Data, Drones, and Digital Mascots – Episode 32 of Unbuffered の全文書き起こし

state of Pennsylvania. So Christopher Mitchell (05:09) Right, and you can you can read this story on CommunityNetworks.org by the way. So Sascha, ⁓ how do you react to this just knowing what you've done in Pennsylvania and ⁓ and knowing the world out there that is bereft of decent data? Sascha Meinrath | X-Lab (05:23) Yeah, yeah. A and it and it's important to celebrate these leaps forward. I'll start there, right? It's like foreshadowing, right? Which is to say I think it's great that Vermont is doing this and they absolutely should. And it's essential, right? If you're if you view things like BEAD as an intervention, and I think we can all agree that it's meant to be an intervention, then knowing where things were before and where things are after the intervention would be kind of a fundamental and essential series of data collection efforts to see how eff effective, how efficacious the intervention was. Without that Christopher Mitchell (05:59) Right. Speed tests don't have to be perfect. They can still tell us a lot. Sascha Meinrath | X-Lab (06:03) That's correct. That's correct. So here we are, we've we're we're investing tens of billions of dollars and we don't have baseline data really and nobody's thinking about like, well, what's the evaluation here? How do we know if we succeeded? And how do we know where there's differential success? And how do we know like which business models and intervention types work best and what is the level of intervention achieved? Is there a difference between say satellite or wireless or fiber? I think yes, physics says yes. But like, you know, that we can test that hypothesis empirically. And to do that you kind of need no one pay attention. Christopher Mitchell (06:41) And then no one can no one will pay attention. That is the inevitable conclusion to this. I agree with you, Sascha. Those are answers that we should have. But once we have them and we see that Vermont is doing much better than many other states because of this approach that they've developed to develop, ⁓ you know, not just embracing fiber, but embracing community ownership of it. ⁓ you know, I'm I am despairing that no one will bother to go back and say, wow. Sascha Meinrath | X-Lab (06:48) Yeah. Christopher Mitchell (07:09) Vermont and New Hampshire, but they're so different. Why is that? No, no one's gonna care. They're just gonna be like, Yeah, New Hampshire sucks. Like Sascha Meinrath | X-Lab (07:15) Yeah, so I'm more optimistic. I think people have rightfully tuned out assessments that are not meaningful or contrary to their lived realities. Right? And so it's important to tease apart. I think people care a lot when they find really hard hitting information and data that says like, yeah, in fact this thing that you've been told doesn't exist. This digital divide that we've declared is non-existent and all of our data show that your lived reality is wrong. Sean Gonsalves (07:25) Mm-hmm. Sascha Meinrath | X-Lab (07:44) Well then yeah, people don't pay attention. But time and again what I see is that when you have meaningful, scientifically grounded assessments that in fact properly capture the lived reality of people all around the country that they pay a lot of attention. And again, I think that's one of the reasons why our government has worked so diligently to prevent meaningful assayance and assessment. That everyone that I've ever talked with in rural America understands the digital divide. And yet that lived reality is officially papered over. And into that is why it's so important that Vermont and other places do these independent, you know, assessments of what what's actually going on on the ground. Now, the caveat that I started with is really based not on wanting to poo-poo anything that Vermont is doing, but to point out like Sean Gonsalves (08:18) Mm. What Sascha Meinrath | X-Lab (08:42) It would be even better if we had open methodologies and open data. Because again, one of the problems that we have seen, and we see this with the national broadband map itself, is that like there's a lot of valuable data in the national broadband map that we can't access. And the reason for that, I would say, is because were we to be at access, we would do better assessments. It would have a much more hard-hitting analysis that people would pay attention to and they would say, Hey, FCC. What the heck are you doing? Hey, NTAA, why are you sucking so badly in fixing these problems that your own data document? And so, please. Christopher Mitchell (09:19) Right. Let me hit let me hit let me hit up on this for a second because and then Sean, I know you had a question, but just based on that, Sascha, it's important for people to understand that this data is being collected with Ookla. And so if for instance we wanted to write a grant to say, give us a bunch of money that we need in order to do this, first of all, we'd have to ask for an extra $30 or $50,000 just to write a check to Ookla to get access to the data. And then in getting access to the data, we would have to sign a contract that says we would not do anything with that data that could do any number of things, which includes stuff that would make an individual ISP look bad. And so we literally not only is there a cost barrier, but we are literally contractually obligated not to do certain research questions if we wanted to and if we had the money. So like that's a pretty big deal. Now, Sean, what were you going to ask about? Sean Gonsalves (10:10) Mm-hmm. I wasn't gonna ask. I was just gonna comment just to your point earlier that ⁓ you know, to your point about you know, no no one will pay attention, no one will care. Like I get that sort of like in the aggregate, but one thing I do think people care about is whether or not they're getting what they pay for. So that that is fundamental to me. And so that's what these speed tests I think represent sort of the a measure of accountability of of you know, it's not I'm not just

accepting on faith that the, you know. Sascha Meinrath | X-Lab (10:10) Yeah. Sean Gonsalves (10:36) 500 megabits that I'm paying for a month, I'm actually getting a and and people want to verify that. And so I I so that this is why I like things like this is that it it it sets up a a structure where there's some there can be ⁓ some level of accountability, which is a separate question as to whether or not you're gonna take this data in aggregate and and refine maps and get a better sense of where, you know, places should be and you know, there should be more investment in or something like that. But just on a basic consumer level, it's it's it's worth knowing whether or not people are getting what they pa what they're paying for. Christopher Mitchell (11:11) Sascha, are we just No, no. So I wanted to I wanted to ask Sascha. So you know, I I just talked, let me just give it a clean break from when Sean started. so Sascha, you talked about open research, open methodologies. Is there anything else that besides my criticism of using Ookla as a partner, what else are we missing that if another state was like, you know what, we we're we just had a big election and we have a new mandate. Sascha Meinrath | X-Lab (11:11) And all that is before we get into pricing. Christopher Mitchell (11:39) ⁓ we're gonna do this thing. What should they be doing as opposed to just replicating Vermont's partial successes here? Sascha Meinrath | X-Lab (11:45) Yeah, so Ookla measures a lot of really valuable elements, a lot of very valuable factors and variables that go into the broadband experience. And people often mistake, they're like, ⁓ so you're for measurement lab. And I'm like, Yeah, I'm for measurement lab. They're like, So hate Ookla? I'm like, No, no, I would use both. And it's crazy not to use both, right? if I'm like, okay, I'm a doctor and I have one diagnostic and that's it, I'm like, I'm probably gonna miss some stuff. So obviously, you want these extensible frameworks for collecting information. It's not even that Ookla and M-Lab are going to be the end all be all either, like saying, I got two diagnostics and now I'm good. no. You want to build measurement frameworks that have baseline standardized data collection and extensibility that can add in additional modules to look for certain phenomena that allow me to diagnose specific problems to my locality or my time. period or what have you. ⁓ and the reason why I'm against say using the Ookla tool is if that is the foundation for everything. It turns out that's a walled garden. You never get this extensible framework. You're always stuck in Ookla world. And as good as it is, it is extremely limited. And over time, and we already talked about the licensure, but over time it really diminishes one's ability Sean Gonsalves (12:59) Mm. Sascha Meinrath | X-Lab (13:13) To ascertain what fundamentally can be improved in my network, if it's not already part of the Ookla framework, you can now measure it via that. And so I would take a big step back and say, okay, let's build a fundamental, a foundation for layering multiple different things, of which Ookla absolutely should be one of the verticals. But it should have a lot of areas for people to put in. new exciting things. And again, like I I'll bring up pricing because I think this is one of these elements is just completely missing. It's been expunged from broadband. So we're like, how do we understand the digital divide? I'm like, well we gotta understand whether connectivity exists. And everyone agrees that. Like that's important. It's like what about price? Yeah, price is the second most important criteria on whether people adopt broadband. Well what do we collect? Well we collect existence poorly and pricing not at all. Sean Gonsalves (14:02) Mm-hmm. Right. Sascha Meinrath | X-Lab (14:10) So now let's diagnose the digital divide. I'm like, that you're gonna be very limited in what you can do given those limit those problems. Sean Gonsalves (14:17) Well a and as they say, if it's not affordable, it's not accessible. Christopher Mitchell (14:21) Well, let me let me ask if do you remember back in the day, twenty years ago, Connect Kentucky was like passing encouraging states to pass registr legislation in which a non profit would be able to get state funds to engage in mapping and data collection activities. And ⁓ they framed it that way to make sure that they were only ⁓ competing against a very small group of entities because there was very few nonprofits that did that work at the time. ⁓ Sascha Meinrath | X-Lab (14:21) Hundred percent. Huh, yeah. Christopher Mitchell (14:47) If a state wanted to f was like, you know what, we'd like to do that, but we don't have the capacity in house, how are we gonna do it? I don't know. Is X-Lab available for work along those lines, Sascha? Sascha Meinrath | X-Lab (14:59) yeah, ⁓ but our biases are towards open data, open methodologies, open technologies. ⁓ it's weird that that's the radical position. I go to states, I'm like, hey, how about we build everything so that you can control it and you own it and you can do whatever you want with it. And states are like, I don't know, I think I want the black box that I have no control over that totally disempowers me and I want it to cost ten times as much as it'll cost to hire X-Lab. But it's not just X-Lab, right? X-Lab is one purveyor in an open Sean Gonsalves (15:18) Yeah. Sascha Meinrath | X-Lab (15:28) scientific research community that has been consistently sidelined, even though we literally built and pioneered the technologies that Ookla is now ⁓ commercializing. M-Lab figured this out. And long story short, you can have better, cheaper data, you can have more extensible frameworks, you can have, in essence, and it you don't need to come to me to do this, but we need a critical mass of people to fund these open alternatives at a level I would guess it's probably 10, maybe 20% what we're pouring into the

proprietary systems that are worse for us. And Christopher Mitchell (16:08) And I'll just I I do want to note ⁓ Sascha is not just speaking hypothetically. You've received money to build a system. I've used the system and that system did multiple tests because the tests are free effectively. And so for every test that you did at measurement labs, you did a nuclear test. You filled their database up for them. Sascha Meinrath | X-Lab (16:25) Correct. That's 100%. Again, like I don't wish them ill. I just want to point out that there's these severe limitations that cost us extra to have less. I'm like, well, that's that feels like American healthcare, maybe. But like it's it's a bad deal for states to go the route of yet another proprietary platform where you're like, man, how do we navigate these licensure limitations? I'm like, well, use a platform that doesn't have them would seem an obvious choice. And so you have this real oddity in America, right? We've spent over $500 million on broadband mapping and have very little to show for it. And we have been able to show consistently that an effort that would cost less than 10% of that would be far better and you would be more empowered. Christopher Mitchell (17:22) No, no, you're missing the point, Sascha. And this goes back to was it March of twenty twenty when the pandemic was just setting in and President Trump said if you stop testing, we won't have a problem. And so, Sascha, I tell you that we do not have any problems with Internet access because we do not do any testing. We don't collect pricing. We just announced that, hey, we've solved it. And and that is a deliberate policy choice from ⁓ a whole community of people. So that's where we are. Sascha Meinrath | X-Lab (17:24) Yeah. Sean Gonsalves (17:34) That's right. Sascha Meinrath | X-Lab (17:35) Right. That's right, that's right. No, I I remember arguing with ⁓ I won't name but like some FCC ⁓ senior FCC officials about the then Pennsylvania map, which showed universal connectivity. And I was like, That's wrong. And they're like, We've seen no evidence of that. And I was like, I want you to come with me to rural Pennsylvania, stand up in front of a community there and declare this. Right? Just j just come come with I'll pay for the trip. And they're like, No, no, no. I was like, so you know Christopher Mitchell (18:09) Ha ha. Sean Gonsalves (18:10) Should we submit that? ⁓ Christopher Mitchell (18:16) Mm-hmm. Yep. Okay. Sascha Meinrath | X-Lab (18:19) The FCC knows that it's its data is wrong. ⁓ but it took independent efforts to expose this. And the thing is, after a while, I mean, we've all been through the ringer of the challenge processes, right? And it's like, well that that piece of data is wrong in the in the national broadband map. As long as the ISP consistently lies, you get overruled. And I was like, That's just a bad adjudication process. Like that's just Sean Gonsalves (18:20) Okay. Mm. Mm. Christopher Mitchell (18:45) Yes. That

and ⁓ and the idea that if someone claims to have service in an area and doesn't, you need to find one of their customers to demonstrate that their service doesn't work, only they don't have any customers, so you can't find anyone to test. It's like a very it's a very Joseph Heller kind of ⁓ policy. I do. Sean Gonsalves (19:00) Yeah. Sascha Meinrath | X-Lab (19:02) Yeah. Yeah. Sean Gonsalves (19:04) Yeah, yeah, that's what Sascha Meinrath | X-Lab (19:06) Yeah. I mean remember barrier free, right? These are the guys that turned in data showing that they had universal gig connectivity across six states. And the FCC was like, That that scans, right? They just ingr ingressed it and it took Derek Turner over at Free Press to say, That's wrong. Like you you've got Christopher Mitchell (19:25) Yeah, no, we I remember I remember separately, we before we knew that Derek had turned it in, we also reached out because ⁓ I remember there's the the woman on my team that was our our mapping person, ⁓ Hannah. ⁓ and this is one of the many Hannah. We had three Hannah's on the staff at that point. And ⁓ one of them was like, Would you look at this? All right. I'm gonna I'm gonna draw this to a close though. Sascha Meinrath | X-Lab (19:31) Yeah. Sean Gonsalves (19:41) Mm-hmm. Sascha Meinrath | X-Lab (19:46) Right. And it it like face validity. It f it fails face validity. You nobody has any state in the country with universal gig connectivity. Period. So the fact that the FCC completely just was like, yeah, that's that's legitimate data. And we'll put it in our national broadband map. And it will take an extra effort by an independent group of nonprofits to say, no, no, no, that that's That's too wrong, right? Because they never corrected the mechanisms by which these overstatements could be sent in. They were just like, okay, okay, okay. In the case that you are too hyperbolically wrong, and somebody petitions us, we reserve the right to remove that one data point, not to look for other mechanisms by which wrong. That is just a broken methodology through and through. Christopher Mitchell (20:40) Yeah. Yep. Okay. We are going to move on though. And this is the one that I think is gonna be quicker, which is I got this question today. ⁓ interest rates. So the interest rates are going up, and boy, I I don't think they're gonna go in a different direction anytime soon. I don't know, maybe the bubble bursts will see them go down briefly, but ⁓ I actually think that there's larger issues that are gonna keep driving them up. And I have this question of whether ⁓ higher interest rates meant that cities were gonna be scared off from building municipal fiber networks because of the cost of borrowing. And so I don't know, Sean or Sascha, I had a I had a response because I've been thinking about this for a while, but I'm curious what you guys came up what you guys come up with with five seconds of warning. Sean Gonsalves (21:24) Well, as as everyone knows, I I am a scholar of

municipal finance. No, seriously, I have no idea. Although it does seem like, yeah, it will, because already, you know, communities get sticker shocked when they hear, you know, about the cost of borrowing and wanting to do this kind of stuff. So so yes, I think, you know, it would lend itself to putting another quiver in the in the whatever you call it. It ⁓ arrow in the quiver. Thank you. Man. An arrow in the quiver. Sascha Meinrath | X-Lab (21:24) yes. Christopher Mitchell (21:47) An arrow in the quiver. Come on, man, think Hawkeye. Sean Gonsalves (21:52) ⁓ it it in in in saying, yep, see, this is why we can't do it and we're just gonna have to hope and pray that ⁓ you know, one of the big cable companies just, you know, does the right thing. Christopher Mitchell (22:03) Sascha Sascha Meinrath | X-Lab (22:03) Yeah. Yeah, a hundred percent yes, ⁓ for debt serviced capital expenditures, right? And and that's across the board, right? So people will buy fewer houses, fewer cars, fewer municipal broadband networks, like anything that's gonna be deficit serviced, ⁓ and deficit funded is gonna be harder to come by. Same with the US national debt, which is to say like, you know, when interest rates go up. borrowing goes down and that's kind of a general truism. ⁓ we can talk about the hyperinflation problem there at some point. But like long story short, that is certainly going to be the case, but it won't just be broadband. Broadband will be one of a whole portfolio of public investments that if deficit funded is going to diminish. Period. Christopher Mitchell (22:56) That's spoken. Both of you are snowflakes. You're you're young. You're naive. ⁓ Sean, when did you finance your first car? I'm curious. Do you remember what year? Sean Gonsalves (23:06) Model A Ford, maybe. ⁓ no, I don't I no, ⁓ I don't remember. ⁓ in I don't. Not at not at all 'cause this was back in this was it would have been back it would have been back in the nineties. So I I I have I don't remember at all. No. Christopher Mitchell (23:11) You don't remember the interest rate was? I'm curious. Sascha Meinrath | X-Lab (23:15) He had a musk ox followed by a you know mule team, so Christopher Mitchell (23:23) Yeah, so that's what I'm curious about. that's the thing, is that ⁓ I people like me, you know, and Sascha and and Sean to some extent ⁓ came of age in a time where right now everyone thinks 5.8%, you know, ⁓ is a high interest rate. It's not. It's a perfectly normal interest rate. It's just that we've been spoiled from 20 years of abnormally, bizarrely low interest rates that we will probably never see again. ⁓ I mean, I would say let's hope. Sean Gonsalves (23:44) Mm. Mm-hmm. Christopher Mitchell (23:51) We never see again because that would be a real sign that something's gone wrong if we get them that low again. ⁓ and so I think there might be some short term, but like the thing is is just that like it makes a lot more sense for private companies to gamble on building networks when the interest rates are low. And as the interest rates get higher, I think it is harder. And so private equity needs to find a better return. And and so I think that what that means is that all in all, we may see cities more likely. To build because we will see fewer private builds. I think we are hindering heading into this period of consolidation. And so I do think it would be harder for cities to justify. They may have to do more work to explain why. And that may actually result in fewer because we don't have as many people that are willing to do that, perhaps. ⁓ I mean, this is this is not something where it's like directly related to the interest rate. The real question is how many places do we have citizens that want to like mount campaigns and like make this a priority? Sean Gonsalves (24:27) That's a good point. Mm-hmm. Right, right. That's true. Christopher Mitchell (24:48) Whether that's for, you know, like a neighborhood or whether it's for the entire city. But like ultimately, the interest rate is not extreme. We've just been spoiled for a long time. And in the nineteen seventies and nineteen eighties, the idea of building right now seemed like this is a great interest rate. So everything's relative. And I just want I wanted to throw that out there. But like for people that are used to like two and three percent, yeah, those days are gone, I think. Sean Gonsalves (25:11) Yeah, well, I mean it's a good point that you raised and and and it could very well be the case. I just ⁓ you know, one of the things that I I get ⁓ frustrated about in in talking about, you know, financing municipal bills and whatnot is that very rarely does it take into like it's it's it's always sort of looked at in this this narrow sort of accounting, how much does it cost? What's the good you know, what's the debt. as opposed to raising other questions related to it, which is like w w ⁓ what's the cost of not doing this? And and Christopher Mitchell (25:42) Yeah, how much are we spending on that like those kids going to school who are less effective and then they're dragging down other people because they can't do their homework and they're having to ask more questions and get more attention? Sean Gonsalves (25:48) Right. Sascha Meinrath | X-Lab (25:53) Well, as Sean mentions, I mean the opportunity costs are very measurable. We don't measure them, and therefore the lack of expenditure seems like a cost savings and therefore the benefit of the bargain. If you ever say, hey, look, we're gonna save twenty billion dollars, but it's gonna cost us fifty billion dollars in increased r tax revenue. Well, it doesn't take a genius to like, so your benefit of the bargains costing the treasury thirty billion dollars, like that's not good. That's a that's a bad Sean Gonsalves (26:04) Right. Exactly. Sascha Meinrath | X-Lab (26:21) That's a false economy. Bad mathematics. ⁓ that's in essence what the lack of broadband build in the United States translates to. A

massive Christopher Mitchell (26:30) Let's just be very clear about what you just said, Sascha, so people get it, right? You if if if I could break into what you just said, you said we're gonna cut the amount of money we're spending on broadband, and that's a good thing. That's what that's what NTIA is saying, Arielle, Arielle Roth with her savings that they're talking about. And what you're saying is that ⁓ they're not spending that money means Medicare, Medicaid, the Department of Veterans Affairs, all of those services are gonna be less efficient. Sean Gonsalves (26:36) Mm-hmm. Christopher Mitchell (26:56) People are gonna have worse outcomes. We're gonna have more veteran suicides from lack of telemental health and all those other things. And and when you take those into account, we are not saving money. We will spend more money and people's lives will be worse. Both of those things are not savings. Sascha Meinrath | X-Lab (27:13) And and in fact cost us quite a bit more over time. I mean, and w you know, America's pioneered a lot of a lot of the understanding around this, right? It's why we pioneered public education. We understood, hey, look, if we invest in this, it comes back to the nation state, right? And and to fail to appreciate that a critical resource like broadband brings more into the country than it costs to build as a country. or a municipality is really kind of myopic. What's that? Doesn't have to, but it should. We could still screw this up, that's true. Yes. Christopher Mitchell (27:44) It doesn't have to, but it should. Doesn't have to, but it should. Sean Gonsalves (27:53) Ha ha ha. Christopher Mitchell (27:55) All right. I do want to I don't want to move back over to Starlink then. We're running out of time. ⁓ Starlink shut off. I mean, so we were preparing for this. Like Sascha was trying to explain what happened. I was wait, do we actually know that this would happen with like with in now it's been several weeks? ⁓ but ⁓ Ukraine, ⁓ and and I'll let me just say that some of this stuff might turn out not to be entirely correct. I'll be curious to see what actually happens. But it does seem that that Elon Musk threw a switch and Ukraine was not able to use the Starlink slash Star Shield system on their drones, which are used both for offensive and defensive capabilities that were traveling faster than a certain threshold of like 100 kilometers an hour. If they went faster than 160 kilometers an hour, 100 miles an hour, then it stopped working, which meant that they were no longer able to intercept. Russian attacks and Russian missiles and Russian jet drones that are being developed now. And it also seems that Russia seems to have known this because they launched an extra heavy barrage after that happened. And now and people died. And and Elon and and Donald Trump, President Trump, said both of them were Yeah, we don't know what happened there. And then they started both sort of taking saying Elon did it. Sascha Meinrath | X-Lab (29:00) Immediately, that night. Christopher Mitchell (29:15) Only one of the things that Elon said was that he couldn't make decisions about Star Shield that had to come from the the chain of command within the the secret the the Department of Defense. And so, like, I don't know, there's a bunch of stuff that, like, just the statements that have been made that are not true, which is not the first time. Like, of course, Elon Musk lies about everything all the time. And so we don't actually know all of what's going on. But the thing that Sascha brought up, which I want to I wanna come back to you on, Sascha, is that It is profoundly disturbing that it seems like one person could decide whether or not Ukraine can stop Russian attacks on its cities and electric infrastructure. Sascha Meinrath | X-Lab (29:49) Correct. Correct. And and you know, again, you have this confluence of civilian and military utilization of SpaceX assets. It's very clear that these two networks are directly linked in their architecting. And it does appear from you know the best available n info that we have that somehow. A civilian outside of the military chain of command was able to directly degrade the capabilities of a military communications infrastructure. And I and I think that is v well worth investigating. Because if true, I'm like, what is the US military thinking? A, but B, like, I'm pretty sure if you or I decided to screw with military communications. there would be a dark van outside our house the very next day, and we would just never see the light of day again, because to do that is downright treasonous. Right? To aid your enemy, and it doesn't matter what Trump thinks, because Congress passed the law about this. To give aid to your enemy is the very definition of treason, and yet. in polite DC circles, this is not a question that is even being asked, much less investigated, as to what actually transpired here. What what actually happened that led to, it appears, the degradation of our military communications infrastructure. Sean Gonsalves (31:34) Yeah. Yeah. And well, and then yeah, I mean to and so to me, related, you know, in in the grand scheme of things with Starlink and some of these questions. I mean, there is this question about, you know, the the the the military use and how it comes into play with UK Ukraine Russia, but I feel like you know it's there's something much bigger going on too. You know, the Gizmodo had the story that I just recently came across on this research about this Israeli ⁓ ad firm, ⁓ Discover Tech, where you know it's it's It's potentially and and and l let's be clear, they haven't found like the smoking gun, but there seems to be an association based on the research that they found about how these ads are bought online. And this is the spyware that where you don't even have to click on anything for it to infect a device. And the the research is is is raising a question and suggesting that it may have ⁓ you know affected

Starlink devices in a hundred and something like over a hundred countries in And primarily in places Argentina, the Philippines, Japan, ⁓ Syria, Libya, Yemen, South Sudan. I mean, th this kind of stuff, it's like it you know, it goes way beyond ⁓ just, you know, ⁓ telecommunication and and and folks having access to the Internet. And you know, it it it in fact the the research that's suggesting this this link between this this spy f this spyware stuff was enough to have for Google to cut off Discover Tech's access to its ad exchanges and for IAB Europe. To remove it from their global global vendor lists. And so there seems to be some, you know, kind of where there's smoke, there's fire here. And it's it's it's it's kind of stunning that all this is going on. And and and the and the thing of it is it it all relates back to this question of like trust. And it's it's almost like big tech is like hellbent on destroying trust. And and and and which I don't know, it that seems like a good way to destroy a business, but you know, it's like the arrogance. Christopher Mitchell (33:21) Yes. No, no, it's how you build a business. Like, this is the key. I was I was waiting to like figure out how I could weigh in on this, Sean. But like the connection of this to our work, I think the best explanation that I ever heard for big tech, like when you think about what is big tech, big tech takes a business and says, How can I make other people pay my costs? And then they can that's called scaling, right? Like, how do I how do I scale this up? Scale this up means I don't have to take questions anymore, right? So Sean Gonsalves (33:33) Mm. Christopher Mitchell (33:53) Starlink's number one reason, as far as I'm concerned, for why more people don't use it. Terrible customer service. No customer service, right? And like that, there's a lot of people in ⁓ rural areas who would like to use it, I think, and they don't because of that. And there's other factors as well that drive people away, but that's some big one. But like they're talking about launching a mobile phone service. You know who's not gonna have customer service? Them. Like, this is this is how they work, Facebook, etc. Sean Gonsalves (34:12) Mm-hmm. Mm. Christopher Mitchell (34:21) These companies exist at that scale because they have said we're gonna screw over our users. If something goes wrong, they cannot reach us for help. And this is this is what when they talk about scaling the business, that's the big part of big tech. It is about figuring out how to not be able to like pay the costs of having a human being sitting at a desk somewhere, answering the phone and dealing with the consequences of a product that doesn't work well. And it comes into this ad tech stuff, I think, where Starlink is just asking this question of. Sean Gonsalves (34:36) Mm-hmm. Christopher Mitchell (34:51) What is the minimum viable product to have a really cool connection? But we're not going to think about how to harden it. We're not going to make, we're not going to ask all the questions that have traditionally gone into telecom, which is is it going to work all the time? Is it going to work when someone's depending on it in an emergency and all these other sorts of things? If you, if you just ignore all of that, it's a lot easier to scale up, go big, and then run your run your rivals out of the market. And so we see this across big tech, Sascha. Sascha Meinrath | X-Lab (35:19) Yeah, no, that that that's exactly it. And and but it's also predicated upon never having anywhere else to go. Right. So you're gonna have Starlink, SpaceX, etc., running the tables for at least the next eighteen, twenty four months, but Kifan, Gua Wei, like there's these giant Chinese mega constellations that are gonna be coming online between now and say twenty thirty. At which point we're gonna be told like you can't use them. Sean Gonsalves (35:25) Right. Sascha Meinrath | X-Lab (35:46) And we're like, well, why can't I use them? They're like, it's g they're surveilling you. I'm like, and you guys aren't? Like who's like what and you know, can I just VPN through? Like that seems about as secure as encryption is these days. Like, why would I I I I suspect that's right. They they they're they're not answering the questions that I have about their network architecture, but I can't Sean Gonsalves (35:50) Yeah. Christopher Mitchell (35:52) Ha ha ha. I I think that's what Star Shield is. I mean, Star Shield uses the same satellites. I think they just have a higher level of of user terminal. Sascha Meinrath | X-Lab (36:11) I I there's not enough satellite launches to make for a second mega constellation. ⁓ but long story short, you know, we've we saw this with like TikTok. We've already seen like Deep Seek, the Chinese open weighted models where we're like the US government has declared nobody in the US government can use them anymore. Like we're gonna see this with satellite. Like more and more there's Sean Gonsalves (36:21) Mm-hmm. Christopher Mitchell (36:32) R I think that's the right thing to do by the way. I mean I think they're not always wrong is just what I'm gonna throw out there. I'm Sascha Meinrath | X-Lab (36:37) My point is that more and more places where you're like, hey, ⁓ we don't like you having options to use a functional equivalent that we don't control because dangerous. And we're never tackling again the under ⁓ undergirding ailment, which is we should be making this kind of warrantless surveillance an illegal business practice. And then we wouldn't have to worry about all of these things. I want to also mention, you know, like Sean Gonsalves (36:46) Right. Sascha Meinrath | X-Lab (37:05) Israeli firms have a very long and terrible history of kind of driving these privacy invasive root kits for telecommunications. Ron Diebert and the whole Citizens Lab folks up in Toronto have been documenting at least a dozen years worth of the next, the next, the next terrible system

that's targeting civil society and et cetera. This latest iteration, very real as it is, is just one of a whole assortment of products that like when North Korea does it, we rightfully say, like, no, that's illegal. When Israel does it, we sort of are like, ⁓ well, you know, boys will be boys. I think again, we need clarity of what laws and rule of law looks like in a space where now, through AI mediated development, these things are going to proliferate at an unprofitable. Sean Gonsalves (37:33) Mm. Hey. Sascha Meinrath | X-Lab (38:02) precedented rate. And again, we have purposefully refused to adopt international treaty rights around these kinds of interventions, place them, I think, where they should be, which is like along with chemical, biological, radiological weaponry. ⁓ we just have turned a blind eye, and that will haunt us for generation. Christopher Mitchell (38:25) Well, I think the answer there is for people to have a takeaway, not just being ⁓ upset, but open methodology, like you know, open research, ⁓ having a clear set of rules that apply to everyone, having good privacy standards that are enforced. I would say having a significant public investment into research along these lines that is consistent with an open democratic society, to develop these tools. ⁓ these are all things that we should be thinking about. Sascha Meinrath | X-Lab (38:53) I love it. You're like, I'm for science and consent of the governed. Yeah, I I yeah, a hundred percent, man. Sean Gonsalves (38:58) No doubt. Christopher Mitchell (38:59) I have I have one last question though. Sean, What the Tech?! Sean Gonsalves (39:04) All right. What the Tech!? So let so let check this out. So have you two ever heard of the phrase baby schema? Exactly. All right, good, good, good. So I don't speak German, but I learned it comes from the German word kindenschema, something like that. ⁓ and it's a concept that comes from the famed ethologist Konrad Lorenz, who proposed it like back in like 1943. The idea is that. Christopher Mitchell (39:12) No. And I haven't looked it up while we were talking either. Sean Gonsalves (39:31) Infant-like features trigger this like instinctive caregiving like affection response in adults. And it's the route that Meta and these other AI companies are taking. Baby schema. You've seen Meta's AI agent, that little cute thing that you know can answer email and buy groceries and coordinate plans with friends. So according to Forbes magazine, ⁓ Meta's adorable AI agent got 10 million users in just over a month, and they named that that that Laboo Boo ⁓ mascot jolly other now other companies are doing the same thing. Open AI's got these colorful mascots for its dots agent. Space X AI's got the Grok bot series. So I so I'm saying so why now? So polling shows that 20 of Americans are con are are are only only 20 of Americans are confident in big tech companies. 67 according to Pew magazine look at Zuckerberg unfavorably. Remember Meta's AI glasses, d

the backlash thing not too long ago, like unsolicited filming of women? Yeah, exactly. Perv glasses. And you know, Gen Z kind of pushing back, calling this kind of stuff AI maxing and whatnot. And so it's all in this context of like this wider resistance to data centers and whatnot. So when you think about it, AI agents need way more access to than than a chat bot. You know, they need your email, your calendar, your contacts, permission to act on your behalf. And this cute little mascot. Christopher Mitchell (40:31) Herv glasses. Sean Gonsalves (40:55) Is trying to mask that reality when in fact Jolly and his AI cousins are just this friendly face for this extreme data collection. One recent poll found that ⁓ a that ads with mascots in distinctive characters are six times more likely to be remembered. But I'm just gonna say this: I'm with Microsoft's AI's Mustafa Suleiman, who warned us about treating AI as human-like. So I'm gonna so I'm gonna be so I'm gonna say it. Maybe keep some of the bathwater, but throw out this baby. Christopher Mitchell (41:29) I was thinking that you were gonna say that we should force all the AI companies to adopt the paperclip as the mascot because that will drive down utilization. Sean Gonsalves (41:36) All right, that's a little happy though. Sascha Meinrath | X-Lab (41:38) Well, I just like the idea that if all the AI looks like Ren and Stimpy were saved, right? Sean Gonsalves (41:43) Yeah. Christopher Mitchell (41:44) What's the plan for done this evening? Sascha Meinrath | X-Lab (41:47) Make it really ugly. Christopher Mitchell (41:50) I do I would do want to ask a question, Sascha, out of the blue, which is what are you using AI for lately? And which AI are you using or are you not? Sascha Meinrath | X-Lab (41:59) gosh. Well, so ⁓ for those that are using Android, ⁓ you know, they've just basically made it so that any form of non completely privacy invasive AI is impossible. So I just removed mine completely now. So now I'm running Dicio, which is a local based phone AI that allows me some of the functionality, a lot of the functionality that I actually used and liked, ⁓ and I'm learning it. There's all these skills that basically you can download and and what have you. for my personal utilization, you know, I I I have an oddity, which is to say I run something called the Center for Agentedic AI research. So we have, you know, a few millions of dollars worth of infrastructure that we get to build things with. So I'm using a highly bespoke model running ⁓ some fusion. Sean Gonsalves (42:29) Mm-hmm. Sascha Meinrath | X-Lab (42:55) behind the scenes to build AIs that co combine several AIs and then I'm running that within a fabric called Pi or Pi fabric, I should say. ⁓ that's a lot of words that basically say like I'm not using off the shelf and I I I would not recommend using off the shelf. But the other big thing that I'm seeing is that off the shelf and bespoke AI are now bifurcating pretty rapidly. That our access to knowledge Is being curated through the AI. So that when I ask the same prompt in off the shelf AI, I'm getting systematically worse answers and access to less information that the AI has determined you shouldn't have access to. And I've challenged it on a lot of this, whether it's copyright, I'm a professor, so I have a fair use right to use copyrighted materials in my class. AI doesn't care. And when I argue with it, yeah. Christopher Mitchell (43:50) No let me ask you a clarifying question. Are you talking when you say off the shelf, do you mean Gemini, Claude, Copilot, Chat GPT, those? Okay. Sascha Meinrath | X-Lab (43:57) all the ones that people know of, I'm not talking about going to Hugging Face and downloading, you know, some open-weighted models which are often functionally equivalent and give you better results. What we're seeing is that just like we s like just like a sensor that the AIs are doing that, but all behind the scenes. You don't even see it. And when you know that it's happening and start interrogating and say, hey, were you programmed to not allow me my fair use rights. It's like, no, I just had ethical concerns. And I'm like, you did not. You you're you were programmed. And and it will argue with you. Be no, no, I just I I felt that, you know, there could be to the point that it was arguing with me a series of legal precedents backing up why I did not have the right to this. But unfortunately I knew the legal precedents that said I did. So when I challenged the AI, I was like, Sean Gonsalves (44:29) Mm-hmm. Sascha Meinrath | X-Lab (44:52) Yes, okay, that's correct. That is the precedent, but and then it would default back to I have ethical concerns about, you know, I just I feel I that I don't want to help you do this thing. That again, the FCC's already ruled I have a legal right to legal information online. So what does it mean when the AI is directly, explicitly, verifiably, unrelentingly Removing my right to legal material online. It would seem on the face of it that this is a very clear violation of the FCC's mandate. And yet when it comes to AI, we're like, you know, it's just magical. So don't ask, don't tell, don't don't interrogate, don't hold these companies to account. Long story short, the companies undergirding all of these AIs are now actively censoring. US access to legal information online at a rate that, you know, probably makes the biggest censors on in the planet are dwarfed by this. To what end? To the end of their corporate ⁓ priorities. So I am certain that there are deals being hit all over the place. ⁓ you know, there was just a recent Christopher Mitchell (45:58) But what to what end? To what end? Sascha Meinrath | X-Lab (46:16) federal court ruling, the Daniels case that said, I, the general public, have access and a right to access the addresses of my elected officials. This was contended in court. The federal court said yes you do. Didn't matter. AI within days of that court ruling

started eliminating access to the addresses of our elected officials from the results that it said it could access. And when you Christopher Mitchell (46:42) I mean, coming to you from Minnesota, I'll just say I'm not super sympathetic on that one in particular. Sascha Meinrath | X-Lab (46:47) No, I get I mean, it is absolutely a piece of knowledge that has been used for terrible ends. There's no question about that. The point is, if you have a right to something and some th something's running and purposefully removing that right, in this case access to information, ⁓ you know, again, if it happens in Russia or North Korea or China or Iran, we just call it censorship. And these are the biggest sensors on the planet right now. They dwarf anything else that we see. And it's measurable and verifiable. And it's not even part of the conversation of anyone that I've talked with in DC thus far. Christopher Mitchell (47:29) No, I hadn't heard anyone talking about that. Although my my initial reaction is that we're heading to a world. So if you remember, Chat GPT was never gonna have ads in the results. Now they have ads in the results. I presume we're next going to the yeah, I could get that information for you, Chris. It's just extra two ninety nine a month or two ninety nine a week, or like, you know, like it's gonna be like, you want the sports scores? Well, that's an extra like three ninety nine, you know. So that's that's probably I mean, like some of it is is ⁓ like you said, Sascha, for their their business ends. Sean Gonsalves (47:46) Right. Mm-hmm. Mm-hmm. Christopher Mitchell (47:59) Like, you know, these are companies that have to justify an insane amount of money over time. And I feel like, ⁓ that's one of the things that we'll see is that the models will become much less useful unless you get that the three add-ons. Sascha Meinrath | X-Lab (48:04) Yeah. Well, let me let me give you the the reality now that is gonna blow your mind. A lot of the open-weighted models out of China have freer access to information than the US off-the-shelf models. They censor the crap out of TNMN Square and all this other stuff, no doubt. ⁓ 100%. Throw that into Deep Seek and you can see it for yourself. But like the open-weighted obliterated models out of China are way freer than the US models from Anthropic and Google. Sean Gonsalves (48:11) Mm-hmm. Mm. Christopher Mitchell (48:21) Yeah, no. I'm I'm yeah. Yeah, I've actually been playing with ⁓ Quen three, ⁓ and trying to get a sense of Sascha Meinrath | X-Lab (48:41) Yeah, Quen's, yeah. Quent Quen ask it about TNMN though. Quen would be like, nope. Yeah, it it still senses that, unless you're using an open weighted obliterated model. But like Quen again is an example of how these other models coming out of China actually have some really powerful features and functionality ⁓ that I anticipate we are going to be forbidden from using in the United States. Christopher Mitchell (48:47) All right. Yeah. I mean it's gonna be interesting to see what's coming next. I was just reminded and we'll wrap up here in a second, but I was just reminded that like it's not that long ago when we were told that like massive AI driven secure cybersecurity events would were not gonna happen. And ⁓ and now it's gotten to the point where people are like, Yeah, just be ready for one week and so Sascha Meinrath | X-Lab (49:34) Yeah. But we're also being sort of We're we're being misled in many ways. So is as long as we're looking for like the Skynet-esque Terminator like, you know, existential threat, that we're not gonna pay attention to the everyday pickpocketing, the shiv to the back that AI actually is one hundred percent doing right now, which is to say, yes, I you know, there is a non-zero chance of an existential threat from AI, it is very low. And it's very low just because we're just not Christopher Mitchell (49:53) Mm-hmm. Sascha Meinrath | X-Lab (50:07) There's no big kill switch on humanity. It's difficult to do. ⁓ But it's guaranteed with 100% probability that AI is going to screw you in a thousand little ways. You talked about like bots and not being able to reach a human and all like that's what it's doing. But you may allow that everyday frustration because you're really worried about like the big existential threat. And I think that's the era that we're in right now. We are. Sean Gonsalves (50:23) Mm-hmm. Sascha Meinrath | X-Lab (50:34) kind of grinning and bearing all these petty, petulant AI, right, that just won't solve your problem or screws you just a little bit, but you're like, well, at least it didn't like throw a terminator at me. ⁓ it creates this sort of acceptance of the everyday banalities of evil that are AI. Christopher Mitchell (50:55) This is this is what I was predicting actually was that the ⁓ the era of it's ⁓ the we're we're having a backlog, we're you know, we have too many callers and you have to wait twenty-five minutes on hold before you talk to someone, that's about to disappear. We're not gonna be waiting on hold. More not gonna be talking to anyone either. We're gonna be talking to an agent and it's gonna get real frustrating real quick. Sascha Meinrath | X-Lab (51:17) Yeah, I I just tried to to to return something to Amazon, a company some of you've heard of, right? Like it was nightmarishly difficult to get the AI to accept that it should accept my return. It clearly has been instrumented to waste a certain amount of my time. And I'm like I'm an expert in AI, and I'm just like, how do I get this thing to acknowledge that I have a right to do this? And what finally did it was like Christopher Mitchell (51:25) Mm-hmm. Sean Gonsalves (51:37) ⁓ Sascha Meinrath | X-Lab (51:46) Yeah, basically just told it like, I have my council here and we want to know like do we need to bring lawsuit? And was like, Boom And I'm like, that's that's terrible Like Sean Gonsalves (51:52) Yeah. Mm. Christopher Mitchell (51:55) I I will say that I also

just ran into something like this, which will be entertaining for listeners perhaps, which is ⁓ my son hit me at a moment where I wasn't paying as much attention. He wanted to order something online. And so like my my wife or I do that for him. And he's like, Here's the website. I'm like, Okay, seems reasonable. I go to the website, it's got a really good deal. Should have been the first thing I saw where I was just like, Whoa, wait a second here. There's no way they're this is way too good of a deal. But I was just like, Okay, kid found a deal. I'm gonna I put on my credit card information. And ⁓ wheels spinning and it's like going through this whole thing. I'm like, this is a weird processing thing. It's like talking about like the extra and this, and it's just going through it, doesn't go through. And I was like, man, that didn't work. I'm putting a second credit card because I'm an idiot. And as I put in the second credit card, it's literally as I'm clicking the button, I was like, wait a minute, there's too many trip wires right here that I'm going off in my head. And I was like, I just gave two credit cards to a scammer. That's what I did because it wasn't a real sight. And it was like, it was like it looked right. Sascha Meinrath | X-Lab (52:34) Yeah. Christopher Mitchell (52:51) You know, look real professional and and ⁓ man. And so like and so I went to the credit card companies and I wanted to be and they were like, Is your card stolen? And I was like, Not yet. Like I was like, I just gave it away at the card number like an idiot to like, you know, like the but like I didn't try to talk to anyone because I knew that that was foolish. And ⁓ and I tried to use the online tools and like none of the drop downs made sense, right? It didn't fit in there of me being like, Hey, like you should Sean Gonsalves (52:53) ⁓ man. Sascha Meinrath | X-Lab (53:01) Yeah. Sean Gonsalves (53:01) Right. Yeah. Mm-hmm. Right. Christopher Mitchell (53:18) put fresh like charges under f under extra scrutiny for the next six months as we watch to see what happens. But that's not an option, right? So now I'm just watching, waiting for the fraud to happen. you know, so Sean Gonsalves (53:30) ⁓ man. Sascha Meinrath | X-Lab (53:31) Yeah, I just moved banks to my local bank here because I got so sick and tired of just these petty things where I'm like, I just need to solve this one little thing, whether it's a statement or like, you know, had a question. And I was like, I don't need this to be a two-hour like sisyfician task. Like it like I just need to talk to a human. And so now like the thing that I eschewed, I hate talking to humans. I'm like, please give me a human because like I just want to get the thing solved. Christopher Mitchell (53:59) Mm-hmm. Sascha Meinrath | X-Lab (54:00) And there was this moment way back in like the early twenty twenties where like for a moment humans were slower. Right? And now I'm like, no, if I have a problem, I need to find the human as fast as possible. There's the only thing that will solve this without wasting extraordinary and that is the exact opposite of what we were promised all these systems were gonna do. Sean Gonsalves (54:13) Right. Christopher Mitchell (54:23) Yeah, buffers. They're buffers. They're buffers to annoy us and to get us to just abandon it and take the hit and move on with our lives. Sascha Meinrath | X-Lab (54:31) Yeah, it's a war of attrition by customer service, a hundred percent. Christopher Mitchell (54:34) Yeah. All right. Thank you guys. It was a fun ⁓ like to have those meandering ones at the end. So if people who are are bored by it, they can just move on their way, but everyone else can sit there and enjoy laughing at my stupidity of like, yeah, anyone else want my credit card information? It's right here. Yeah. Sean Gonsalves (54:42) Yeah. Sascha Meinrath | X-Lab (54:49) Can you just hold it up to the camera? Sean Gonsalves (54:52) Well, yeah. Rod Rod Rod Serling signing off from the Twilight Zone. Sascha Meinrath | X-Lab (54:57) Yeah. Cheers. Christopher Mitchell (54:58) Thank you guys.