Episode 16: The Shifts Shaping Tech: Satellite Spectrum, Foldable Devices & the Race to Manage AI

Peter Jarich (00:00)
Hi everyone, and welcome to Intelligence and Noise, the GSMA Intelligence podcast, where we take a look at the week's news and talk about what you're probably already talking about, and give some of our insights into why what you're talking about matters, and then maybe what you're not talking about, and maybe why you should be. We build that off of the GSMA CEO newsletter, which, if you're getting it, means you're important. Congratulations. If not, we're doing the heavy lifting to tell you about it.

I'm Peter Jarich, the head of GSMA Intelligence, and as often, I've got Tim Hatt with me. Tim, how are you doing?

Tim Hatt (00:40)
Happy Monday, Peter. All good, how are you?

Peter Jarich (00:43)
Good. We're gonna pause here. They release these on Thursday, Tim. So let's just wind back and put a pause here so Yulia can edit that out, just in case we don't want her to say "Happy Monday." But she can decide the best way to deal with that.

Tim Hatt (00:51)
They do, they do.

Peter Jarich (01:01)
I'm sure.

Let's give it a five-second pause and then we'll get back. Yulia's listening in and probably laughing. She does that a lot.

As is often the case — I take that back. As is always the case, I'm going to take my privilege of kicking off with a bit of an anecdote. If you find these cringey, let us know. If you love them, we can make it longer.

But many of you guys know that last week I was just coming off of holiday. Holiday is one of the few times when I take a chance to read stuff just for pleasure. And it was a bit weird when folks at the resort were asking me, "Hey, what are you reading?" And I was saying I was reading about the history of the Yugo.

And if they were older and North American, they knew what that meant. If they were older and not North American, they thought I was talking about people from Yugoslavia. If they were younger, they had no idea what the heck I was talking about.

It was a book, believe it or not, about the history of the vehicle, the Yugo, which was a Yugoslav-made car that was imported into the US. It's a fascinating read — a bit about life in the '80s, a bit about how you bring a car brand to market, a bit about trying to sell economical cars when people didn't want that.

But what was interesting is, and for those of you who don't know, it was obviously sold as a very low-cost car. But one of the things they said was that what dealers were doing is that, in many cases, dealers were using this as bait.

So they would advertise that you could get this $4,000 or less car at their dealership, and people would come in and they would upsell them on something else. They might upsell them to a nicer used car, they might upsell them to a nicer car, but it was bait to bring them in. The cheap thing got them into the dealership, at which point they could upsell them on other things.

And I mention that because later on in the podcast, we might have an opportunity to talk about another upsell opportunity to get people into shops who may not end up buying the thing that they came into the store in the first place for.

So we will come back to that.

But the first piece of news, Tim. The first thing we're gonna talk about. So the two things that made it into the CEO newsletter this week. There was a story about Australia's continuing moves to put some limitations on social media, in this case limiting whether or not people see algorithmic results and if they can opt out of that.

But probably the bigger piece was the reported — and I say this, reported — partnership around MSS spectrum by the European operators.

So my understanding is this is based off the May proposal from the EU to break the EU's MSS spectrum — that's 30 megahertz of spectrum in the two gigahertz band — into three different blocks of 10 megahertz: 10 megahertz for EU government uses, 10 megahertz for commercial operators that are operating out of the EU, and then 10 sort of broader.

And this was really, for folks who didn't see the news, reporting from, I think Bloomberg broke it first, that said the big European operators — that would be Orange, Telefonica, Vodafone and DT — were considering a partnership to bid on it as a block and essentially be the European operator.

Did I get that right? And if so, why? What's the thinking here?

Tim Hatt (04:51)
You did get it right. This is a rumor.

You're one for one on that, Peter. Good work there.

It is a rumored announcement that would be of the big four mobile operators in Europe — Vodafone, Deutsche Telekom, Orange and Telefonica — that they would, according to the story, jointly bid for a portion of the upcoming European MSS spectrum at 2 gigahertz.

Now, the interest, I think, in the story, Peter, is that there's a few points of relevance here, even though it isn't yet confirmed.

But as you noted, Europe has designed a spectrum auction that's a little bit unique, actually, in the way they set it up, where the allocations are a third for government and critical services. Most of that would be serviced by the IR² plan constellation.

A third would be for commercial use, but reserved for European-domiciled participants. And a third is open market, meaning AST could bid, Starlink could bid, any other at-large bidder.

And the interest here — and we'll get into compare and contrast — but the interest here is that if this type of vehicle were to take shape, it would probably be going after that third that is reserved for European companies. And therefore there's commercial and sovereign implications.

Peter Jarich (06:16)
But I think — and you hit something there which I think is kind of important, right? — is this a reported or rumored, a rumored proposal, a rumored set of work by operators around a proposal.

So how exactly — we haven't seen the movement to actually say this is what the EU is going to do in terms of how it's going to be allocated. So we're talking about a rumor around a proposal.

And the fact that we're all talking about it so much highlights the stakes, you know, what's at stake here and that this is a big deal, right? Sort of an if, an if around an if, which tells you how really the profile of all this, right?

Tim Hatt (07:03)
100%. And let's just put some facts around the debate so that people understand the scale.

So we talk about those big four operators. They account for around 900 million mobile connections between them. It's a big number. I mean, that's coming up on like 15% of the global market share.

Two, and we show this in our quarterly satellite report that we release, they account for the lion's share of existing and planned D2D partnerships in Europe already. They have over 30 between them spread across different providers.

And why that matters is that if this type of move were to take shape, it would then require open procurement across the range of D2D providers.

Now, I think the interesting point here is when people start to compare this with the US JV announced by the operators four or five months ago.

And on the one hand, it's tempting to compare them, but actually when you peel back the onion, they are a little bit different.

And just at a high level, get your reaction on it. The US is mostly proposed as a joint access layer for satellite companies using existing spectrum. This is proposed to be a scale play to acquire new spectrum and really drive home that sovereignty agenda.

Peter Jarich (08:29)
Yeah, no, I mean, it's interesting. I think there's a similarity to the extent that we're talking about telcos getting together and partnering to do something on the NTN front.

You know, one is developing its own capacity, right, with the European operators. The other is really around joint collaboration to acquire and allocate capacity.

Right, of, "Hey, we're going to work with these folks, and here's how we're going to allocate or acquire capacity across them." So very, I mean, very different.

But what I think is interesting is just going back for me as I think about how this stuff might roll out.

You know, you talked about the sort of third tranche of spectrum that could be acquired by, if the proposal goes through, could be acquired by a Starlink, you know, SpaceX, Starlink, an AST.

But it's interesting because AST already has a pretty tight relationship with Vodafone. AST already has a tight relationship with operators in the US.

And so I don't want to say I worry, but I wonder how tricky does this make things, right? How tricky does this make things when you've got operator — I mean, so would you be pulling AST into that spectrum tranche in Europe? Or if so, would they have to divest of that?

Or would it be — it feels like it could become pretty tricky pretty quickly. And then even if you didn't have to divest that, would you, if I'm the other people in that partnership, the other operators, do I go, "Hey, hey, hey, like, does this — what does this do to your incentives as part of our tranche versus the others?"

It makes it kind of — it just feels like it's going, and none of these things, no operator partnerships are ever an easy thing, but it feels like this one's kind of fraught from the beginning.

Tim Hatt (10:33)
It does feel like that, and there is a lot of risk.

I think you bring up a good point around, you know, the Vodafone-AST example, because that kind of speaks to a situational awareness that this would have.

Like, go back to the comparison. Fundamentally, if this were to happen, it would be about trying to ensure that the EU-led players are acquiring the spectrum and that you were reducing the entry risk of a US company or another non-European country.

The US is the opposite. It's more trying to prevent the market power of one satellite.

If working on that basis, if you're Vodafone or if you're Telefonica, who has partnerships with OQ and a range of providers, how do you navigate those risks?

I don't think they know that yet. And I think the fact that the European Commission has not yet settled on the auction structure tells you a lot of uncertainty that is there.

Peter Jarich (11:37)
Yeah, and I think that just adds to — I mean, so, like any of these things, right? And we're talking about acquiring spectrum to support NTN services, right?

You've got all your basics of costs of network integration, of how do you handle roaming, of the fact that although they would potentially acquire spectrum across the EU, they do compete in some of the same markets, right?

And so, you know, you've got that of how do we operationally make that work. And I trust that they can figure that out. Right, right. That's fine. They've got partnerships.

On top of that, I'm sure there'd probably be some competitive issues, or the EU would look at that to go, "Hey, are you guys coming together, and what does this mean?"

No idea what that would look like. But for me, and again, I feel that that could probably be something that they've thought about.

I think the tricky part, though, is how do you deal with existing assets which are not, you know, which are not European, right? I mean, you say, "Hey, look, you've got a relationship with this satellite company. How does that fit with the plan that you need this to solve?"

Now, on the other hand, the whole — as I understand it, right, the key argument for a Vodafone piece of work with an AST is to build a sovereign capability.

So there's no reason why you can't build sovereign capabilities with people that aren't European, but it becomes difficult when the, if, as the proposal says, there's a requirement for that spectrum to be held by, you know, European or EU-domiciled company.

So yeah, it's gonna be — it's gonna be a fun one.

Tim Hatt (13:12)
Lots of questions still to be answered, but if nothing else, it tells us the urgency of moving in this space.

And would we have seen this two years ago? I'm not so sure.

Now that we've seen the level of activity, and especially the SpaceX IPO, things are moving a lot faster.

Peter Jarich (13:31)
Yeah, yeah. And I think — and I think also because, and to be fair, right, we're just talking about — I mean, at the end of the day, and it's all important, but at the end of the day we're talking about what, 10 megahertz of spectrum?

I mean, so it's not like we're talking about, you know, giant swaths. But as we see folks like Starlink doing some, or at least talking about — and we've talked about this in the past in the podcast — talking about interesting architectures like leveraging small cells alongside the satellite connection.

You can guess that if I'm an operator, I might find that particularly interesting.

So we'll see, we'll see where that goes.

So, over the stories. And for folks who don't know the stories — and you wouldn't know because it's all behind the scenes — two stories were proposed, and if you've been watching the news, you probably might have guessed at least what one of them was that didn't make it into the newsletter.

One of them is really around a broader focus around AI becoming more of an ecosystem play.

So the story was around Qualcomm's work with AWS to find a customer for its AI silicon.

And the broader story, right, is that we need to be thinking about AI as obviously — we all talk about the model makers and we think about the AI data centers, but it's obviously an ecosystem play.

And we talked about that as we talked about the Hugging Face acquisition by NVIDIA, really this reminder. And I think maybe that's why it probably wasn't worth highlighting.

I think we see this every day, that AI is more of an ecosystem play.

But probably the story that is worth covering, and didn't quite make it into it, was the story that dominated most consumer gadget technology discussions and media over the last week: the launch of the iPhone Duo.

So for those folks who went on Kalshi and made their bets on what it would be called, I hope a lot of you lost money, just because I just don't bet on Kalshi. So I kind of feel bad. You know, a little bit of FOMO there.

No. I know that that wasn't, you know, one of the names people were bandying about.

Not surprisingly, I think this was well anticipated. People kind of knew that we were gonna see a folding iPhone launch. They kind of had a feeling that it would be the passport form, the book form.

I think most folks thought it would probably start around the $2,000 mark, and in fact it did.

Prices go from $2,000 up to, I think it's $3,200. $3,200 if you want the version with two terabytes of memory.

Go figure. The two terabytes of memory alone is probably what justifies the cost there.

I mean, from my perspective, the quick background here is that, you know, if you look globally, and this is the story that I think we heard a lot about, right, is that, you know, from a global context, yeah, foldables have been around for a while.

You look at the iteration number that Samsung is on. It's been years upon years that we've been talking about this. You know, eight or so years, almost a decade now.

There's still a rarity when you go around. You see a friend or you see someone in your office who has one, you're probably still looking at scant numbers.

It's because I think most market numbers or people who track forecasts state around 2%, 2% of the market. So still a small share, and that's probably dominated in, you know, developed versus developing.

But against this background, you've got Samsung, who recently said that they saw 30% growth in their foldable market, their foldable sort of shipments last year.

And more importantly, they said that I think it was a quarter of the people who are buying their foldables are coming from other companies.

So it's a way to steal customers. And if I'm Apple, then I go, "Hey, I can't ignore this forever."

Is this a way for me to steal folks from other manufacturers?

And let's not forget, if we look at Apple's revenues. Okay, Tim, here's the quiz part. We're gonna see if you're one for one. What share of Apple revenue comes from the iPhone?

Tim Hatt (18:01)
I believe, well, certainly of income, I think it's around 55 or 60%.

Peter Jarich (18:05)
Yeah, close. It's 50 — at least at last. That's much closer than I think most people — you'd probably win in a pub quiz.

It is 52.6%. Fifty—

Tim Hatt (18:14)
I'm useful for something. I could do it, but—

Peter Jarich (18:19)
52.6 rounds up to 55. So we're gonna give it to you.

Surprisingly, iPad around 6%. I would have thought so, yeah.

So when you've got over half of your revenue coming from something and you've got competitors who are claiming big growth, I mean, I guess the big question is, Tim, are you going to buy one?

Or maybe more importantly, will your wife divorce you if you come home and you're and you're—

Tim Hatt (18:45)
It's on my Christmas list of Santa, so I'm hoping it shows up in my stocking.

I was gonna ask you the same question, but maybe we can get into the Secret Santa for that one. Price tag might be a little high, but, you know, we can look at it.

Peter Jarich (18:56)
Yeah. We'll add our Director General to our Secret Santa pool. You never know. Maybe he can pony up.

Tim Hatt (19:03)
Yeah.

But it is interesting. And you point out a couple things there that I think are worth mentioning.

You know, number one, the price tag is high, but I don't think Apple cares about that because they have such high customer loyalty already, that for them this is sort of an entry point to a class of device they think probably has legs.

You know, the interesting thing for me is the timing of this.

You know, Apple is no stranger to being late to the party on various product categories. Let's not forget the iPhone was not the first smartphone, far from it. Nokia had been experimenting, Samsung — it came late and changed the world, but it was not the first.

We saw that with VR, arguably, we saw it with AI. And I think we're now seeing it with foldables.

But what are they looking at here? Ultimately, they've got a kind of get-out-of-jail-free card. Other companies don't have it.

Which is the fact, 90% of their customers come back every single time to get another iPhone, and it's the most dependable revenue stream in the world.

So even if they screw up on the first iteration of this, and I think they've probably taken time to make sure that doesn't happen, they've got room to give.

Peter Jarich (20:19)
No, you hit it. There's, I think, huge expectations about what they do and the quality side of things, right?

And you've seen that, I think, to some extent they could sit back and look at the complaints. I think I've read a bunch of the reviews.

And not surprisingly, I think what people expected was, yeah, you know, the thing's going to feel solid. It's going to have a seamless experience between the front cover and when you unfold it. It's going to be really great with the applications, and it was.

But then small things like, you know, if you've used a folding phone, the crease is always an issue. And you see the crease and you feel the crease, and the crease is — you can sort of ignore it, but you can't ignore it.

They're like, "Look, with this thing, the crease kind of disappeared." And you can feel it, you can tell it's there, but you don't really see it because they use texture on the screen.

So it's little things like that, right? Where they can focus on those details.

But you hit it. They also get a pass because, you know, they've got some time to get it right.

But also, this thing's gonna be a status symbol.

You know, and I think we all sort of saw this when they launched the last iPhone and, you know, the Pro Edition only came out in the orange.

You knew that if you saw someone with the orange iPhone, that it was a bit of a, you know, it was a bit of a status play, right? Yeah, hey, I've got the cool one, right?

And you saw some other companies and developed phones that look just like it in orange.

And I can't say that, you know, I didn't look at those and go, okay, if someone was carrying that, I wouldn't know. And would I think differently of them? I'm not quite sure.

But I think it's — I think for me, probably the two biggest things for them are gonna be, you know, they're going to sell a bunch of these.

And I've seen some analysts forecast at around five million.

So five million times $2,000 minimum, maybe round that up. We're talking about $11–12 billion if they don't just sell the $2,000 version.

So that's not bad, and they'll sell it to people who wanna be early, early, early users, status symbol folks.

You will look differently upon the person who has the iPhone Duo when you see them out in the wild.

Tim Hatt (22:41)
You will. And is my big question here, and I'd be interested in your perspective on this, is does it change the market?

Do you get to a point where Apple itself and its halo effect inflects things in a way that wasn't previously possible?

And right now, 2% market share, you're right, that's essentially nothing in the smartphone — well, it's not nothing, but it's very small.

But do you get to a point where if Apple eventually converts over, let's say five to seven years, even half their base, they've got 20% market share?

You're all of a sudden in a situation where foldables, you apply that to their groups, or maybe 15 to 20% of shipments.

That's an open question, but I think that's something we gotta watch.

Peter Jarich (23:20)
Yeah, yeah, yeah.

So I think there's a couple routes for that to happen, if we think about — or a couple of vectors on which this could impact the market.

One, yeah, mainstream validation, right? So it sets this idea that, okay, this is a thing that everyone needs to have. Apple's doing it.

Does it make other folks sort of get in? Does it make them up their game, right? Going, "Okay, yeah, I was kind of niche. You were gonna buy these just to be the person with a foldable phone, but now they've kind of set the tone in terms of how my software needs to work, how I need to think about treating the physical side of things."

I don't think it's gonna get to a point where it drives competition so much that we see prices going down.

But I think the bigger piece would be, you know, we know Apple's got the developer ecosystem down where people jump to develop for the iPhone when they can.

Does a larger — you know, if we see foldables now becoming something where iOS developers start prioritizing, does it make them think more about split-screen development?

Does it make them think about doing things for different angles for different screen sizes?

Which then you see that, and I think that's where you see the sort of rising tide lifts all boats potential effect, where, okay, developers kind of get on board to support the iPhone because it's the iPhone.

And then that benefits everyone.

So it could be — I mean, I think that's probably the best-case scenario here, because I'm not expecting prices to fall anytime soon.

Tim Hatt (24:55)
No, no. I think you're 100% right.

I think it's gonna have to be something we see in the coming kind of 12 to 18 months to get an initial indication this Christmas as well, and go from there.

The last thing I think is an open question that we'll probably come back to in other context on different episodes is what does it mean for processing and chip capabilities in the AI era?

We know right now the silicon readiness is there for some AI capabilities, but not others. Does this throw a new factor into the mix?

Peter Jarich (25:25)
Yeah, yeah.

And then I think also, for those telcos that still have retail shops, does this get people to come into those shops?

I think we heard this when the first round — you know, when the first round of foldables came out. I think you'll remember the work that we did around retail and retail presence.

I remember hearing from folks going, "Look, whether or not they're going to buy that folding phone, they're coming in to actually play with it, to touch it, to go, okay, here's how this thing works."

And that's a touch point to engage them on any number of things, whether or not they actually buy that phone.

So time will tell.

But with that, let's move on to the next story.

And this is now, for those folks who are familiar with our format here, this is the "On the Road and On Our Minds," aka the Willie Nelson Memorial edition or segment, where we talk about either our travels or things that are just smaller news pieces. Neither of these are sort of linked to them.

So maybe I'll go first on this one. It's probably the smaller of the two stories, if I remember correctly what you want to talk about.

It was Nokia's Cognitive Operations launch. So if you saw this, it's a new solution that Nokia rolled out for a number of vertical industries, mostly focused on mining, public safety and defence right now, with a focus on how you're looking at combining sort of critical communications infrastructure with edge and AI.

So edge AI for critical communications verticals.

Think about all the classic AI — all the AI use cases that people are talking about. So site and asset intelligence, video analytics, maintenance prediction, downtime prediction.

If you look at the launch and you look at the press release, there's a box that you can launch on site. It's obviously hardened for those use cases. You could deploy it on the cloud with Azure.

I think for me, and again, like I said, this is, you know, we don't need to spend too much time on it because I think some of the main points are there.

This factors into a lot of the work that we've done not too long ago around edge AI and all the importance of launching edge AI.

Most importantly for us, and this goes to some of the work that we did — and we can flag some of the reports in the show notes — that as much as people talk about edge, and the first thing they talk about with edge is latency, that for edge AI applications, much more than latency, it's things like resilience.

Right, or, you know, resilience, or the cost — keeping backhaul costs or keeping energy costs in check, or whatever it is.

But in this case, resilience and having that thing on-prem so that if something goes down, you've actually got that AI workload on-prem, as well as all the value of data sovereignty, data privacy.

Again, feeds into all of that, particularly with those critical use cases like public safety, defence and mining.

But what I think is interesting is, and of course, this fits into the whole focus of B2B for telco being the 5G monetization side of things.

You know, for me, though, I will say, a little bit of that, the head-scratcher here, though, is where this fits into the broader story around Nokia and what they're doing with their enterprise business.

I think for those folks who, you know, those folks listening, you probably remember the story, but it was 2025 when Nokia moved its enterprise campus edge business into its portfolio business segment, which was meant to be a holder for things that were not strategic.

It was Q2 of this year that they basically classified the Enterprise Campus Edge business as sort of for sale.

And so where exactly this fits in and how exactly Nokia is looking at verticals, I think it's another reminder that they're still important.

And that, as much as you may — it's a bit of a quandary there, or a bit of a dilemma, where it may not be core and strategic, but it's still important.

I think it's a reminder.

I don't know. Tim, you did a lot of the work with us on the edge side of things. What's your — moving the needle, not moving the needle?

Tim Hatt (30:12)
Well, on its own, I think it's one of several things that are moving the needle.

I think the jury is now increasingly — I think the rationale is much clearer for an edge AI situation.

And I think ultimately this gotta play through a set of proof points for enterprise customers, where they see the resiliency in practice, they see the efficiency gains, they see the TCO savings versus running these workloads in the cloud compared to on-prem or further out the edge.

And I think we're getting to a point, especially with 5G SA developments, where — and now AI factory deployments — investments are being made to put it into practice from the operators.

As you say, Nokia has changed their strategy to support that industry move. Their competitors have as well.

I think ultimately the proof will be in what we see happening with the B2B revenues and AI deployments for the operators coming in a couple of years.

Peter Jarich (31:12)
Yeah, and these are big companies, obviously, where when you've got a product, whatever division is running it and launching it, in an ideal world connects with everyone, but that's always hard.

So I was a little bit surprised, but also I was surprised that we didn't see — but wouldn't be surprised if we see in the future — you know, as much as there's been the discussion of putting the enterprise campus edge business up for sale.

I think there's also a follow-on that, yes, but we'll still sell RAN equipment that could be used to support private networks.

Bit surprised that we didn't see a connection there, right?

That obviously Nokia's been at the forefront of talking about AI RAN and developing RAN capabilities off of AI, off of AI.

So feels like either there was a potentially missed opportunity here to connect this to that, saying, "Hey, all this could be combined. You've got a nice hardened unit that can support the RAN needs, as well as the AI needs of these verticals."

Maybe that's coming. Maybe we're talking about working with different suppliers.

I guess that's what makes it fun. Time will tell.

Tim Hatt (32:20)
So—

Peter Jarich (32:21)
It feels like we've only, you know, we've mentioned AI sparingly on this call. So it feels like we're gonna get in trouble with someone if we don't talk AI.

So what did you want to focus on, Tim, that didn't get enough attention?

Tim Hatt (32:37)
Well, I wanted to bring up the essay that was released by the Anthropic CEO, Dario Amodei, over the last few days.

And just to be, just for our listeners here, if you haven't seen it, or even if you have, this is a kind of expose slash essay published by the CEO of Anthropic to highlight what he sees as risks in AI development that need to be contained through cross-sector collaboration, and indeed, government regulation.

And I think it's interesting because Anthropic, of course, is one of a handful of AI leader companies that are at the forefront of frontier model development, has an IPO in the works, is a major provider to some of the biggest companies and governments in the world, but equally has been very vocal, increasingly vocal, around the security risks.

And from his essay, what was interesting is his main point was essentially that AI's productivity gains are going so fast that the risks are growing at a point where we can't necessarily contain them.

And therefore, he's put out kind of three key recommendations.

One of which is to slow down development and allow outside evaluators into frontier model companies to audit those.

Two is alignment with government around AI regulation domestically.

And then three is an international treaty or set of practices.

I've got a view on that, but I think it's an interesting timing given the IPO, given what's happening in China.

Peter Jarich (34:16)
I think so.

I think part of what's key is you talking about this as an essay, because it is not small, right? It is a long piece of work that, you know, I don't want to say it's well thought out because I don't know how much time, but it feels well thought out.

And I think Dario takes pains, I think, to highlight that it's well thought out.

You sort of see from the beginning going, "Okay, AI is really, really important, and I've written about this before. But there's risks to AI, which I've written about before," right?

So you sort of, like, right up front, there's a preface of it's good, it's bad. I've already talked about a lot of this, right?

And then, you know, sort of the question is: okay, I've talked about a lot of this. What has changed?

And to your point about AI productivity, it's less about AI's support for productivity gains as much as it is the productivity about developing AI.

Right. And this whole recursive learning where AI is — you're seeing the pace of development and then the risks.

And I think I give them credit for calling out the risks from a bunch of folks.

But sort of he leads with the — and, you know, this summer the thing that made me think about this was the Hugging Face attack from OpenAI, right?

It's like, yeah, yeah, yeah. There's stuff that Anthropic's done as well. Don't, don't get me wrong.

But the thing that made me really think about this, and I think you're right, I think then you've got his sort of three pieces: the embedded evaluators, where you want people to sit in here and watch what's going on, and we're unilaterally committing to this right now.

Tim Hatt (35:51)
Yep. Yep. Yep.

Peter Jarich (35:54)
What I think is interesting is you talked about, and this goes to your point about sort of the politics here, it wasn't positioned as, you know, okay, one, embedded evaluators, number two, coordination between frontier model makers.

No, it was democratic coordination, right?

It was very much positioned as coordination between all the people here in democratic countries.

Not saying who we're not talking about, but then of course, number three, global coordination was talked about as the US and other democratic governments coordinating with authoritarian governments, right?

It wasn't just coordinating, it was very much positioned as, you know, democracies versus authoritarian governments.

And, you know, China's mentioned later on.

But it's a very clear message.

But as much as you could sort of see this as, okay, are we trying to talk up sort of maybe some US principles here?

You know, I think on the one hand you can feel that, that you mentioned the timing, right?

It's not just the IPO timing, it's also the fact that they released data about other ways that they've seen AI being misused in their own AI, including weapons development.

Which, you know, they've been a — you know, and to be fair, they've highlighted responsible AI from the very beginning, right?

But then, two, if there's any question this is about sort of getting on board with US principles or the US administration, the fact that the US administration, or at least the president, has said, "Nah, don't worry, AI's fine," has downplayed any of the AI risks.

Tim Hatt (37:49)
Yeah.

Peter Jarich (37:50)
Yeah, is sort of fun, right?

Tim Hatt (37:52)
It's—

Peter Jarich (37:53)
It—

Tim Hatt (37:54)
And to your point, I can't really think of an instance in technology's recent history where a government has been at the point of saying, "Hey, we don't need to regulate this," and the NGOs do saying, "Hey, I think you do."

And so the open question becomes, how do you get that alignment? What does that mean for the IPO time?

I mean, of course, there's probably ripple effects into the telco sector as well. We could talk about those in another episode.

Peter Jarich (38:15)
I mean, when you've got — I mean, we've got, and I've talked about the folks that, you know, the report they put out, which they call their most detailed threat intelligence report to date, right?

They're talking about users in Iran, Russia, China, Yemen using this.

So I think that they see, like, and it's not just them, they see some of the potential here, and they can't not see the fact that the general population has questions about AI, and AI is becoming less and less popular.

And I mean that's whether it's development of AI data centers or data centers in general or AI use, that if nothing else, if they want to have consumer enterprise businesses and sell to folks, they need to make sure that they're being clear that this stuff that they're at least — they're at least doing the right thing, or at least trying to.

I mean, I guess I have some questions about what you can do globally.

If they're gonna have problems with what they can do in the US, I wonder what they can do globally.

But it seems like, in not long after launching that, OpenAI said, yes, we agree to this. We think it makes sense. We agree to evaluators.

Elon Musk, in theory, still head of xAI, said yep — or at least I think he said Dario is right.

So who knows what he meant.

Maybe he meant Dario's right about any number of things that he said, but it seems like the US industry is on board.

The devil will always be in the details, but fingers crossed.

Tim Hatt (39:53)
We will see where it goes. A rare show of industry alignment.

Peter Jarich (39:57)
Yeah.

So with that, I think, look forward to next week.

We don't have any travel imminent in the next week, but we're gonna be busy not too long after that as we head to the end of September and early October, but we'll share that for next week.

So let me just say for now, thank you all for tuning in.

Again, if you like it, leave a rating somewhere on your favorite podcast app. Send us a note about how we can do better, what we can do, what we can do at all if we're not doing it right now.

Thank you, Tim.

Tim Hatt (40:36)
Thank you.

Peter Jarich (40:37)
And hope to see you all next week.

And in the meantime, let me remind you that I will probably not get this outro correct ever.

But I'll say: keep searching for the intelligence, but don't be afraid to embrace a little noise.

Thank you all very much, and we'll see you soon.

Tim Hatt (41:00)
Cheers, everybody.