's not like the first effort of its kind. No, there's a bunch of these. Honda Research Institute, USA announced a pilot program in December that would use data from cameras and GPS in its cars to help monitor road conditions and identify repairs for local transportation departments. That pilot program is taking place in Ohio. In December 2020, Europe's Data for Road Safety Initiative announced a long term deployment of a safety related traffic information system that would integrate data from cars and traffic systems. To create warnings for drivers of dangerous conditions. And the International Transport Forum runs the International Road Traffic and Accident Database, ERTOD, IRTAD, that contains a validated crash and exposure data from 32 countries, including the United States, going back to 1970. And it makes that data available for safety agencies to use as well. I feel like a lot of people will go, well, okay, we all want construction workers and anybody who might be crossing the road when cars are going by to be as safe as possible. But if you're collecting the data of how I was exceeding the speed limit for the last three days, is it going to come back to me in some way? I think that's where some of this, like, hmm, what are we really doing when this data comes in? Yeah, to me, I don't know. I just probably not a fair comparison, but it feels like browser data going back some years where the big concern was, well, with this new advent of cookies back when that was a new thing. And these ideas that you can kind of be tracked across the site you're on, as well as when you left, where did you go? And this sort of thing made a lot of people go, well, I don't want to be followed around. I don't want people to know everywhere that I'm going, whether or not you're up to something skeety or weird or whatever it doesn't matter. I just don't want people following me, especially, you know, government or big companies or whatever. And I feel like this is a little like that. I don't want to call it fear, but whatever that feeling of like easy now, don't, you know, don't give every little piece of data away. Because some people maybe don't want someone else to know that they, I don't know, they take a certain road on their commute or whatever the reasons may be. I can't help but feel the parallels from that, from that conversation for some reason. Yeah, and that's why it's important to, this is a great example of data collection does not immediately mean privacy violation. And in fact, data collection doesn't immediately mean ad targeting, which we all kind of jumped to the conclusion. Sometimes it's not about surveillance, it's about safety. Safety. This is important data to have to literally save people's lives, especially construction workers. People just kind of forget how much risk those construction workers are at when they're out doing highway work. So it's important, you guys are absolutely right, it's important that this data be privacy first. Most of these efforts have good privacy policies in place. And all of them at least say, and I haven't looked to see if they're audited, but they say that they are only aggregating data incidents. They're not tying it to like a license plate number or even a VIN or anything like that. They're just saying like this, you know, these amount of cars passed this place at this speed, that sort of stuff there. Yeah, there'll always be some percentage of people who just say, well, I don't believe them when they tell me that. And that's fine, whatever. There's always going to be that. But personally, I think data is important in almost every aspect of life these days. And I think this kind of data would go a long way to not just help us ease into the, you know, more energy efficient electric car era that we're slowly entering into or automation. I mean, these things are going to benefit from that data, of course, but just good old fashioned. How's that stretch of the highway performing data is really interesting and not only interesting, but saves lives, saves money, saves a lot of headache. So I want my car to tell a database that will use that knowledge that this road needs work. And this road is full of dangerous potholes. You know, this road is crumbling like, yeah, let me opt in, I guess, but do that and integrate it more. Let's use organizations like ERTOD to compile this and make it available to everybody who can use it. Transportation agencies, car companies, road repair, etc. Music streaming services are a convenient way to listen to music you want to listen to. I think we all agree on that. But sometimes it's really difficult to find good new stuff. That's always the trick. How good is the recommendation? Sounds like SoundCloud's got an idea, Sarah. Indeed, yeah. So I'm one of those people who I tend to kind of listen to the same stuff over and over because discovery can be overwhelming to me. And I don't always get what I want if I'm just kind of trying new music unless it was specifically recommended by a friend, etc. SoundCloud has acquired a company called Museo. That's M-U-S-I-I-O. It's an audio-AI company that can listen to new music and identify what makes a hit a hit. SoundCloud says it wants to use Museo to sort through its library of largely amateur music. Not everything on SoundCloud is amateur music, but there's a lot of it. And also help people discover things that they might love but might not find on their own. Because Museo's tools can sift through hours of music, countless hours in fact, and then pick up the songs that have certain patterns and characteristics that correlate with being popular. This is a bit of a math project, really. Yeah, and they're not the only ones doing it, right? Who else is out there doing this? Yeah, so Music Distribution Platform Tunecore announced back in February that it's partnering with LAB's music startup Fuego. That's F-W-A-Y-G-E-O. Get it? It also uses AI to match listeners with creators. So it's a little bit different, but the idea is to get a creator to be able to have a deeper connection with somebody who really likes the music that would not have found them otherwise. Tunecore is a division of global digital music company called Believe. It's based in Paris, France. Fuego is a mobile music discovery platform which helps independent artists increase their exposure, also their earnings potentially, and they launched in beta last year. Along the same lines, DIY music distributor DistroKid, it's been around for some time now, has an AI bot already in place that bot is named Dave. And it reviews tracks and ranks qualities like danceability, speechiness. Now you might say, well, okay, what does all that mean? Results, at least, as far as I checked it out, a little varied. So for example, Dave's review of the song, if you want to build a legacy, you're going to have to break some legs. That is by an artist called Madison Minor. I included 3% speechiness, meaning the song probably doesn't have talking. I listened to the song. It's like a heavy metal song. So there was a lot of yelling, screaming. You might say, oh, okay, that's not rapping. You know, maybe there's some nuances there. But yeah, DistroKid invites people who have submitted music on its platform already to submit to Dave and see where people fall in the percentages of what is this music, how good is it to me based on what I want to listen to right now? One takeaway I have from just the recent years of streaming music, wherever the source is, that they all do okay with their algorithms and that approach. Spotify, Apple Music and others seem to do okay in that regard. But I'm a little surprised this announcement or this, let's say these movements are happening in the smaller, what we would consider the smaller areas of music recommendation on the internet. It isn't Spotify announcing a big AI initiative to improve their searchability or findability. It's not Apple announcing it either. It's coming from smaller places. I wonder if we'll see the big boys adopt a similar plan, something with AI. They just don't make as big of a deal about it, right? SoundCloud had to acquire a company. Spotify and Apple have both bought AI companies in the past. I think we're just less surprised and we don't talk about those as much because we expect those companies to buy AI. And we already know Spotify uses algorithms to recommend stuff. I think what's distinctive about this is SoundCloud is saying not just discovery. Jerdstar in our chat said, I don't think math solves music discovery well. It's a subjective issue. We don't just want more of the same. But that's not what SoundCloud's doing. They are looking through and saying, there's a lot of bad music here. Sorry guys, but there is. We can narrow it down for you and say, these are most like the kinds of music that succeed. And then still leave it up to you to subjectively decide which ones you like. And that's a little different than Spotify, which is trying to say, oh, well, we know you like this. Let me give you more like that. Yeah. Also, this is, you know, just to be clear for all of us here. Yes, this can be helpful for discovery, but this is also really helpful for artists who already have a presence who are trying to figure out, oh, who's the next person to collab with or a manager. There's a lot of business, you know, senses of this that make a lot of sense because SoundCloud is a place where I mean, you know, many artists have kind of made a name for themselves. There are a lot of artists who obviously are trying to make a name for themselves who can't. And it's not because they're not talented. It's because there's just a lot of junk in there. Yeah, you're back to the discoverability problem. And it happens if every medium podcastings this way, certainly music. It's getting, I think to the point where streaming services have enough original content that they're struggling with discoverability of their newer shows if they're not, you know, marketing them right or whatever. So, you know, can we get the next weekend or the next post Malone or the next whoever to spring out of SoundCloud the way they have in the past, probably, but they were also among a fewer base then they had some advantages in that regard. And now those those places are full of really talented people. How do they get noticed? I think this is the track. This is the way you get it. It's like a diagnostic assistant, right? The doctor still determines whether you have the disease or not. But the AI can help narrow it down where to look, right? And give an advice of like, yep, might be that. And the doctor signs off SoundCloud is right now having to just kind of look into finding things. I'm sure people at SoundCloud go through tons of bad songs. They're like, Nope, Nope, Nope. Before they finally hit one like, Oh, that was pretty good. We can recommend that on the homepage. So this is going to make that job easier. Narrow it down so they're only listening to the likely candidates, which would theoretically make their homepage recommendations and their other recommendations better because they can now choose a wider variety because they have more to choose from. Whereas before they wasted a lot of time listening to things they were never going to highlight. It would be cool if they broke some real ground here so we could see that some of these ideas bleed into other mediums. I think podcasting in particular is really hard to get noticed if you're new. But even old timers like us, we do a new show. A lot of times we're counting on our current base to know about it and to spread the word. Very hard to get noticed for new people. So these kinds of initiatives I think are going to be really important for user-created stuff moving forward. Well, we have a little more to talk about on the ethics of using these kind of algorithms in a second. But first, I want your ideas. If you have ideas about hosts and guests from other podcasts and shows that we should have on this show, let us know. Check out our guest survey and put in your recommendations at dailytechnewshow.com. Well, in 2021 there were more than 400,000 research papers published on subjects relating to artificial intelligence. And that number continues to rise. As the work expands, so do the concerns around it. Ethics and bias in the field have led to multiple controversies. At Google, for example, and certainly elsewhere. We're among the outlets who have emphasized the importance of accounting for human bias in training data. It happens. How do we identify it? And how do we make it better? Protocol has an interview out today called Stop Worrying About AI Bias All The Time. What is going on with that? Yeah, no, it's not great. It's like, we're, like you said, we're always saying like, gotta be careful of bias. Here's a person saying, don't worry about it. It's not exactly what she's saying, though. It's an interview with Bina Aminath, Executive Director of the Global Deloitte AI Institute, and head of Trustworthy AI and Ethical Tech. You should read it. It's an excellent read over a protocol doing great work. Aminath knows what she's talking about. She's a database and SQL developer. She's worked in data science and AI for more than 10 years at Bank of America, GE, Hewlett Packard, and then joined Deloitte in 2019 and just wrote a book called Trustworthy AI. In that book, she writes the following. What passes for public scrutiny is too often just seductive clickbait headlines that fret over AI bias and point to a discreet use case. There's a lot of noise on AI ethics and trust, and it does not move us closer to clarity or consensus on how we keep trust in AI, commensurate with its power. Alright, so if you're someone who says, well, shouldn't we be concerned about AI bias? What direction is she going in? She's pointing out that it's not that there isn't bias or that we shouldn't be concerned about it. It's that we only think of bias in the cases where it matters. And there are lots of cases out there where other ethical concerns are important and those are getting overlooked. For example, she points out predicting jet engine failure or wind turbine power generation. Reliability, security, and safety are more important in those cases. She even uses a medical example practicing, you could say like, oh, well, the medicine, it should always be about bias. Well, whether an MRI machine might fail or not, bias might play a part, but it's probably not the most important part. If human data isn't involved in a decision, bias probably doesn't play an important part. AI ethics also need to account for situations where other concerns might outweigh bias concerns even when bias is a big concern. And she uses the example of identifying potential human trafficking victims as a place where there might be bias in your data set, but that bias might be tolerable if that algorithm is better at identifying victims of human trafficking over not using it. It's a question anyway. And I think what she's trying to say is not, don't take account of bias, but don't assume that's the only problem. So she's not saying that bias is a problem, just that bias needs to be, you need to identify it. And in certain situations it's helpful and in others it is not. Yeah. She wants the public conversation to not identify bias as the only problem. We need to add to the conversation. So she recommends that ethical practices should be part of every employee's training. If you're using AI in your business, the intern should know about it just as much as the data scientist, because every employee is going to use the tools that the data scientists create. The interns in HR is going to use that weed out algorithm that the data scientists created for HR, so they need to know what questions to ask and everything like that. So it strikes me that the concern here might be, this is my average brain speaking here, so I'm happy to be corrected on this, but it seems like the bias in AI is something that, let's say, lay folks, people who don't understand this stuff or haven't really paid a huge attention to it, they're going to be the ones with the least amount of information about what this even means, like what does it even mean to say an AI has bias? I feel like there's a big opportunity for education kind of across the board, and that's what she's doing. This is good, right? This is a step in that direction, certainly. But so that people understand the more and more our world gets wrapped up in automated systems and AI driven systems that we have as clear understanding as possible about what that actually means so that we can back off of the nerves about it a little bit. That seems like that's always kind of part of this problem. This is such advanced stuff. Whether we're talking about vehicle automation or a million other tech advancements, having a voice assistant in every room, this sort of thing, these scare people who haven't really been filled in. And so some of that is on them because they're just not bothering to fill themselves in. But this one feels like a complicated one, and I hope there's more of this, more talking about it, and less saying, well, it's AI, no one can understand. This is really high end stuff. And then everyone feels like, well, wait a minute, you can't do that. I need to know. I mean, it's built by humans. Right. It's not that hard to understand. I agree. There's a nuance to what she's saying, which I think gets lost in our simplified, 280 character conversations out there, which is, there's bias in AI. Well, then we should stop using it, right? If that's the extent of the conversation, you're going to come to the wrong conclusion because as she said, we can use these algorithms to predict jet engine failure. We can use them to stop human traffic. We use them for a lot of very good things. We just need to be responsible about how we use them. And that doesn't mean, oh, just ignore the bias. It means fix the bias and also make sure that you're paying attention to all of the ethical concerns and the other things you do as well. I think that's worth paying attention to. Yeah. It's not the same as saying we can't have uranium based decorations at this restaurant. No, we're going to go back and figure out how we can do it. This isn't that. This is an alterable, tweakable, look at it from a different perspective, find something to go, oh, that is kind of weird. And we don't want that bias in there and go and tweak it again. Like I got to have that attitude, I think. You have a lot of places with uranium based decorations. No, I don't know why I thought of that. In what world? An example we can all identify with. I mean, if you're really a fan of uranium and you're like, yeah, you're not had a uranium party lately. They're really fun. I just run in the wrong circles. Apparently. Hey, wait, you're in my circles. Never mind. Oh, yeah. Google plus for out a little liquor. Okay. So US based movie studios and streaming services. This is not funny. Have largely shut down in Russia and are no longer supplying their content to the country. And that's in protest over the war in Ukraine. We all know this. However, many folks in Russia were still able to go see the Batman. And it wasn't because Warner Brothers or DC changed a policy or because theaters there broke the law. So what's going on here? Dan Campos has an explainer. Hello, DTNS crew. Movie theaters in Russia continue to release films like the Batman. Don't look up or turn in red. Thanks to the use of Torrents. While the locates by Hollywood studios or streaming services organizations such as the Association of Cinema Owners in Russia issue the statement to indicate that the block threatened to crush their industry. At the end of March, Russia approved an act known as the priority action plan to ensure the development of the Russian economy in conditions of pressure from external sanctions, where it eliminates fines for violating copyright laws in digital assets such as movies, music, software or video games. For more information about this, listen every day to Noticias de Tecnología Express, the Duolingo of Tech Podcasts. I love the Duolingo of Tech Podcasts. Go sign up for NTX and brush up on your Spanish. Or if you already speak Spanish, it's going to be perfect for you. I had heard about this law, but I'm glad Dan brought this back to our attention to be like, yeah, they are in fact using it. It's not just a law they put on the books. They are showing the Batman in theaters. And yeah, it's crazy to think that they went to the extent to change copyright law because copyright law feels like that immutable thing that even the theater owners might be like, well, don't change it because if you open that door, then piracy is going to get worse. That's a weird and crazy thing to think about. Yeah. And some people, you just need the Batman. I get it, man. You've got to have the Batman. So I'm not saying this is good. I'm not saying anybody has anything. I'm just saying the Batman. Sometimes you need him. Yeah. I mean, it's just, you know, great responsibility. I don't know. Yeah. It's an illegal bat signal. But it's not the movie they deserve. There you go. That's great. God. But he's the movie they pirate it. All right. Let's check out the mailbag. Let's do it. Okay. So, I'm going to go back to the code and this was in response to our conversation Monday. Where, where Rich declared his love for Yen's like men. Mattia says, Yen's like men will never go stale. Great deep dive into the music scene and why artists are increasingly releasing older products. Of course the labels have been doing this for years with greatest hits compilations. But the cynic in me wants to point out remixes. Put out a new product and increase your revenue. Ultimately, the only correct response to this conversation is Hans shot first. Very nice. The real quick note on that. It's not that different than, you know, what we used to call reruns back in the 80s. And I'm watching the fall guy for some reason. Well, no, it's different from reruns. It's like if they went and reshot the entire show. That's a better way. Yeah. And I wouldn't mind that. I'm not saying that the fall guys totally out of date, but it's pretty out of date. So. Up your game. Isn't that what we get though? Everything gets reimagined. Yeah. And you know that thing. More than ever. Yeah. Things are being reimagined quite often. In fact. Yeah. Who knows how many shelved scripts there are for the fall guy movie. I know they exist. They there. I don't know the last time. Anyone watched back to the future, but that's one of my favorite scenes where the kids like, what's a rerun? And he's like, Oh, never mind. Yeah. Back to that because it's not a rerun to watch for a lot of kids to watch friends or the office. It's the first time through. Yeah. Exactly. Yeah. Well, thanks to you, Scott Johnson for being the best person ever. Let folks know where they keep up with everything else that you do. Yeah. Take that every other day's guests. You guys are the worst. Just kidding. You guys are all great. Yeah. There's a lot going on. If you listen to the show, there's a good chance you like daily content. I have to figure. And so as a result, I thought today I would just mention that I have a daily morning show that I do with my good friend, Brian Ibbitt. In fact, Tom's on Wednesday as we does a tech segment before we all come together here today for this. And that happens every week. So if you are looking for some good old fashion, fun morning entertainment that doesn't sound like every morning show you've ever heard. Come check it out. We're over at frogpants.com slash TMS. It's the morning stream. If you're searching for it for your, for your podcast apps or whatever. So again, that is frogpants.com slash TMS. And for everything else, you can find me on Twitter at Scott Johnson. Excellent. We also want to thank our brand new bosses, Christopher and George just started backing us on Patreon. Give a warm welcome to Christopher and George. Very warm. That's warm. I don't know. No, I didn't hit the wrong button. This is an actual audience that we have here. And they blew the air horn. So yeah. There's a longer version of the show called good day internet. We'll probably laugh about this and lots of other things available at patreon.com slash D T N S for all the live viewers and those after roles right after D T N S. But just a reminder, we're live Monday through Friday at four PM Eastern 20 hundred UTC on D T N S. And you can find out more at daily tech news show.com slash live. We are back doing it all again tomorrow because that's what we do on weekdays. Just remember young movie joining us. Talk to you then. This show is part of the frog pants network. Get more at frogpants.com.