 Daily Tech News show is made possible by its listeners. Thanks to all of you, including Scott Hepburn, Bjorn Andre, and Jeff Wilkes. Coming up on DTNS, Apple wants NFL Sunday ticket and more, but the more might be too much for the NFL. Plus, Andrea Jones-Roy explains why scientists don't really know how AI works. It's a black box. This is the Daily Tech News for Tuesday, Tuesday. It's Tuesday. It says Tuesday. Tuesday, November 22nd, 2022 at Los Angeles, at job merit. From Columbus, Ohio, I'm Rob Denwood. And heart skipping a beat is the producer, Roger King, which is me. Roger at it right. It's all good. Sarah Lane helped produce today's show. She's not with us right now, but her fingers are all over it. We got Andrea Jones-Roy coming up. Let's start with the quick hits. Microsoft told The New York Times that on November 11th, it offered Sony a 10-year deal to keep the Call of Duty franchise on its PlayStation platform. Since announcing its intention to acquire Activision Blizzard, Microsoft has repeatedly said it plans to continue releasing Call of Duty titles for Sony's consoles. Access to the popular game franchise has frequently been brought up in regulatory reviews of the deal. Automatic CEO Matt Mullenwick said that Social Network Tumblr will soon support Activity Pub, the decentralized social networking protocol used by many so-called Fediverse apps like Mastodon. This would allow users on Mastodon or any other Activity Pub-supported app to follow a Tumblr account from their Mastodon instance and vice versa. Mullenwick said support would be added ASAP, but offered no specific date. So Mastodon uses Activity Pub to do the Fediverse. So if Automatic puts Tumblr on Activity Pub, would it then be part of the Mastodon Fediverse? Philosophers want to know. The UK's Competition of Markets Authority opened an in-depth Phase 2 investigation that will focus on the supply of mobile browsers in regards to Apple and Google's market dominance, as well as Apple's restrictions on cloud gaming services on mobile. The investigation could take up to 18 months to complete and then forever in the courts, but they're looking at the duopoly. That's the takeaway here. The markup reports that major tax filing services including H&R Block, Tax Act, and Tax Slayer transmit financial information to Meta through the MetaPixel. Data sent varied by company. Tax Act sent gross adjusted income, names, emails, and refund amount. This information was also sent to Google Analytics but omitted names. H&R Block sent health savings account usage and college tuition information. Tax Slayer used Facebook's advanced matching system to try and link site visitors to Facebook accounts, sending phone numbers, name of the filer, and dependent names. Intuit's TurboTec used the Pixel but only to sign in and send usernames at device sign in time. After being contacted by the markup, Tax Slayer removed the Pixel to evaluate its use. Intuit modified it to no longer send usernames and Tax Act no longer sent financial details. Oh, that's nice. It's nice that they're not doing that anymore. He got caught. Oops. Reuters sources say the People's Bank of China is readying to fine FinTech giant Ant Group more than a billion dollars, likely over alleged violations regarding disorderly expansion of capital. That's apparently a crime in China and resulting financial risk to the country. Ant Group has been under intense government scrutiny since it scuttled its 2022 IPO. The scrutiny began and then the IPO got scuttled. The fine may single that the crackdown is near its end. Usually when you write the check, then the problem is resolved. We'll see. All right. Oh, that's the end of the quick hits. Let's talk a little more about what's going on with Apple and the NFL, Rob. So the athletics sources say Apple is still the front-runner to get the rights to NFL Sunday ticket, but the negotiations are taking longer than expected. NFL Sunday ticket lets viewers pay to see all the NFL games happening, except those in their home market. Those are exclusive to the local CBS and Fox channels. Direct TV has had the product for decades, but the contract is up after this season and Apple, Amazon, ESPN and YouTube are all reportedly making bids. Now, Apple has been reported to be the front-runner for a while now, but supposedly what the athletic is saying is it wants rights that aren't in the current package that Direct TV has. And that is what's making the negotiations take a while to resolve. That probably includes things like VR and AR, right? Those wouldn't have been in the Direct TV agreement because there weren't any rights to give there. And it's a pretty good bet that Apple was going to come out with some kind of mixed reality headset early next year. So they might want to write that in. NFL Sunday ticket also doesn't include international streaming rights right now. So it might be that Apple wants to work that into the deal because Apple likes everything on Apple to be available worldwide. And Apple might want to get the rights to those local games, the so-called blackout games, like it did with its deal with Major League Soccer. Major League Soccer, all the games are going to be on Apple TV, whether they're in your local market or not. So they might be trying to get the NFL to give in on that, which would require Fox and CBS to weigh in, which seems like a difficult negotiation. And another thing complicating matters is that the NFL is selling its stake in NFL media, which operates the NFL website and the cable networks, NFL Network and Red Zone. Red Zone switches between games as teams get close to scoring. Last December, NFL Commissioner Roger Goodell told the Sports Business Journal that the NFL was months away from selling that stake. The NFL may not want Sunday ticket and the NFL media stake in the same hands, but Apple does. From an office sports reported in March that Apple wants to buy both Sunday ticket and that stake in NFL media. Yeah, that reminds me of what Netflix was saying about wanting to control all the aspects of a relationship with the sport. So we haven't heard Netflix involved in this one, but it does sound like Apple. Don't know, Rob, what's your best guess? Is Apple just trying to get as much as it can and it'll settle for what it can get? Or is this going to be an all or nothing thing? No, there's no question. Apple wants everything it can get. And we will see whether or not they, you know, what they're going to be willing to accept. But the NFL is one of the most powerful brands in America, if not the planet. And I think they're trying to be very careful that we don't want to give up too much of that power to another single entity. I mean, you know, to an even more powerful. They don't want to give too much juice to another company that already has a lot of juice. So I think that that's probably a big part of what's going on here. Yeah, I think Apple has got a lot of cash to wave around in front of the NFL. But the NFL is also a brand that has a lot of influence and sway and a lot of cash of its own. So maybe Apple's cash wave and doesn't work as well with the NFL as it does with an AI startup in Silicon Valley, which is probably dragging out these negotiations. That's a really good point. I like what I like the idea. We don't really know what's going on. This is all based on reports, but I like the idea that Apple would like to make it simple. That's a very Apple-y thing to do. And they have done it with Major League Soccer to great effect. You subscribe to Major League Soccer, you get all the games done simple. Don't have to figure out where you are if you're traveling or whatever. You just get the games. If they can pull that off with the NFL, I mean, that would be huge, right? It would be huge. The issue that I've seen with this is forever you've been able to watch your local team play with some exceptions. I mean, there definitely are some blackout dumbness that the NFL has. But for the most part, if you live where your team is, you can watch your teams play on Sunday afternoon or whatever day they're playing. That would change if now Apple controls this. That means that you now have to get a subscription to see even potentially your local games. It wouldn't have to. They've done a deal with Major League Soccer where some games can be broadcast. They could do a deal with the NFL that says, yeah, Fox and CBS can keep carrying the local games. That's fine. We just want to be able to stream them in that same market. So you have a choice. OK. Yeah. OK, so that's a choice. That's where Fox and CBS are going to say like, oh, hell no. We don't want anybody going to watch it on Apple TV when it should be us. In which case Apple would then have to come and say like, well, what if we let Nielsen credit you, the viewers that are on our stream for your local advertisements? And I don't know that Apple's going to want to do that. That's where the headbutting is. It's not that Apple wants you to stop the broadcast. It's that Apple wants to be able to stream it. Because Apple can work its own advertising. Just promote its own stuff. So yeah, this is it's taking longer than it should. And there's a reason why you take longer than it should. You're talking about two of the most recognizable brands on earth and neither one wants to allow the other to have too much. Well, Apple wants it. The NFL doesn't want to give it up. That's that's what I think a lot of this is, you know, how can we how can we make the money and also keep a bit of the power? Yeah, and it gets confusing, especially if you don't follow it closely, because there's blackouts and blackouts. There's the NFL tradition of a blackout because they didn't sell enough tickets, right? The Raiders historically never had their home games shown because they didn't sell enough tickets in Oakland back in the day. That's a different situation. And I don't think it happens very much if at all anymore, right? The blackout we're talking about here is if you have NFL Sunday ticket on Direct TV and I live in L.A. I can't watch the Chargers or the Rams on NFL Sunday ticket because they want me to watch Fox or CBS for those. But I can watch all the other games and then it's different. If you if you live in Columbus, I assume the Browns are going to be blacked out or maybe the Bengals too, right? But but but you're going to be able to watch all the Bengals are Indianapolis. You get a bunch of teams close here in Columbus, you know, just centrally where it sits. So yeah, this is it's interesting. But yeah, those those old we didn't sell enough tickets. The Browns used to get caught up with that when they played in the old Cleveland Municipal Stadium because that thing sat what 90,000 people or something like that. Infinite number of people. It's just it's like just I mean they were given tickets away to school kids. People walk over the street. Hey, come in and watch this. Just so our local market, you know, you know, can actually watch the game. So I'm glad that that's pretty much solved with smaller stadiums. It does feel like Apple's going to going to get something out of this. I doubt they walk away just because they can't buy. I mean, we didn't even talk about them buying NFL Network. I could see them taking that and just putting it on on Apple TV and saying NFL Network now available to everybody. That would be an interesting play. Red zone. Same way. I doubt they'll walk away from this if they don't get everything they want. I assume it's still worth it even if they just get the same deal as Direct TV. My guess is they'll be able to get more than what Direct TV had, even if it's just the mixed reality stuff. What would be the thing that you that you would like to see as an NFL fan? I don't have a problem with Apple getting it so long as I can easily subscribe to the stuff that I was able to easily subscribe to before. And Apple, they may want everything. I don't think they would walk away because I have to remember that there's a big company named Disney that is right behind them in the negotiation of this as well with ESPN. So there are other players, so they can't play too hard of hardball just because, okay, we'll just go ahead and do that deal with Disney. It was more advantageous to the NFL. So they've got to tether this really carefully to make sure they don't close themselves out by trying to have everything. Yeah. Apple could write the bigger check. I think that's why they're the front runner. Even the Amazon, they can write the bigger check. Certainly over Disney's ESPN or even Google's YouTube these days. So it's a matter of how much they can get. And that's actually a pretty good bargaining position. Would you like to take less from Disney or would you like to give us what we want? All right. There's lots of ways. If you got thoughts about this, let us know your favorite team and which streaming service you would like to have get the NFL. How can they folks get in touch with us on the socials? So feeling social, get in touch with the DTNS audience on socials at DTNS on Twitter, at Daily Tech News Show on TikTok and at DTNS pics on Instagram. Machine learning, deep learning. All these different kinds of algorithms out there that people call AI are used a lot more often these days. They're in your smart devices. They're in your smart homes. They're in your text to image generators if you're out there doing the Dolly. There's a rising concern though, it's been rising for a while about the fact that we really don't know how these algorithms get their answers. Recently, Chloe Zhang wrote a story called scientists increasingly can't explain how AI works. We asked Andrea Jones Roy, data scientist, comedian, circus performer and host of majoring in everything to join us again to talk about why that is and how big of a problem this might be. Andrea, thank you for doing this. Thank you so much for having me. I always love talking about problems with data science. So it's interesting, but it's very interesting. So thanks very much. We're happy to benefit from that being your thing. So what did you think of this article, first of all? So it's an excellent article. I'm really glad that it's out there and circulating around. As you said, it's not something that's a new problem or something that is new in terms of data scientist awareness of the black boxness. But I think we really cannot say enough out loud and as far and wide as we can that we do not know what's going on in the vast majority of these models. In some cases, that doesn't matter. But increasingly, as you said, if we are going to apply data science machine learning AI and we use those words, maybe problematically interchangeably as well. If we're going to be using all of that in our everyday lives, we really should have an understanding of what's going on in them and an understanding of the limitations of those models. And how those models might actually be incorrect and what that might mean for all of us. So I'm so glad that that it was written and I'm so glad that you're spotlighting it. Oh, thank you. It is something that I think is becoming more well known that it's a problem. If you can, is there an easy way for people to understand why that is? I think a lot of people say, well, you built the thing, why can't you tell how it works? Right. Well, I was thinking about this and one of the easiest ways in, I think, is to think about machine learning in contrast with statistics. And for now, I'll focus on machine learning and we can sort of leave aside that AI is a deeper deep learning or more complicated machine learning. But even a simple machine learning model, one of the things I say to my students is even those, we don't exactly know how they work. So if viewers remember their statistics class or even if you don't generally in old fashioned statistics, what we're doing is we're saying here's a set of variables that I think matter, right? Here are a couple of data points that I think might explain the GDP of a country, you know, the population, the this, the natural resources, all that good stuff. And in statistics, we the researchers specify that model, we say, I think this is going to be positively associated with GDP. And we build it out. And then we test it and estimate it and evaluate it. And we kind of do some diagnostics. Obviously I'm oversimplifying, but that's what we're doing. Sure. With machine learning, we're basically saying, here's what the ingredients are. This is sort of the simplest version of machine learning. We say, I'm going to take those exact same data points. I'm going to take GDP, population, natural resources, all that stuff. But I'm going to say, hey computer, you tell me how they fit together. I'm going to tell you which ones to look at. And then you find out what the patterns are that best explains that thing that I want to understand or predict, which is GDP. And in that case, even in this very simple model where we've been very clear about what the ingredients are, we maybe collected the ingredients, we can look at every data point. We only get output that tells me, okay, the computer says this is the estimated effect. The computer says this is the predicted value two years from now. Or if we increase natural resources by this amount, this is what we would expect the value to be in the future. And we don't really know from those models as easily as we do with statistics. We don't really know the why. We're very good at point predictions and we can evaluate how good our predictions are by waiting until the future arrives and then saying, hey, we got it right or we got it wrong. And there are other techniques we can use to test within the model, but we never walk away with a sense of like, ah, this is why it works. You really have to do a lot of gymnastics and machine learning to walk away with something like, now I understand why natural resources is a strong predictor of GDP. Or I have a sense of what the deeper underlying reason is for such and such playing a role. We tend not to get that. And in this particular example, it may not matter that much because if all I want to do is make some investments in some countries that I just care that I'm right. But if we're doing things like hiring algorithms, assigning scores to students, evaluating whether people are ready for parole, deciding whether or not someone is a reliable person to loan and make a mortgage out to, which are all things that companies use data science and machine learning for. And we don't know why they're making those decisions. We can't actually unpack whether or not those predictions or decisions are are right or fair or going to lead to longer term problems down the road. We just we just don't know. Tell me if I have this right. So in the first example, the not the not AI related example, I create a formula, a very complex formula where I'm like, I will put in these kinds of cows and this kind of grass, and it will tell me estimated milk production of the cows. And if it if it's far off, I made the math so I can look and go, oh, I waited that too much or I did this calculation wrong. Whereas with machine learning with deep learning, we're saying here are a bunch of cows. Here's a bunch of grass. We know the math is good at associating things, but it's not good, particularly at associating cows the way our original example was. So it just outputs stuff that looks right. But you can't look at the equations it uses because the equations are for associating things, not specifically for cows. And so you don't know how it got to its answer. That's basically the idea for statistics. We also don't necessarily go in and say, oh, I'm going to wait this 2x or 3x, you know, when we estimate, you know, a regression line, it can tell us that the slope should be two or three or whatever. So there's still, you know, an element of guessing there, but generally that's the idea. And if we think about, and you might say, well, in this simple example, I can probably work out backwards what the coefficient on cow and sheep and whatever should be. And sometimes we have like a settlers of Catan model or something like that. But you think about the kinds of things that we do for like facial recognition and various, you know, things in your home, you're using maybe thousands of variables. And there's just no way as a human to sit down and work out how all those variables are being combined and in which way like we could get the computer to spit out a whole bunch of coefficients. So that's probably an in deep learning. Basically what you're doing is you're doing one of those algorithms, and then you're using that result to feed into a new algorithm and then using that to feed into another sort of this nested machine learning situation. So you simply, the amount of time it would take for a researcher to tease out what any one prediction is doing would probably be, I don't know, maybe a full week of work. But that would be for every point prediction that you're making. Like it quickly becomes impossible and intractable, which is why we use computers to do it. And part of the reason data science, it's kind of like that. You just need a deep learning algorithm that can tease out why the deep learning algorithm did its thing, but then you'll need an algorithm to tease out that one. And yeah, yeah, yeah. Exactly. It's like the, I don't know who the philosopher is, but it's, you know, if the brain were simple enough for us to understand, we wouldn't be smart enough to understand it. Like that's kind of what's going on here is that what's cool is that it can do a lot more than humans can do as far as taking all these variables. You know, we have an intuitive sense of, yeah, I think you want more cows or more sheep. And the computer, it can look at all the sheep that have ever existed in the whole wide world and combine it with everything else that we decide to feed to it. And we just don't know why we got those outcomes. And as I said, in many cases, if what we're trying to do is predict, maybe we don't care another limitation on sort of the machine learning revolution that I talked to my students about in addition to the very profound biases as far as like who gets sentenced to how much time in prison and all of that which shows is very racially biased, for example, is, is we really don't walk away with much of an understanding and so regular statistics, call me old fashioned, but regular statistics, tends to be more useful for saying, well, like, in a more generalizable way, what are some of the mechanisms by which increasing cows in our country would give rise to GDP? I'm not saying you can't get that from machine learning, but you can almost, you can in many cases learn more about the world through a simpler model. And because it's simpler, we can then generalize it and actually think about policies in another way. I'm not saying one's better than the other, but you know, explanation and point prediction are two very different exercises. That's great. I mean, we haven't even gotten into so what do we do, we're going to have to have you back to talk about that because that's a whole topic of its own but would you say if you had to leave the listeners with with one thing, it's, it's that you should just be aware that these tools are useful but they're not perfect or something like that. That's pretty close to what I would say. I would say the number one thing and as I was reading the the excellent article and vice that that prompted this conversation, I think it's so important. They put it extremely well. It's just so important that we as humans realize that even though machine learning deep learning AI sounds incredibly sophisticated and it is very complicated. It's not the same thing as saying it's a correct or capital T truth interpretation of the world. And one of the things that I really get concerned about is when I work with companies and CEOs and really smart people who stare at data all the time and make really big decisions using data and using the output of machine learning variables is we've got to remember that all of these, all of these are wrong. I mean they're not useful but every prediction is an estimate. We the researchers the humans have chosen. I've still told the computer hey pay attention to cows and sheep as opposed to chickens and cars or whatever it is right and I can make that list really broad but I'm deciding what data to look at and I'm letting the computer do what it will with it. But that doesn't mean that I've actually gotten to a capital T truth answer because maybe I'm missing a whole bunch of data that I left out. Maybe the data itself is super biased. I over counted cows. I under counted chickens because most of them were underground. I don't know how chickens work right but I think the humility you know it sounds so technical and it is very technical and it can get very very very precise but that's not the same thing as being right and we've got to use our own human judgment and our own ability to say. This is prediction makes sense. What am I leaving out what am I over counting when I look at the results of my study am I systematically denying more people of a certain type. Loans from the bank is that a problem can we look into it and so it really you know machine learning outputs are information and they're useful but they're not correct. So algorithms are like people basically some of them are good but nobody's perfect. I love that. Andrea thank you so much for helping us understand this if folks want to find more of what you do where should they go. I am at Jones Roy J. O. N. E. S. R. O. O. Y. On all the social medias and at Jones Roy dot com. Fantastic thank you so much. Thanks for having me. Real quick. I wanted to correct myself earlier I said that the Ant Group scuttled its IPO in 2022. That would be this year they did it back in 2020 so sorry for the verbal typo. Rob you ready to check the mail bag. Let's do it. We got an email from Will who says I'm a 40 year old father and I was lukewarm about 3D programming until I saw Beauty and the Beast in 3D. It was like a moving pop up book and was incredibly entertaining. Animation is incredible in 3D and for that alone I was super sad it didn't catch on better. I'd be thrilled if it came back and would pay a hefty premium for it in Disney Plus in particular if they offered that content. Thanks again for the wonderful work. He isn't the only one who emailed us about 3D TV. No David sent a letter in and says as someone who bought a 3D TV during the 2010s craze and enthusiastically tried to use it. The Sandra Bullock movie Gravity was fantastically immersive in 3D and so were sports like football soccer to the rest of the world. Formula 1 antennas I can tell you what I think killed the 3D. Twitter. If you recall those were also the heydays of Twitter when we were all talking about the second screen. We were all live commenting on sports events as we watched them and you just can't do that with 3D glasses on. I remember filling torn technology or community in Twitter 1. But now that Twitter is dying maybe 3D can make a comeback. Maybe greatly exaggerated but good one good one David fact that you want to look down at something else kept the glasses off your face which is which is a good one. And then we also got this one regarding the chicken and egg problem with content for 3D televisions. There's already a driving force for new 3D content VR virtual reality is the perfect medium for 3D movies shows sports events etc. And by targeting VR platforms content providers would at the same time target 3D TVs. Finally I think the biggest question is what sort of glasses free technology will be used. If it's something like a light field display then we're not talking about just 3D televisions but full on holographic displays and those would be awesome. That is also a very good point. Thank you Sergei for sending that one along. Rob you weren't here yesterday when we were talking about this. What do you think you ready for convenient glasses free affordable 3D television or do you not care. Well those are all three good descriptors of what it has to be convenient affordable and I'm not wearing a helmet to just watch television. Just just just not going to happen. So I think that you know that this email that was talking about VR that might be the place where 3D really starts to take off because you know granted those headsets are not going to be inexpensive but they're not going to be as expensive as a you know as a 85 inch you know 3D television and it doesn't require you to wear glasses or headset. So I think that it's just I just don't know the technology has caught up to any other part is that you have to have all of the you know all of the content has to be created in 3D. Last thing I wanted to have this really expensive thing that I watched 3D content on and there's no 3D content to watch on it. It's another good reason why Sergey may be right either VR or AR or mixed reality right that could do both because then then that addresses David's point where well I can look at mastodon Twitter Facebook whatever in my mixed reality glasses I don't have to take them off to look at I can get both my 3D television and the mixed reality stuff in the same place and then if the content isn't available 3D. I don't feel like I wasted it because I'm using the headset for other things too right. Absolutely like I said it's you know like these things are not inexpensive. So if you if there's multi purpose for it. I think that you know you probably win there. I could see buying you know a 3D headset much more quickly than buying a 3D television that does not require a 3D headset to wear and watch with it. Well thanks again to Andrea Jones Roy for talking to us about AI earlier in the black box. If you want that vice article be sure to head to DailyTechNewShow.com and get the link. It's a good article and thank you Rob. Great having you along. Tell folks what you've got going on these days. I'm just a couple of other tech podcast so you can find me always over on SMR podcast at smrpodcast.com and my newest show you know the tech John at thetechjohn.com that's the tech J A W N excellent. I want to give a big thank you to our brand new boss William. William just entered the patreon. Thank you William for backing us at patreon.com slash D T N S extending the ability of this show to continue to serve you and it's down to William and every other single patron who helps us. We give you value. We give you value. You give us a little value back and we keep going for almost nine years now. So thank you William. Welcome to the fold and stick around William because as a patron you get the extended show Good Day Internet. You can also catch the show live Monday through Friday 4pm Eastern 2100 UTC. Find out more at DailyTechNewShow.com slash live. We're talking tech layoffs tomorrow with Megan Maroney who just experienced one herself and Scott Johnson will be here too. Talk to you then. This show is part of the frog pants network. Get more at frog pants dot com. Diamond Club hopes you have enjoyed this program.