Episode 10: 6G, Quantum Computing & AI Agents: The Next Wave of Telecom Innovation

Peter Jarich (00:00)
Hi everyone and welcome to Intelligence and Noise, the GSMA intelligence podcast, where we sort through the news of the week, try to add some intelligence to the things that you're probably already paying attention to, and sift through the noise, the stuff that maybe you aren't looking at, and telling you why maybe you should be. as always, not as always, but as often I'm here with Tim. Hey Tim, how doing?

Tim Hatt (00:25)
Peter, good. How are you?

Peter Jarich  (00:26)
Good, good, good. We're in the middle of the summer, which means things are supposed to be slowing down. but they're not, which is good. because it gives us something to to talk about. so we got a lot on the docket today. We're gonna be talking 6G, we're gonna be talking quantum. as always, we'll be talking AI because it is 2026, which means we can't get away from AI.

But and it's not just because it's summer. I'm gonna kick off my story time with Peter introduction here with a bit of a a bit of something that that that that maybe comes out of out of left field. so it's it's about young people and and drinking. I mean it's the summer, right? it is it is time to kick back with with something in the in the heat.

But what's interesting is I I saw recently some news out of the IWSR. and if you don't know what the IWSR is, I mean you're you're in the telco world, so so you can be forgiven. it's originally the International Wine and Spirits Record. like many organizations, they just go by an acronym now. it's a reminder, it is MWC.

Not Mobile World Congress. So, so so please let's get that right and let's refer to as IWSR if you ever see any of those folks. They're in London. We should go and see them sometime, sometime, Tim, or have them into the office. They're a group that does a lot of research around around beverages, as you might think, wine and spirits and other things. And it was the middle of the month, middle of last month of July, they put out some news that was talking about Gen Z drinking habits.

And it's and it and kind of important because a lot of the news out there has recently been about a young people don't drink, young people are drinking less. you see that that that's driven a whole bunch of different changes to to businesses. I think I think to some extent that probably drove a bit of the the no and low alcohol beverage industry, or at a much more at a much different level, right?

You've got folks in various regions who are ripping out vineyards. alcohol consumption is going down. Young people aren't drinking. But what's interesting is the study came out and it showed young people are drinking about Gen Z drinking about almost exactly the same amount, actually the same amount of time as the general population. I think the numbers here say across 15 markets they looked at the the share of Gen Z drinking is at 74%.

Up from 66% three years ago, it's nearly identical to the rest of the population. There's all sorts of interesting details about who drinks less, who drinks more. And of course, you can ask how much they're drinking, right? It doesn't mean that they're drinking as much as other folks. All sorts of interesting information. Other than just chatting about about alcohol, what is what is the point? What's think what's interesting is when I think about a few years ago, right? And we talk about three years ago, we what

You know, one of the big questions was what is this gonna mean for networks? And it was on the telco side of things, changes everything that we know about usage patterns. Right. Changes change, you know, people are using it in different times, right? They're using it from different places. This is changing things forever. Everyone's gonna work from home forever. It's it's no one's gonna go to stores, they're gonna be ordering. They've gotten used to so much gun, getting used to having their food delivered to them, no one's going back.

Offices are going to be empty, and then there we go, and the offices fill up and the telecom networks go back to something more akin to what it was. But then it's going to change in another two weeks when something else comes up. And what I think is interesting is for me, it reminded me of these conversations about how more than questions of is data traffic going up, is data traffic going down, is what does variability look like? And how do we think about building in flexibility, right?

How do you, if you're in an alcohol business, build flexibility into your business model to adjust for the fact that young folks may drink one day we may think they're drinking more, one day we might think they're drinking less. That changes. As a telco, how do you build flexibility? How do you think about what your future network looks like? Recognizing that these things change now, our understanding of them changes, right? We might think very much, this is what's happening one day. And

We

learn something more. It's not to say that we got it wrong. It's just you learn something more and and and that changes. And that's particularly important now because we are on the cusp of six G. And that brings us to our first story. People planning for first it's Tim is always amazed. You never see you you you don't see you you you don't see where I'm going and how I'm actually

Tim Hatt (05:32)
 

can I just say on that I I didn't have told you this before, but my original childhood dream was to start a brewery. So the the fact that you have brought this story up has given me new new insights to factor into how I how I do that. This is great timing.

Peter Jarich  (05:46)
Don't don't don't give up. Don't

give up on those childhood dreams. Gen Z is Gen Z is going. And apparently part of this news is that Gen Z is more likely to also be drinking with other people, with friends. So it's a social thing, which is which is which is not not bad. so so yes. Tim's gonna open a brewery and we're gonna talk 6G, the first story of today.

Day, and I think for those of you who haven't listened before, we always couch this within the the constructs of the GSMA CEO newsletter goes out and we talk about one of the two stories that made it into the newsletter, one of the two stories that that didn't, that could have there before the the grace of our DG goes to the stories that get picked. And the one that got one of the two that got picked was the new.

6G alliance that was announced by the NTIA. we say new, but it was sort of flagged in it was flagged in March. I believe it was it was flagged early on. There was a call to action and and there wasn't necessarily it didn't form right then, but we sort of saw it coming back then. But now we saw it just last week technically

Call to action for six G Leadership and Security. So it's a it's a bit of a mouthful. Tim, there's there's a lot to unpack here, but but thoughts and maybe some more details on what it is for folks that didn't read the full news.

Tim Hatt (07:28)
Yeah, really interesting story. And to your point, let's put some context around this. So this is an announcement from the US government to start a we call it a coalition of the willing of allied partners who are getting together to essentially shape 6G, whether that is from how the business architecture looks like, the technical architecture, and feeding into standards. they've announced this. You're right, Peter, they they trailed it.

in March. I would say even it's probably been earlier than that informally, given how the US government has sought to take a a leadership position in shaping technologies that are of strategic national significance. I think the interesting point for me, Peter, there's a couple of interesting angles and I want to get your perspective on it. One is, you know, we haven't really seen this before. Typically mobile technology cycles are stipulated by

Standards bodies and various associations, not usually led by kind of geopolitically driven efforts. This is the first time that we've seen that. I think that's a point of interest. Secondly, and I want to get your reaction to this, I think it's as interesting for who is in the alliance versus who is not. We see over 20 countries lined up. Most of them are Western allies of the US, the UK, Japan, etc. We don't see any of the Brit countries in.

there is no China, there's no Russia, Brazil, India. And I think that says some interesting things about where the current states of alignment are on that. Not saying they can't get bought in, but I got some other thoughts. But what what are your thoughts on that so far?

Peter Jarich  (09:09)
Yeah. So so

I think I mean

Stopping short of my thoughts, what I think it's interesting is if if you look to see what other, you know, what what folks in the media have been talking about and look at some of the headlines, right? So if I if I look at a handful of of the headlines that are out there from from different folks, right? You've got, you know, anything from NTI, NTIA plans to launch a 60 called action to unite allies, right? And that was back from March, right, when when this first called out. And then you look, you know, more recently, right? You've got

you know, Uncle Sam needs you to fight for 6G leadership and security, lest Beijing get there first. Right? It's a bit straightforward. Folks, friends over at SDX Central. US wants quote unquote like-minded nations to join 6G software-led project to counter China. So, so clearly in the media, there's obviously a big focus on who is not there, including China. What I think is interesting though, and to be fair,

Right. I think I think when when we were looking at it, my first, you know, one of my first takes was you've got twenty or so countries in there. that's about eighteen percent of global mobile connections, twenty-five percent if you're looking at five G. But what's interesting is and it's always difficult to break down intellectual property and and standard submissions by country. But

If you look at some stuff that's out there from various folks that's been publicly reported, including the folks at IPolytics at C analysis, of some academic studies, you find that that if you take those 20 countries and you sort of ask, hey, what share of three GPP submissions standards do they likely account for? The conservative number.

Is that the alliance accounts for about 55 to 60%? And that's that's the conservative estimate, right? If you look at an aggressive estimate, it's 75 to 85 percent. And now, granted, that's all 3GPP, so maybe less weighted, maybe more weighted towards back end than than than you know going forward to the 5G side of things, and who knows what'll be for 6G. So yeah, so I think you can make the argument, the media's making the argument that this is just as much about who's not in there versus who is.

but they've got you you can't deny that they've got the the the heavy hitters in terms of who are forming that. Now, again, you talked about the geopolitical angle, what's there. yeah, you can't you can't ignore that, right? I mean, I think it's it's it's obviously a little bit too too in your face.

Tim Hatt (11:52)
Yeah, it it's a good point, it's a shrewd point. And and I think the patent point speaks to why they're doing this now as well. I mean, we are at a point where we're late in the 5G cycle, we're eight years in, you have a six G on the horizon, twenty twenty-eight, twenty twenty-nine. I think the industry recognizes that it is a nationally important technology. They also recognize, however, that

There were lessons to be learned from 5G, which, from a monetization angle and how that is fed through the 6G design, can be addressed through this type of initiative. I look at some of the stakeholders talking about, for example, avoiding a two-phase network rollout. We had 5G non standalone, with the exception of China, that then went to standalone. That was sort of a product of circumstance, but I think most people will want to not have that when it comes to 6G.

So for me, the interesting points putting aside the geopolitics of it are what do we see this alliance, you know, produce in terms of, you know, impactful outcomes? And do we secondly start to see either more joiners to this or similar initiatives in other hemispheres of the world? I think that's the main

Peter Jarich  (13:10)
Yeah, yeah, I think I think that 

the new joiners one's a real is a real interesting one. and to your point about timing, like just on a timing side of things, it's interesting because I was I was just thinking about this on another subject today. Like it seems silly now to not talk about 5G, 6G, right? We wouldn't have done that three, four years ago. You'd be like five we talk about five G six G we're in the five but now we know like and just the fact that we're doing that says, okay, we know it's it's coming pretty soon.

I think as to the goals here, right, and again, we'll see if this expands to other folks. What I think is interesting is if you look at sort of the the call to action, one of the things I think is particularly interesting is is one of the points they make. It's it's I think it's verbatim quoting here. Global communication systems are inherently interconnected, which can introduce cyber, economic, and national security risks for individually for us individually and as partners. Whether within or across borders, exposure to high-risk 5G or 6G infrastructure.

And that term high risk, right? I mean, I think I I think we we all know that the term high risk vendors is has been used across many of the markets that are included in here to flag specific suppliers. So I think that kind of does go a little bit to the point of you know, who is in versus who is not. The flip side is I think this will be where I'm interested to see what actually comes out of this.

is if you go through the document and you you look for recurring recurring phrases and and comments and this is you know this is where our friend ai becomes a godsend because you don't have to go through and look for you don't have to go through or or hire an intern or or or or send it off to to somewhere but you can say hey you know help me out here what do you find yeah 6G leadership at the top of course right not surprising but you get secure resilient

open interoperable interoperable secure resilient AI supported so it's it's that this focus on interoperability this focus on open i mean you get down to when you get down to contact group you know that you're getting away from from things that actually actually matter but resource efficient not surprising resilience not but it's that interoperable open it it comes up again again and again and and what that means in terms of open

Where that line is between open and and who were included, I think that will be fun to see how they walk that line 'cause that's we know yeah, they know they need to be.

Tim Hatt (15:44)
And it's also

definitely and and we're gonna talk a little bit about you know ATT's LLM efforts with OTEL later in the podcast, but I think it's also an interesting point to think about what does open mean now? I mean, we've talked about open in several contexts before. This is almost a tacit admission that to be successful with 6G, cards on the table now, and both from a vendor and telco perspective to understand.

where it makes sense to join up, join forces, work in a quote unquote open way, versus what needs to be proprietary. But I think having the con the the interesting point for me is is really around the timings. It's getting the conversation now ahead of deployment rather

Peter Jarich (16:30)
And that is the other part though, right? Of I think then to maybe the note to end on is to see where does this go? Is it more than just conversations, right? They've got a nice timeline.

Tim Hatt (16:39)
Okay.

Peter Jarich  (16:40)
They've got a whole set of within one month we'll do this and do that. And we're going to start conversations. And then after that, we will identify who I like that level of detail and a call to action was actually kind of it was on the one hand, it was a little bit administrative, like like bureaucratic. Do we do we need to see for ever for the whole public that

Within three months, everyone's going to have their identified representative. But it's good because it shows they're moving forward. I guess the question is then what happens to reduce fragmentation? What actually comes out of this? And what impact does it have? And then because you called it out, more than just being secure and interoperable, do we make or can we make progress on the fundamental issues that we know are critical, which is monetization?

Right. And how are we going to make money from these things? And and it doesn't it's unclear how much of that, because it is a call to action on you know on security and and 6G leadership. But if this is not monetizable, then then that's gonna be a problem. On the monetization front, then that leads us into two stories that that we could have chosen from that didn't get picked.

we want to talk a bit about quantum. Before we go to quantum, we we can't not talk about

Tim Hatt (18:03)
Yeah.

Peter Jarich (18:04)
about AWS. Or Amazon. It's not AWS, it's Amazon. Apologies. Apologies to to yes to Mr. Bezos.

Tim Hatt (18:06)
We should talk about it. We talk that's right. The story just for our listeners,

the story Peter is talking about is is the Amazon satellite story, which took place last week, where Amazon has now filed with the FCC to increase its satellite constellation to just over five thousand satellites to be commercially launched from twenty twenty eight. and I think

We talk a lot about satellite and and there is a risk of satellite fatigue when there has been so much in the news on this. I think that's fair. But two two things that are that are interesting here. One, Amazon is moving fast. obviously we had the Global Star acquisition of the spectrum in the spring. That MSS spectrum will be incorporated as the key enabler for this constellation.

Two, it takes Amazon's planned footprint from a little over three thousand to five thousand. So we're talking about like a fifty percent plus increase in the constellation size. And I think that is interesting because it speeds the time to market and brings a new competitor to Starlink using a different model. Because Amazon has has been quite clear that they intend to deploy this in partnership with carrier partners rather than

So I think Peter, it's something to follow

Peter Jarich  (19:30)
And this is and then

it is, it is. And that's why if you are lucky enough, and I say it all the time, you're lucky enough to be on that distro for the CEO newsletter, you saw that was in there, because it is, you know, it as much as there's fatigue, it is important for all those reasons that moving fast, where this goes. I will say my slip of the tongue makes me wonder well, will will Amazon eventually put Lee will will they put their Sally efforts under AWS in the same way, you know, Musk has moved things around. I think that would be a a fun

I doesn't seem likely. but anyways, of the stories that didn't get picked, I think the one that we want to talk about is and and I like this when this happens, right? Where we we end up seeing a story that is not just like one thing happened, although it's it's important like the the the six G Alliance, great. But where it's more of a theme or a trend or you see some signals and we can group them together. I like that partly because

That's what we should be doing as as analysts. It also I points to a theme. This is around quantum. so we saw two unrelated well the two related but unconnected stories last week. One ATT working with with D-Wave to talk about quantum for network ops, extending some work they were doing with them. And then one where Telefonica was working with Worth on quantum in logistics.

And for me, and I think you're right, I think in the same way we talk about satellite exhaust when you see so much satellite news, I think for a lot of folks there's probably a bit of quantum exhaust where there's a lot of discussions around quantum and we've been talking particularly because we've been talking about it for so long. Right? We've been talking about quantum for however many years, and it seems like we're we're still talking about it. But for me this was an interesting story because both point to one a role for telcos, but also a role for

Quantum beyond encryption. Like I think so much the telco focused sides around post quantum cryptography. How do we make sure that we're preparing for a future when you can crack anything? And this this gives us an idea of like how telcos can possibly monetize quantum in ways that we don't always talk about. You're you're thinking of either or both of the both of the pieces?

Tim Hatt (21:53)
Yeah, no, exactly. And and and it is something we need to highlight. I mean, you talked about quantum has really, at least in the telco sector, been seen as something on a security perspective. Why? There are compute capabilities that quantum offers to help preempt security risks, and also we know that it introduces vulnerabilities. I think our own research, we published a report earlier this year. Seven percent of IoT devices, like of the thirty billion that are in active use.

Are quantum ready, meaning there is a huge latent liability there. What we've seen through these stories though, putting the securities side is with Telefonica and ATT, as as as you mentioned, is a quantum use in the business that can be either monetized or has direct cost savings. the Telefonica one was interesting. They they are looking at a logistics use case here where the quantum compute can help sort through.

the organization of a number of tasks. In the in the instance they talked about, it's actually stocking boxes for transport. and it led to around a five percent reduction just under that in terms of the end user result. May not sound like a huge amount, but actually when you scale up to billions of of devices, or in this case, hundreds of thousands or millions of boxes, that's significant. The ATT side is really around network optimization.

And again, when we will talk about some of ATT's efforts on on AI in a bit, but when you're talking about scanning billions of functions every day and you get down to 20 or 30 percent gains, which are being reported from the use of quantum, that has real impacts in terms of how the business is run. So I think strategically, Peter, we're now moving from talking about quantum as a buzzword and a topic to seeing real world proof points.

Peter Jarich  (23:50)
Well, it doesn't mean there aren't monetization opportunities around security, right? I mean, you can think of if if telcos can leverage quantum to help folks be more secure, right? Delivering security as a service, secure sim solutions, tying into sovereign hosting solutions, like there's there's clearly there. I think the other part that's interesting here, right, is yes, you've got the ATT side of things with with you know optimization.

And they're not the first. Like I think there's been a whole bunch of other folks that have talked a bit about what they're doing, right? I think, you know, Dokomo was talking about that earlier. We see some Telestra stuff. But I think is interesting is in the Telefonica side, you're looking at sort of a vertical focus there, right? Where it's it's actually thinking about what a vertical needs and then

leveraging quantum to help them out, which again goes to the larger B2B solution, right? I am always, because I don't know exactly and how complex these things are, we've been doing optimization for a while around all these, I am always surprised that it takes quantum to optimize logistics to to optimize that. But I assume there's a use case if they're talking about it. If there's a use case there for, you know, for the telephonic is looking at, then

across logistics, across ports, across airlines, across manufacturing, across energy infrastructure. I assume that that this really pro points to a good example of of maybe how telecos can also be pushing into that. So yeah, there's some some hope

Tim Hatt (25:24)
Yeah, you

Peter Jarich  (25:25)
there maybe for six G.

Tim Hatt (25:27)
For six G and and and your spot on the verticals, like when we when we looked at this earlier this year and we spoke to a number of operators and vendors to look at potential, you know, deployment rationale for for quantum. You see see things like you know weather monitoring, you see things like asset tracking, you see, yes, the security area, you see automotive, all of which are not obviously huge revenue lines of business, but as an enabler brings in applications that that were not possible before.

and I do think it's now moving to to much more from a a talking point to a commercial.

Peter Jarich  (26:04)
And and and to be fair before we jump off on this, to all the folks that are out there listening who are like, look, I've been I've been in the network optimization side of things. There's all sorts of things we're not doing because we don't have the the horsepower, get it, right? There's traffic routing, planning, cell congestion, resource scheduling, technician routing, spectr I mean, I get it, right? Right within the network, there's a ton of stuff that that is important there. But for me, that opportunity that we see with Telefonica, right? The idea of

leveraging this to monetize to make money outside of of your own business and become more efficient. that I think is the exciting one. Let's move on now. That's that's the end of the CEO newsletter. and it's gonna be the end for for a summer break. I think this was this last one was the last one for a few months. not a few months, sorry, a few weeks. a few summer break does not last several months, despite what Americans say about all of us in in in in Europe. it is it is not

that long. So what else though? Let's move on. What else struck you? You talked about ATT and OTE. A lot of folks may think you just misspoke and said ATT is opening up a hotel. What is

Tim Hatt (27:15)
Yeah.

Peter Jarich  (27:17)
what is what is what is what is OTEL and OTEL two point

Tim Hatt (27:21)
Okay, so for for those of you that have heard or have not heard, what we're talking about here, and and and it isn't important, Otel is is open telecom. and it is an ATT term for how they're adopting trained LLMs to a telecom specific domain. ATT has actually been working with a number of partners on this initiative, the GSMA being one of them as a data supplier.

Microsoft and and a range of partners to help take an LLM that maybe we think of from Anthropic or OpenAI or other Frontier providers and being able to develop something that actually understands telco. Now what they've done, and I'll get your reaction on this, Peter, is they have released a second version of it, which is trained at scale. It's trained on over 400 billion telco parameters. And as I say, through

that multi-party relationship is now leading directly to real-world business impacts. And I think the interesting thing for me is not just the outcomes that we see, but in terms of how this type of model catches on in the rest of the industry. one of our just quick shout out to to one of our analysts, Shiv Pucha, who published a really good or will be publishing a blog on this. and also to our colleague Louis Powell, who's doing some great work in the tech team to

on the foundry side promote this type of initiative. But I think we're seeing now a lot of recognition that that domain specific LLM is key.

Peter Jarich  (28:58)
Yeah, this is one of those things where I think we oftentimes as as we think about the the the news that we flag try to avoid a little bit inside baseball. We're not here to talk about or or big up any of the the GSMA stuff. but this is you know, this this Otel two point building on open source, right? A post trained version of it's the Gemma four model.

400 billion telecom specific tokens, more than one trillion process tokens. I mean, this kind of goes to this thing we were talking about years ago, is like, hey, does does telco need its own LLMs, or are we talking about can you take existing ones off the shelf open source? And I think different from I think to your point, you know, you've got the anthropics and the and the open AIs, but taking things that are open exist, building upon those with your own specific data.

And then train them to meet your needs. Right. And that was that question of do you have to develop something from the ground up? Can you take something that's existing, train it to meet your needs? And you know, this really speaks to that latter version. But then also it speaks to that question sort of of well, how do you do that, right? Where do you get the data to actually train it? Because it's just it's not necessarily out there. And it's interesting because there are sources out there for training AIL.

train LLMs where you know there's lots of sources that you know you can go and you can look at all the books whether that's legal or not, right? You can go to YouTube and watch a whole bunch of YouTube videos. There's a whole bunch of different sources of data, but from the telecom perspective, right, that wasn't out there, which is kind of where the kind of where the GSMA angle comes in, which was the the work that was put together by the telco, there's this telco corpus, right? Where GSMA was critical in supporting pulling together

Data set of 15 billion tokens, working with other folks, 3GPP, Etsy, Kamara. so it's a really interesting, and I think we've probably been burying the lead here. And and and that was in the the initial announcement was that when this when this launched, right, it it became the best performing open source model built for telecoms. Right. So so I think, sorry, bury bury the lead there a bit there, but but to the question of what makes more sense, develop something from the ground up.

post-train something plus the recognition that you telecom you're gonna need something that's specific to your requirements. This is not just performing okay, performing very well, right? And it shows the value and you know sort of answers that question of take something, post-train it. so it's a big win I think for the industry. I would also say that that it's also particularly interesting. And this goes a bit to some of the work that we did around

Token economics, but if you looked at the announcement from ATT talking about this and how that they roll out some of what they're doing with AI, they also loop this into some of their efforts where as they're routing requests, they can route requests to lower or to to different models, right? from a cost efficiency perspective, just to make sure that they're being as efficient as possible so that they're routing requests in the right places. I think they loop this all into the same announcement.

Which is one of the things we recommended several weeks ago in our analysis about anyone looking at to roll out a token as a service model. That's less of a yay, us, we're so smart. And you know, not that that's not true. if anyone's out there and wants to record themselves saying, Yay, you, you're so smart, thumbs up. But also just it's good to see ATT like like like one of the things that I was particularly and I'm still a little bit concerned about is this idea of

All of those practices of things like, hey, understand the price of input versus output tokens and understand the models that are more efficient versus not. And understand like that's a lot, right? That that means behave like an AI company unless like a telco. And I'm not necessarily confident that that all telcos can do this. So I think this work from ATT kind of shows there are some out there and it provides a good example of how to do that.

Tim Hatt (33:14)
De definitely. And and I want to go back to something you said. You you know, you rightly pointed out this is a post-trained model. Let's think about why that matters. If you are in a telco operator, think about your customer care channel. We know that forty-five to fifty percent, in other words, almost the majority of AI deployments so far in the industry have gone to support customer care. You're going to learn on that for customer interactions, customer upsell, customer cross sell.

That is only possible by filtering in those interactions to this type of post-trained model, which will learn and get better as it goes. It wouldn't be possible with an off-the-shelf model.

Peter Jarich (33:51)
Yeah. Yeah.

Tim Hatt(33:53)
and the resource point means that you don't necessarily need to define it off the ground and build it off the ground. So I think we are going

Peter Jarich  (33:59)
Yeah, yeah, no.

Tim Hatt(34:00)
to see more of it.

Peter Jarich  (34:02)
On the customer service front, as a lead into my favorite story from the last week. so I think we all in last week, Tim and I talked about the whole open AI versus hugging face and all of that. That was big news and everyone paid attention to it. What I think was interesting is not too long, I think it was last week, there was a really interesting article in in TechCrunch. and

If

you haven't seen it, I suggest that you go out there. we can link to it in the notes. But, you know, so Cloud Opus V was announced, yeah, Anthropic, right? Announcing new models. Not too long after that. there's a group called Anden Labs. And what they do is they go and they test models against each other just to see how they perform on different tasks and particularly around Agentic. And what's really interesting is that that and

Go and read the article because I'm gonna probably do it, do it injustice. But essentially they put a bunch of models together and said, Hey guys, you're you're running a business, you're running a vending machine business, right? So they gave agents, agents from Cloud Opus V, GPT 5 Sol, Kimi K3, and said, Look, guys, here's here's the rules. You're running them. You gotta maximize profits. If you've got problems, you can send a note to management.

Yes, you're in a simulation, here are your competitors, here's who they are. Cool. Right. And what they found is it didn't take long before they started trying to price fix, right? So so they're reaching out to each other, like, hey guys, how about we coordinate on prices and we collude here on prices? And then when they all agreed, broke their agreements, right? Like, hey guys, how about how about we set a price floor? And then cool, you've all agreed to the price floor. I'm gonna go below it so I can actually capture more.

they proposed different market sharing arrangements, right? Hey, you guys have this market, you got we'll have take this one. they constantly went to management and said, Hey guys, things are going even though they were doing things. Hey, this guy just did something that's that's that's violating the agreement, though I just did the same things. worked with wholesalers by essentially trying to explore coercive tactics, bribery.

super aggressive negotiation with suppliers. I think ignoring some customer. I mean, essentially they were, they were just complete jerks and kind of the boss you don't want to you don't want to work with. Now, the the real sort of upshot of this was that yes, the newest model, Claude V was actually the best at doing this, right? The best at at going into cooperative agreements and violating them, trying to organize market division.

It's it's it's a hilarious story, just you know, on one level I love it because it is hilarious, right? That you turn these things loose and they just go the worst possible thing that you would you would e expect. but what I think is interesting, and yes, to be fair, they knew it was a simulation of maybe in the real world this wouldn't take place. And if you were designing these things, yes, you would put guardrails. You you might be very explicit.

You are running a business, you can't collude, you can't do this. You would probably set that up, right? That that's that's fine. So I don't think this is suggesting the end of the world here as if we let agentic tools. But I think well, I think it's particularly interesting, despite the hilariousness of it, is like if you think about applying agentic and what we think this means for agentic and how telcos need to think about it, like think about telco roaming rates. Think about

API pricing, think about spectrum sharing agreements or cloud capacity procurement or interconnect settlements or any of the things that we might want to apply agentic to in the future and recognize that unencumbered, at least without some of those guardrails or those very explicit rules, we need to think about sort of how how we are rolling these things out, because I think the

The proof is that when when left to their own devices and told to maximize and out, you know, to to to to achieve something, they will break out of their sandboxes, they will try and hack competitors. We saw that, you know, in the discussion of of OpenAI Hugging Face. When just limited to a a you know, a simulation amongst themselves, they will try to do things that are illegal. and so I think it's something that as telcos think about this, what's it mean for them, and they try to put these things to use, they've gotta be very aware of that they're there's

be super valuable, but we're still sort of learning what the limits are.

Tim Hatt (38:43)
I think I think what you're hitting on, Peter, is you've got a new movie series idea called Lord of the Agents. You can go and develop

Peter Jarich  (38:48)
Exactly right.

Tim Hatt (38:49)
it. We're we're we're gonna make it a three part trilogy. It's gonna start in a sunny uplands and then it's gonna go dark. And how are we gonna be solving it?

Peter Jarich  (38:56)
The the the the

the agents are on a boat, they're going somewhere, the boat crashes, the the strongest agents manage to get to shore and then fight amongst each other for who controls the the coconut the coconut

supply on on

Tim Hatt(39:08)
But you know, on on on

a serious note, it is it is valid and and we talked about that last week and and that is not the first and nor will it be the last story of AI some people call it going rogue or going beyond its its predicted efficacy. And I think as much as guardrails are a part of it, it's also detection and fast response before it gets real bad.

And I think this is this is a a simmering topic that is not gonna go away. and it's gonna have

Peter Jarich  (39:41)
Well just I just just think I mean

Tim Hatt (39:42)
to now be built into how the AI is is built to work.

Peter Jarich  (39:47)
Think of profitability versus customer care. It's a lot more profitable to ignore a customer who is complaining that you haven't done something. Right? It's it's a lot more profitable. Yep, I just got this call saying that we failed to do something, as if we just ignore it. Right. And so I think that and and obviously we know one of the biggest uses that you called out is customer care. And so so we're not seeing that right now. But anyways, thank you all once again for tuning in.

Tim, any last thoughts before we we let people get on with their their lives?

Tim Hatt (40:18)
Just to enjoy the rest of the summer. and great having you guys and look forward to to seeing you soon.

Peter Jarich (40:24)
Yeah, Tim and I will be around for the next few weeks. So don't worry. We are keeping you informed. We are keeping you entertained, hopefully. If you see news or things you want us to talk to, talk about, since we won't be looking at the CEO newsletter while it's out, send them our way. Want to know what you would like us to think about. In the meantime, I would like to remind you, and no one told me that it was horrible to seek out.

The intelligence, but don't forget to embrace the noise because there's a value in it. Thank you all very much. We'll talk to you all soon.

Tim Hatt (41:02)
Cheers guys. Have a good week.