Elma Avdic: Thank you. So my name’s Elma. And I work in the CONNECT Research Centre for Future Networks and Communications based in Trinity College Dublin. What I want to talk to you about is this idea of machines gov­ern­ing things. And one par­tic­u­lar thing that I see gov­erned by the machines is the radio spec­trum. That’s an exam­ple where I see this kind of reg­u­la­tion can work, and I guess the most rad­i­cal thing about it is that I think a robot­ic reg­u­la­tor can actu­al­ly bring more human val­ues to its reg­u­la­tion. So hope­ful­ly I will show you why I think that and how that can hap­pen.

So, I have so many things to tell you with lit­tle time, and I just kind of want to open this box and you take from it what touch­es you and unpack it fur­ther. I did a PhD the­sis on the cul­ture of spec­trum shar­ing, Emancipation of the Hertz,” that was the title. It was a multi-domain analy­sis of spec­trum shar­ing, start­ing from nature, the domain of nature where we find the resource; through the domain of tech­nol­o­gy, which puts the resource to use and in the case of spec­trum it makes it exist; through the domain of busi­ness, where radio spec­trum is used by a ser­vice provider to pro­duce capac­i­ty and sell it out to the cus­tomers, to the end users—to us. That all hap­pens under the domain of reg­u­la­tion under some rules. And up to the domain of soci­ety as spec­trum is used so much in every­day com­mu­ni­ca­tion. And I see spec­trum as a social issue per se.

So I feel that the right way of think­ing about how to imple­ment this AI reg­u­la­tor, the machine-enabled reg­u­la­tion of spec­trum shar­ing is to input the prin­ci­ples into the machine, because the machine has to learn from some­thing, right? So I think it’s about the prin­ci­ples, not rigid if-then-else rules, because they don’t real­ly give space to the machine to be real­ly intel­li­gent and get out of the rules. So, I will talk from where we can extract these prin­ci­ples, how that relates to radio spec­trum, what kind of gov­er­nance and struc­tures are behind the radio spec­trum, where is the prob­lem with that, and what is this lega­cy that you’re sup­posed to go with into the con­nect­ed future. And final­ly, how can we make AI-enabled reg­u­la­tion for spec­trum shar­ing.

We made a myth out of shar­ing. The non-sharing struc­tures gov­ern­ing the under­ly­ing pipes of the dig­i­tal world want us to think that we’re already shar­ing enough by shar­ing bikes and pets. And that’s a pop­ulist dis­course made to hide their hege­mo­ny and pro­tect it from ref­or­ma­tion. So in this map of the Internet, the bright col­ors where sort of the fire­works are, it’s the place of plat­forms where we do share. And it sort of shows how we cel­e­brate and take pride in the way we share on plat­forms. But we just sit silent­ly in the dark when it comes to the pipes, and that’s where the radio spec­trum and the infra­struc­ture are.

So, there is a lot at stake when it comes to radio spec­trum, so let’s start meet­ing the stake­hold­ers and let’s start from the head. The FCC is the reg­u­la­tor, the pol­i­cy­mak­er, and pol­i­cy enforcer. So be like Peter from Family Guy, know every­thing about the FCC. In par­tic­u­lar we need to learn about the FCC his­to­ry and its unfair race with tech­nol­o­gy. Unfair because tech­nol­o­gy’s fly­ing, and the FCC is doing a crab dance.

As much as the Big Spectra wants us to believe it, the balka­nized, entrenched, guard­ed allo­ca­tion chart is not the elec­tro­mag­net­ic spec­trum. Placing own­er­ship and pro­tec­tion in the very def­i­n­i­tion of their work­ing def­i­n­i­tion of elec­tro­mag­net­ic spec­trum, they made it a piece of invis­i­ble land in the sky. But it’s not. It’s a spe­cial resource, usable in bril­liant ways simul­ta­ne­ous­ly, instant­ly renew­able and mag­i­cal­ly exploitable by the ever-progressing tech­nol­o­gy. And yet, they keep it bound in a chart. So spec­trum is not the prob­lem, nor is the tech­nol­o­gy. It’s the humans, the crea­tures of reg­u­la­to­ry habit.

In an effort to please every­one with finan­cial pow­er and not to leave the com­fort zone much, the reg­u­la­tor comes up with these com­plex spec­trum shar­ing sys­tems. It is com­mend­able that they final­ly rec­og­nize the poten­tial of unused spec­trum. And in the case of the depict­ed CBRS spec­trum shar­ing mod­el, it is the rarely-unused mil­i­tary spec­trum in the 3.5 giga­hertz band. But, as we speak the FCC is prepar­ing to announce in a few days that this new space is going to be turned into yet anoth­er park­ing lot for the big car­ri­ers. So, what are the alter­na­tives to that? Making it a nation­al park? Sort of a nod to com­mons. Or a third way?

It’s clear that we can­not go on any­more with the same old same old. A sus­tain­able future asks for new more sus­tain­able busi­ness mod­els. And the resources have start­ed shapeshift­ing. Rigid, non-shapeshifting busi­ness­es won’t be able to process the flu­id resources of the future. In wire­less com­mu­ni­ca­tions, they can not still be just phone com­pa­nies, and not every play­er has to be a big one any­more. Everyone can do a small part and make their mon­ey in the mar­ket with what they have. This democ­ra­ti­za­tion of the mar­ket seems also nat­ur­al and total­ly in the spir­it of net neu­tral­i­ty and oth­er move­ments that are pro­tect­ing small busi­ness­es by using the ben­e­fits of the World Wide Web. And yet we don’t talk about it. So is it too dis­tant, too abstract? It should­n’t be. So you might want a piece of this spec­trum pie in the sky. And you want to become a provider. So by the ben­e­fit of shar­ing, you won’t be fac­ing this dilem­ma of hav­ing your piece or eat­ing it.

So the cur­rent mar­ket is defined by big com­pa­nies bid­ding for chunks of spec­trum for their exclu­sive use under the FCC’s terms. It is ensured that these chunks are far apart, because the bid­ders are afraid of the oth­er users dis­rupt­ing their ser­vice. And they, as well as the reg­u­lar rules, imag­ine these oth­ers have Fred Flintsone-like equip­ment, leaky and loud. And…they don’t. So based on that, they devised these bulky defense mech­a­nisms at the expense of spec­trum.

And as in that famous selec­tive atten­tion video, while you keep an eye on the ball you miss the goril­la in the bas­ket­ball court. So here the guer­ril­la is the spec­trum scarci­ty. Everyone is com­plain­ing there’s not enough spec­trum. But they nev­er join their forces to fight this lack togeth­er and make a prop­er uti­liza­tion of spec­trum a com­mon pri­or­i­ty. So no, they hoard unused spec­trum and then point fin­gers to the oth­ers doing the same. It’s their busi­ness mod­el. So Professor Nash has every right to be dis­ap­point­ed in this scene. The mar­ket is play­ing the wrong game, against the wrong oppo­nent.

Spectrum is an amaz­ing nat­ur­al resource with many beau­ti­ful fea­tures, and yet it is posed as a hos­tile ter­ri­to­ry that needs to be col­o­nized and tamed. That’s wrong. We should team up with the spec­trum and its super­pow­ers.

Okay, so step into the big car­ri­ers’ shoes for a sec­ond. It’s hard to blame them. They are doing what every­one else is doing, fight­ing for prof­it in the only bat­tle­field they know. And the world would col­lapse with­out them; they’re too big to fail. So democ­ra­tize the mar­ket already. If you’re too big to fail per­haps you’re too big to exist.

And so they repeat their argu­ments, like the list in the slide. And because it’s the dog­ma which pro­tects their assets. And yet humans, in their own shoes, they see a dif­fer­ent pic­ture. People see how things can work on human prin­ci­ples. How shar­ing func­tions in the cre­ative dig­i­tal world. So, they speak against big com­pa­ny pro­pri­etary nar­ra­tives. So peo­ple like Richard Stallman, Aaron Swartz, Lawrence Lessig. Lessig spoke about spec­trum as well and rec­og­nized the role of the com­mons and proph­e­sied the suc­cess of WiFi. So we need more Lessigs in spec­trum, we need the enthu­si­asm of Richard Stallman and Aaron Swartz, to take spec­trum from bottom-up.

Screenshot of Ludacris saying, "I just jammed every signal up and down the spectrum."

Because the main­stream does­n’t get spec­trum. It’s mum­bo jum­bo for them so, Ludacris can say some­thing this ludi­crous.

And yet, spec­trum is enabling each and every rung of our needs hier­ar­chy today. The ear­ly Internet start­ed from the top of the pyra­mid. The rev­o­lu­tion that wire­less access brought stormed it all the way to the foun­da­tion.

So you see me get­ting all per­son­al here already. I stud­ied spec­trum in the per­spec­tive of shar­ing economies, detect­ing these five domains where enablers for prop­er shar­ing can grow. So these domains are nature, where the resource is; tech­nol­o­gy, were the net­works are; busi­ness, where the inno­va­tion is the enabler; reg­u­la­tion, where the neigh­bor for shar­ing is dynam­ics and respon­sive­ness; and soci­ety, where the cul­ture of shar­ing is the enabler.

So I looked at the wax and wane of shar­ing, from blood to code. So why do we com­mit more to Git repos­i­to­ries than to blood banks? The log­ic behind a suc­cess­ful shar­ing mod­el needs to rely on two fac­tors. That is prop­er uti­liza­tion of the nature of the resource, and the sec­ond one is sen­si­ble, mean­ing­ful incen­tiviza­tion. Both of these mech­a­nisms mobi­lize the five domains, and they can­not work with­out bal­ance in them.

So bal­ance. That’s what Lessig also said when he devised his own reg­u­la­tion the­o­ry, the one of the pathet­ic dot. So the dot is the behav­ior we want to reg­u­late. The forces that the act on it are law, mar­kets, norms, and archi­tec­ture. Kind of sounds like what I just men­tioned but there were five domains there and here are four forces, but we’ll get to that.

That CBRS mod­el that you saw on the t‑shirt ear­li­er. That’s the best we’ve got, and that says a lot about just how spec­trum shar­ing is not work­ing. So I did a com­pre­hen­sive, rig­or­ous con­tent analy­sis of it’s rule­mak­ing pro­ceed­ing, which is pub­licly avail­able on the FCC web. The cod­ed lan­guage com­pa­nies use to talk to the reg­u­la­tor in order to push their agen­da into the rule­mak­ing, I tab­u­lat­ed their opin­ions issue raised. That helped clus­ter them and estab­lish the clear teams in the game, as well as to see the most impor­tant issues that were sup­posed to make the CBRS shar­ing mod­el dif­fer­ent from any­thing we have seen so far. AI helped with this. So we had a hybrid part­ner­ship where I was the con­text mas­ter and the AI was the con­tent mas­ter.

So read­ing thou­sands of FCC MultiLog pages told me every­thing I need­ed to know about the prob­lems of human reg­u­la­tion of spec­trum. So, why don’t we let AI take care of it? Or should we do it in tan­dem, the way I did the con­tent analy­sis?

We are ter­ri­fied with the machines putting num­bers on us based on some non­trans­par­ent algo­rithms, because who knows what’s in there? We are also ter­ri­fied by the machines putting num­bers on us based on a trans­par­ent algo­rithm, because adver­sar­i­al engi­neer­ing will find a way to game the sys­tem and harm us, the fair play­ers. Even when humans try to metri­cize us, we often say, Hey, you don’t know the con­text. You don’t know the whole sto­ry. So keep your con­text non-aware machines from us.”

So escape one boss and you end up being bossed by an algo­rithm. Is that the future of the shar­ing econ­o­my rev­o­lu­tion? That’s dystopia coat­ed with shar­ing. Can we do bet­ter? And why am I not talk­ing about spec­trum yet?

Data is not going away. Nor are computers—much less math­e­mat­ics. Predictive mod­els are, increas­ing­ly, the tools we will be rely­ing on to run our insti­tu­tions, deploy our resources, and man­age our lives. But as I’ve tried to show through­out this book, these mod­els are con­struct­ed not just from data but from the choic­es we make about which data to pay atten­tion to—and which to leave out. Those choic­es are not just about logis­tics, prof­its, and effi­cien­cy. They are fun­da­men­tal­ly moral.
Cathy O’Neil, Weapons of Math Destruction

So data is the new reli­gion. Just put enough data in, and the result will come. Big data will solve every­thing. And while it may seem like I’m jump­ing on that band­wag­on, let me tell you some­thing. Data is not just data. It is a new dimen­sion of spec­trum. It’s not a descrip­tion of spec­trum, it is spec­trum. It’s the total oppo­site of the allo­ca­tion chart.

And I would­n’t ask for machines if I did­n’t need them. But look at them. All fair and nice if we train them well. And I promise I’ll be there or some­one like me to be a part of this dynam­ic duo, because that is every­thing the cur­rent sys­tem is not. It is a sta­t­ic, sin­gle tone. Even though it’s made by humans, it lacks con­text. And the time need­ed to make a pro­gres­sive change is mea­sured in life­times.

So how can we turn this dystopia to utopia? We don’t have to clone our­selves into dig­i­tal ver­sions. We don’t want a neur­al net­work emu­lat­ing Herbert Hoover. Herbert Hoover is the embod­i­ment of lega­cy in the FCC or a [?] deci­sion tree. This is why the machines won’t be trained using rule­mak­ing pro­ceed­ings and FCC his­to­ry. The bur­den of lega­cy is a curse we do not wish to pass on. In the spir­it of every­thing the shar­ing economies taught me on my voy­age toward suc­cess­ful spec­trum shar­ing, I can tell you that it’s the basic prin­ci­ples the machines need. They can’t under­stand con­text? Well I can. But they can real­ly adapt.

Fast and small changes of rules in dif­fer­ent sit­u­a­tions of the radio envi­ron­ment, it’s a dream come true for spec­trum. The prin­ci­ple of max­i­miz­ing spec­trum uti­liza­tion as the cra­dle lead­ing a flex­i­ble rule­set which fits space-time-frequencies and data like a glove. Yes, please.

So the Tyrell Corporation was a spooky dystopi­an enti­ty in Blade Runner. I mean how twist­ed do you have to be to promise build­ing some­thing more human than a human? I’m not promis­ing that, I’m say­ing it’s already there. If we bun­dle the com­put­er net­work to han­dle spec­trum in Manhattan right now and replace the FCC, it would be more fair, more sus­tain­able, more con­scious, and more pro­gres­sive than the humans of the FCC. And yet these traits are cel­e­brat­ed as human virtues. So I did this. I worked in Manhattan as a use case, and promised in the tease is that first we take Manhattan, so now let’s take Berlin.

I fed a thou­sand pages of vision­ary takes on the future by rev­o­lu­tion­ar­ies into my com­put­er and asked it to make me a slide. This was the result.

You know how I told you that data is a new dimen­sion of spec­trum? Well, peo­ple kept on adding dimen­sions to the def­i­n­i­tion of spec­trum. First it was just fre­quen­cies. Then there was time. Then space. Then pow­er. But data was slip­ping between our fin­gers because it was nev­er that impor­tant. Why would you care about the real-time usage infor­ma­tion from an adja­cent user if they are far enough not to cause inter­rup­tions to your own use and the FCC’s keep­ing them at a con­ve­nient dis­tance? But if the FCC lets the two of us be a bit clos­er and use the spec­trum resource more effi­cient­ly, I will start look­ing at their data to dance with them. Use spec­trum when they don’t, and they will rec­i­p­ro­cate.

So, data is a dimen­sion vital for shar­ing. A strong force that appears once spec­trum users are close. So sum­mon the data, then, and look out of your n‑dimensional space into n+1‑dimensional space.

After I got all the results I want­ed from the analy­sis, I had any image of a futur­is­tic star­ing, where all enablers of shar­ing come togeth­er and make spec­trum shar­ing the real­i­ty. So Lessig was right all along with the pathet­ic dot. It’s just that split­ting the force of archi­tec­ture into two, nature and tech­nol­o­gy, opened the door of per­cep­tion for spec­trum. It’s the nature of spec­trum that has been miss­ing from the reg­u­la­to­ry equa­tion all along. It kept us from build­ing the cul­ture of shar­ing because we did­n’t under­stand it prop­er­ly. Well that’s why I’m evan­ge­liz­ing it.

Spectrum is a unique resource. Get out in the streets and chant that, be aware of the ether in the blood­stream of the mod­ern world, and the rigid con­trol the state keeps over it. Only then will the board­rooms make a point about it and take new busi­ness mod­els on board. They will build machines to talk to the reg­u­la­to­ry machines. A new reg­u­la­to­ry lan­guage, cod­ed in a dif­fer­ent man­ner will emerge. The data will plot his­to­ry and the future of net­works, and show where under­uti­liza­tion can be tar­get­ed. Licenses will be small in space-time-frequency, like rent­ing bikes. No more fifteen-year licens­es, take a fifteen-minute one. Be dynam­ic. Be like spec­trum.

So speak­ing of rent­ing and bor­row­ing bikes, just like Arnie in The Terminator, the machines will put under­uti­lized resources to use.

Citizens of [Telecosmos?], get your spec­trum ana­lyz­er, start talk­ing about spec­trum; at least as much as you talk about Article 13 or net neu­tral­i­ty. So I just want­ed to open this box and throw ideas out there, as I think we’re ready for a new soci­ety. And we need to start more active­ly chal­leng­ing the lega­cy in rule­mak­ing and gov­er­nance. And we need sys­tem­at­ic ways to fight them. Thank you.


Moderator: Thank you. So, questions. Yes. Well.

Audience 1: Well, I have tried to lobby for TV whitespace usage a little bit. I'm probably a terrible lobbyist myself. But I think it's not all that shiny, like you say it. For example you talk about artificial intelligence like something that needs to be centrally localized somewhere. Which is the total fuck-up the FCC made with TV whitespace in the United States. Because that type of communication is relatively short-distance. So actually the devices in the field can actually detect when a channel is used, and then they can use it because it's unused. And they have to be polite to each other.

So it's not like you need a centralized entity that manages all those things, it's the swarm intelligence that actually can build an effective artificial intelligence if you put in the right rules, rather than have it centralized somewhere. Or did I misunderstand your presentation?

Elma Avdic: So, I don't know what you mean when you said you lobbied? for the whitespace. We can maybe talk about it afterwards. But this is not a centralized network. It's a distributed regulator and distributed networks of operators. So it's decentralized, what I had in mind.

Audience 1: I'm just a little confused because towards the end you said no more licenses for fifteen years, but the licenses granted for fifteen minutes. So there is an entity that grants the device, or the user or the business…whatever you named it when you're talked about it, that licensed physically. Radio physics and radio devices are not that frequency-agile that they can operate in a totally different band, for example. Also, the band needs to fit the application that they're built for.

Avdic: So, the licensees that I have here in the picture are granular and modular in space, time, and frequency. What I didn't explain here, as I didn't want to go into too many details—I didn't see this as presenting results of my thesis but more like a talk. So, I have actually a machine replacement of a broker. A machine that is doing auctions— Not all auctions, like auctions on the spot, based on detecting the radio environment, calculating the— Because this data, which becomes really really important to users that are close in time, space, and frequency when it comes to spectrum sharing. So the machines are doing this real-time.

Audience 1: Well we should talk afterwards.

Avdic: Let's talk, yeah.

Moderator: Further questions?

Audience 2: Hi. You talked a lot about the FCC. I was wondering if you could tell us who's in charge of radio frequencies in Europe.

Avdic: So, regulation in Europe is a bit more complex than that. So there is Ofcom, which is the regulatory body in the UK. And it's kind of a mini FCC system. It's built like that; all the boards and offices. But the majority of the rest of Europe actually, everybody besides the UK…it's an EU regulation, a set of countries; there is twenty-eight or twenty-nine set countries that are part of this. And each country has an NRA, that's National Regulatory Authority. So basically each country has its own regulatory body, but they all submit to the European Commission laws of regulation.

Audience 3: Hello. I'm just trying to see if I understand your talk, right. So my understanding with like big companies, they pay loads of money for exclusive access for fifteen years or whatever. And what you're suggesting is a bit more like…a bit how WiFi does channel-hopping. So lots of us might be able to share the same resources over kind of time slices instead. Where do you start with that? Because it sounds cool, but I don't know how you go from here to there.

Avdic: So first there is a lot of work to be done. That's where the magic happens. But your question about the licenses. So, I didn't study sharing economies just like that. So what I learned is the importance of rights to access, and replacing the property rights which are currently the spectrum management model that these licenses are sold in auctions by—you know, it's a property right, sort of. And what the users get in the license are two rights. One right is the right to be protected from interference—guaranteed protection and you know, quality of service. Guarantees and everything. And the other right is the right to exclude others from using their spectrum, even when they are not using it. So this is why it's seen as a property.

So, what sharing economies are teaching us is that you don't necessarily need to own a car anymore to get from Point A to Point B, so it's about the rights to access and just reforming spectrum management models in this way. Because spectrum is a such a resource. It's a fast-cleansing, recyclable resource. Instantly renewable. Perfectly renewable. So, I think that that can work.

Audience 4: In the diagram that you have on the screen right now, you made mention to the emergence of new business models? I was wondering if you had any speculations on what those might look like?

Avdic: I do, but I can't tell you that. [laughs; audience laughter]

Audience 4: Fair enough.

Audience 5: In one slide, you put the sharing economy versus the legacy model. And there were two logos that I didn't like; that was Uber and Airbnb. And I don't recommend this practice, because according to Yochai Benkler (that is an important person that talks about the sharing economy) they call these kind of corporations "unicorns" because they cannot exist. They're existing to destroy some legacy economies that are working, like hotel's, hostel, or in the case of taxis, etc., no? So, these kind of people are probably putting in their presentations, "Oh, Wikipedia, Git, Linux…" and their company. And with this kind presentation I think it's dangerous for us. Because then we are all with that kind of people.

Avdic: So um…I am not saying that sharing economies are great. I use them as an example as I was looking for just an example of successful sharing. So, what you just said about Uber and different kinds of platforms, these logos here are not here necessarily to reflect their ideology but just to show that sharing is happening. So there are different kinds of sharing. I'm not saying I agree with Uber—I don't, of course, or these other types of sharing economies. But it's here to show that a culture of sharing is developed on the platforms that are multidirectional business models. So they work like that. The user gets on the platform and creates content themselves, unlike the pipes which are unidirectional. So the spectrum management and distribution model is an example of a pipe model. So the regulator creates the supply of spectrum and pushes it out to the customers, who don't necessarily have a say in the product; cannot change it or modified it. So yeah, I I take your comment on board. I won't put this anymore here on a slide like this, because I I realize it may confuse people. But it's not the ideology, it's more of something that's feasible right now. It's the only sharing that works.

Moderator: Okay, one more question.

Audience 6: So maybe this is what the AI that you created does, but when I think of machine to machine I think of them having self-sovereignty, like the machine acts on its own. So in your case the spectrum would be able to sell itself to a corporation, where as it sounds like it's still more of an asset in your model where someone's buying it from someone else, whether that's controlled by a centralized service or not. So, have you considered the fact that a spectrum could act in its own service instead of being traded between people?

Avdic: So in most of the societies unfortunately governments own the spectrum. So anything that has to do with spectrum sharing or issuing spectrum under exclusive use has to happen under some kind of rules. The model that I showed is distributed, is decentralized, and it has a automatic sort of auction manager. So governments runs the auctions, and they generate hundreds of billions of dollars of profit. But auctions happen once in ten, fifteen years because there is this norm of entitlement in the spectrum world. And that means really long-term licenses that are renewable forever. And in telecommunications just like in technology, that's a few generations of telecommunications. So I'm suggesting to replace these one-time huge revenue boosts for the governments by business models that are built on sharing from scratch, and when thinking on a long-term scale they actually bring more revenue. So it's automated.

Moderator: Thank you Elma.

Further Reference

Radical Networks 2018 archive site


Help Support Open Transcripts

If you found this useful or interesting, please consider supporting the project monthly at Patreon or once via Square Cash, or even just sharing the link. Thanks.