5G Factor: Making AI Open, Responsible, and Transparent, Episode 62

Hello and welcome everyone to the 5G Factor. I'm Ron Westfall, research director here at the Futurum Group, and I'm joined here today by my distinguished colleague, Olivier Blanchard, and he's our research director focused on very important areas such as devices and semiconductors, including 5G naturally. And today we will be focusing on major 5G ecosystem developments that have caught our eye.

And that will actually include, you guessed it, ai. And we're actually looking at a distinct aspect of it. It really is going to be a conversation about what is being characterized and being promoted as responsible ai, as well as transparent ai.

And you know, the variations very close to that. And so with that, Olivier, thank you so much for joining. How have things been coming along since our last episode together?

You know, we've had eclipses, we've had, uh, we've had monsoons, we've had earthquakes, but I'm good. Yeah, we've lived to tail the tail. And so yes, welcome to the post eclipse era, and hopefully, uh, AI will play a major role here in making it just that a more responsible and say environmentally friendly, uh, era as a result.

And so, with that in mind, uh, we recently participated in the Intel Vision 2024 event. And really at that event, it was Intel really coming out and presenting its AI proposition on a portfolio wide, uh, basis. And in fact, on an ecosystem wide basis that is pretty much underlying, it's bringing AI everywhere vision to basically all the players out there, including partners, customers, anybody who's involved with the technical aspects of making AI work.

And so what I think was very interesting is that it focused on critical sectors that included, you know, uh, finance, manufacturing, healthcare, but also there was, uh, plenty of material related to the mobile ecosystem, especially when it came to generative AI capabilities that is gen ai. And also, you know, taking it from experimental phases to full scale implementation. And that really is the bottom line.

It's about going from, you know, proof of concept and bringing it to productization that is having an impact on, you know, the organization's, uh, say finances and monetization. In fact, uh, Intel shared, I think, a very important takeaway with their own research. Only 10% of organizations have taken their gen ai, uh, basically, you know, tire kicking proof of concepts, uh, and so forth, and have them in a production environment today.

So we're very much at the front end of this. This is something that is clearly going to be evolving dramatically in 2024 and beyond. But also I think that's important to note that we want to see more than 10%.

In fact, some folks even thought, wow, 10% is a, uh, pretty, uh, progressive number. But that aside, I think what is important is what Intel is offering. And to spearhead it, Intel, uh, introduced the GATI three accelerator that is designed really to enable open community based software and open standard or open industry standard ethernet to really enable AI systems to take off to get past that, say 10% threshold and make say, gen AI capabilities more production ready and friendly.

And so how is Intel doing that? Well, they're basically offering an architecture that delivers improved gen AI performance and efficiency, especially in relation to existing implementations. And as we know, that means NVIDIA's AI capabilities, for example, the H 100, uh, GPUs, but also H 200 and, you know, on the horizon, the Blackwell offering.

And so this is gonna be a very interesting, uh, I say contest because in addition to the fact that we know that there's a supply shortage, really when it comes to Nvidia GPUs, we're seeing Intel coming up and saying, Hey, look, not only can we address, uh, this, uh, fundamental issue, but we can come in with something that, uh, can be actually more rewarding from say, a price performance and other aspects. That is, again, the openness, being able to leverage capabilities that align, say, with ethernet networking. And so with that in mind, it's designed to allow activation of all the engines that are being used out there in parallel, that includes the matrix boltsy multiplication engine or MME sensor processor cores, TPCs, as well as network interface cards, or quite simply nicks.

So we're all familiar, uh, with these, uh, engines and how they fundamentally need to come together to, uh, dial gen AI to do its thing in a optimized fashion. And so the key features include, first of all, an AI dedicated compute engine. And so the Intel Gouty three accelerator is purpose built just for that high performance, high efficiency gen AI compute.

7 terabytes of memory bandwidth, N 96 megabytes of onboard static random access, MA memory or sram. And why this is important is because we're seeing today that it's really the memory supply that's kind of wagging, or it is in many key ways wagging the prices of GPUs and how they are built and put together. And so what Intel is proposing is a more efficient way to use memory, and also quite simply a more cost effective way to use memory.

And so this could be, I think, one of the key differentiators as to why Intel when it comes to adopting GOUTY three Accel AI accelerators in combination with Zon CPUs. Now also, in addition, they're all looking at quite simply efficient scaling force, specifically enterprise environments, which also includes, and we'll touch on this operators as well as other mobile ecosystem players. And so that's an offering that has 24, uh, 200 gigabit ethernet ports that are basically, uh, integrated to every GOUTY three accelerator.

Next up is the fact that, again, it's the openness, it's an alternative to say Nvidia Cuda approach. And that is something I think that is a very, uh, keen out there. That's not to say that, you know, supporting say NVIDIA's, uh, InfiniBand capabilities is not gonna go away overnight, but I think, uh, including Nvidia itself is seeing the writing on the wall, this is something that's gonna increasingly become an ethernet fabric that's going to be required to enable this heavy lifting of, you know, for example, gen AI and ai, uh, training as well as overall inferencing.

Uh, so this is something that Intel from, I believe from the inception, can take advantage of. It doesn't have any in Taliban say, uh, legacy issues, uh, to really address in terms of how to move forward in an optimal way. And so this naturally includes integrating Pie torch framework, and that's, uh, as well as, uh, having it provide a optimized hugging face community-based models.

And we find that this is really the most common AI framework out there for Gen AI developers. And that's gonna be simply a key aspect here is, you know, getting the developers on and making sure that they can have a good experience in terms of, you know, enabling gen AI capabilities using, uh, the Intel, uh, platform and also the gout E three PCIE, you know, basically it's the peripheral component interconnect express card that is really designed for lowering power. And obviously that's gonna be another major aspect as we've seen, uh, current GPU cluster designs can take quite a lot of power to really operate.

Yes, there are some improvements on the horizon, but here is a way to really, I think, uh, deliver even more power efficiency within a reasonable timeframe that is within Calendar 2024. Now, the direct mobile e ecosystem impacts include the fact that Barty Airtel is on board using, uh, Intel's AI technology across the portfolio. And so really it's aimed at naturally leveraging their existing telecom data to enhance its overall a AI capabilities to improve the experiences of its customers and partners.

So this is, I think, a, a a direct example of how a major operator is, you know, basically operating as a large enterprise and is already looking at ways to leverage Intel's new capabilities. In addition, emphasis, uh, was also, uh, uh, prioritized as a new customer onboard, and as we know, they provide digital services and consulting, including the targeting of the mobile ecosystem as well as operators. And they're basically using Intel technologies that stem across the fourth and fifth gen Intel Xon processes, as well as Ality two AI accelerator today, as well as Core Ultra to support Emphasis Topaz, which is an AI first set of services solutions, and platforms that seek to accelerate business value using gen AI technologies.

So yeah, that's a lot of information and, and that's actually a summarization of, you know, some of the key takeaways from one of the major announcements. And so I'll stop there because I know there's a lot more to talk about here. And Olivier, from your view, what Intel's doing in terms of bringing AI everywhere and how it's approaching it, and it's an impact in the 5G ecosystem, what are, you know, some of your, uh, key takeaways?

Well, that, that was quite the thorough summary I have to say. Um, and if that was the summary, that's, that's, that's terrifying. No, you know, it's, it's, it's pretty great intel.

Well, okay, two things. One, uh, I just wanna throw in there that we talk about NVIDIA and the H 100, H 200, um, and, and other devices or other, uh, products, uh, on, on the, on the horizon. But I, I, I do want to give a shout out to a MD because they have the MI 300 x, uh, solution, which is also pretty good equivalent, somewhere between the, uh, the H 100 and the H 200, depending on what you wanna do, uh, that's also part of the ecosystem, and that can, and I think will, uh, fill some of the, uh, the demand gaps that, uh, that Nvidia can't fill.

So I, I know that we just kind of like glance over a MD quite a bit, um, but, and, you know, it's, they, they have a good product out there as well that's, that's gaining traction. So that's, that's one kind of like a little footnote that I'm, that's always in the back of my mind when we talk about Nvidia, sort of like this, you know, I know they have 90% of, of the, um, or 90 plus percent of the market right now, but, you know, not necessarily forever. Uh, the second thing though, more importantly is, um, I, I am, I, I do like, um, the sort of not necessarily use case specific design and thinking from pretty much every other, uh, silicon makers, like, how could we compete against Nvidia or how can we compete alongside Nvidia?

Um, and, and so this, this sort of focus on specific verticals, on specific types of applications, on the range of performance and, and finding products that are gonna be not necessarily overkill, uh, and not be underwhelming, but fit in that, in that framework of this is the type of application, this is the kind of performance and bandwidth that we're looking at. And so we're gonna create products specifically to fit in that, in that tier, and just sort of like, create products and, and pro and solutions like combined solutions, um, that aren't necessarily premium to, uh, commodity, but that, that are extremely well tuned to fit in these, in these boxes and these blocks, depending on, on what you're trying to do. So it could be, you know, the, the operators, it could be like 5G Network stuff.

It could be, you know, inference work. It could be, you know, training on LLMs, but maybe not like the, the super, like huge lms. It could be even, um, uh, the lms, the, uh, the mixed models, not just the, uh, the language models.

So, um, what Intel is doing, I think is smart. I, I like their, I don't know that I'd call it a diversification strategy, um, but definitely a more organic, more use case focused approach, uh, to, um, uh, to generative ai, to fabs even, uh, just what Intel is doing to kind of get out of its rut of the last few years and sort of, I wouldn't say jumpstarts, uh, it's, uh, it's revenue engine, it's business engine, but definitely, uh, create more on ramps to opportunities is, is really smart. And this is a really example, a good example of, uh, of Intel doing that and it's ability to execute.

Um, I think what's, what people sometimes forget about Intel, even if they don't necessarily have the, number one, the best solution that everybody wants is they're extremely good at executing. Uh, they're extremely good at finding those niches and those on ramps, uh, to get back in the game when, uh, when, when they've taken out their, their eye off the ball for a couple of years, which happens to pretty much every company. And so, um, this was kind of exciting news in, in my view, and it's exciting news for the carriers as well, I think.

Yeah, in fact, I think those were all, uh, very key topics that were at play at the event. It's like, we have to look at Intel overall, and we know, okay, when it comes to the financials, it's kind of a 90 day scorecard, but that's different from the long view, you know, the strategic implications of building foundries here in the us. And so that makes Intel unique in terms of how they are, you know, uh, competing against the other players.

And that, you know, I think will prove a very fundamental difference, uh, further down the line to the point where these competitors will be turning to Intel for, you know, their Foundry work. And I think that it was an excellent point, Oli, about Intel executing, because we know, uh, that a few months back, uh, Intel basically unveiled, it's bringing AI Everywhere campaign. And here at Intel Vision, it was a, okay, this is how we're executing, these are the details in terms of our portfolio, our partnerships, you know, our overall strategy to really make, you know, enterprise AI not only a, uh, production ready, um, proposition, but to, you know, fundamentally ensure, uh, positive business outcomes are certainly improving them.

And so it's pretty much the first inning in terms of, you know, what is going on here in terms of the competitive mix. And, you know, uh, answering the question, why Intel? And that certainly played into their emphasis on the openness aspect.

I think that is going to resonate more and more now, clearly, uh, for example, open Vno will play an integral role in enabling not just the developers, but also other parties to support these capabilities. And that aligns with, you know, what we kicked off with, you know, responsible ai, uh, transparent ai. You definitely need that openness built in as part of, you know, uh, making sure that those types of initiatives and strategic objectives actually are going to, you know, uh, be built in, uh, organic as, as you put it.

Or, uh, quite simply, you know, make sure that, uh, these outcomes are going to minimize things like drifting and hallucinations, but also, you know, making sure that the bad actors don't flip this on the ecosystem. Something that the rest of the good guys, uh, which I would count as us, uh, would certainly not want to see. And so, with that in mind, let's turn to, I think another, uh, I think important, uh, I guess you can say coming out that, uh, was aligned with this, and that is in a recent blog, Durga Malti, the Qualcomm, um, SVP and GM for Technology Planning and Ed Solutions, detailed how Qualcomm's vision is for shaping the future of AI responsibly.

Now, they've been, uh, certainly emphasizing their commitment to AI for years now, and this is definitely part of their NDNA, and I think that's, this is something that they've certainly been paying attention to, but I think it's also helpful that, uh, with this blog, that it's a sharpening of how Qualcomm is going to make sure, responsible AI is going to be primal in terms of how AI will evolve, including naturally, uh, gen ai. And so one of the core principles of responsible AI is privacy and security. Uh, I think that's understood across the board, but it's like, okay, how can this be quite simply implemented?

And as AI systems collect and analyze, you know, these vast amounts of data, it's going to be essential to protect individual's privacy rights, and also ensure the security of sensitive information. And so, from our view for, uh, by promoting transparency, Qualcomm is targeting the fostering of trust and also enabling users to make informed decisions about AI technologies, ensuring that they align with their values and expectations. So we definitely need to basically dial back some the wild West aspects here.

It's inevitable when we have these innovation bursts, but I think this is gonna be quite simply essential in terms of making sure that AI delivers on an optimal basis, but also, you know, as promised by many of the key players out there. So what Qualcomm is doing in a partnership with, uh, true Epic and with the Coalition for Content Providence and Authenticity, or C two pa, no, not, it's not a Star Wars robot, but C two pa, is that they're advancing the way they verify the authenticity of, for example, photos. And so we know that's gonna be very important when it comes to, you know, minimizing fakes.

And that's gonna be, uh, certainly gonna be a factor, for example, in the 2024 elections here in the us but other parts of the world. So this is something that's vital. And, uh, by using Snapdragon platforms, we see that the TE technology can create a cryptographic seal around photos taken from a smartphone.

And so what this does is that the seal not only includes essential meta metadata, like the date, the time and location, but it can also verify if AI was used and the specific type that was employed. I think many people would fully appreciate that, and thus prevent some of the games that are going on out there with deep fix and so forth. Now, for example, if generative AI was used to manipulate the image, the digital seal can accurately detect it, and even during the transfer of the image to another device, the certificate remains intact ensuring the preservation of that integrity.

And so from my view, this collaboration exemplifies Qualcomm's commitment to transparency and upholding the integrity of content to the era of gen ai. And so with that, I know Olivia, you've been certainly looking at this in detail. Uh, what's your view on, you know, Qualcomm coming out here and say, here are some, uh, key ways to make sure that AI is going to be just that ethical, responsible, and transparent, Right?

Well, I think that's just, it's, it's the type of leadership that, that we desperately need in the industry. Um, right, and no, and, and so it, it, it works on two levels. One, it it works on a level of, of, you know, Qualcomm being a responsible technology company, not only providing, uh, the generative AI capabilities on their devices, right?

Which we've talked about many times on, on this podcast, um, where it essentially the, the, the whole on device AI capabilities that we're seeing, so the, the transition from generative AI just being on the cloud where you, you're on your device and you're, you're doing your work through an app that actually works in the cloud and, you know, sends the information back. Um, a lot of, a lot of what Qualcomm's been working on is having generative AI work directly on the device. It doesn't leave the device, it doesn't go to the cloud.

It's all there for security reasons, for speed reasons, there's almost no lag. Uh, so we're gonna see, uh, the, the arrival of devices, not just mobile phones, but also PCs, uh, and tablets that, that, um, can run a lot of these generat AI apps on, on the device without any connection to the internet. Um, so it, it's important for, for Qualcomm to take a stance there and to create these certificates, these, these modalities, these processes that, um, that capture, first of all, that, um, what type of manipulation has been done to an image or to a video, and also, uh, translates it to users to be able to say, okay, like this, whether you're the New York Times or anybody else, uh, being able to look at an image or a video and say, see exactly how it's been manipulated, if it's just been, you know, edited for a contrast and lighting, it's just a little bit of a color adjustment, or if it's actually been manipulated to change its intent or its content in a way that, uh, that is substantial as opposed to just aesthetic.

Um, the other way that it works, I think, is establishing Qualcomm as a leader in the generative AI space. So it's, it goes beyond the ethics. It also goes to brand positioning, um, and, and company positioning with regards to perhaps investors, but also, uh, industry partners in the public at large.

Um, which is something that, that I think is clever. It's, uh, it's something that more companies should do. Uh, but we don't necessarily think of Qualcomm as a leader in AI and generative AI capabilities, or at least, um, uh, enablement.

We just think of them as the chip maker. We think of them as an IP company. We think of them as, you know, a component maker.

We don't always think of them as the, uh, the enablements company for generative AI devices. And this is sort of a, a reminder with every single file that's out there with a policy like this, um, and with its application every single day, day, uh, for the next few years, uh, is, is a constant reminder that, that, um, that Qualcomm is, is that company, and that it plays in the space and they play such an important role in it. And, but the last thing I would, I would say is just kinda like the caveat and all this, um, yes, if you're transferring the image, the video, whatever the file is from one device to another, the certificate follows it, it, we still need a solution for screenshots, because it's still very easy to just take a screenshot of an image, um, that will not have the certificate, that will not have all this metadata following it around and still post it on, uh, a social network or, or somewhere else.

Um, and so it's still not a completely foolproof model. We still have to find other ways of verifying images and verifying their authenticity, uh, or at least that their context, if they're not fully authentic, but they don't necessarily detract from the story that they're retelling. Um, but that's, that's something that's beyond Qualcomm or any other company's, uh, ability or purview.

I think that's, that, that might be more, that might fall more on the, uh, on publications, on social networks and social platforms, um, and perhaps just the law in general. Um, oh, one last thing. There's a difference between art and, um, and news and news commentary.

And I think, you know, it's, it's important not to censor images that are manipulated. Um, but it's important to create the, the right context. If I wanna manipulate an image for my own artistic expression, um, and use it for that purpose, I shouldn't be allowed to, there shouldn't be a prohibition against that.

But the moment that that image is used to lie to people, to, um, to commit some kind of, you know, fraud, intellectual, political, financial or in, in any other way, then that's when we get into some, uh, some issues where, um, the public at least needs to be notified that, Hey, this, you know, exclamation mark, the, the, the image that you're looking at has been manipulated, and here's how. No, those are all excellent points. And yes, it's definitely an ecosystem dimension here.

Yeah. You know, it's not just Qualcomm. And that's why I think, uh, dur is a blog is so valuable because it's pointing out like, yeah, we're not going to necessarily account for every single instant, but it can be very important for peace of mind when it comes to, for example, a national security application or, you know, a mission critical or a public safety application because, you know, there are bad players out there.

And so, yes, uh, you're, we're gonna have parodies, we're gonna have manipulation, you know, if it's one thing if somebody's doing it from their college dorm room, you know, and, you know, uh, and it's, uh, on a, uh, a topic that is, you know, comet. But, uh, when it comes to these other contexts, when it comes to, you know, these use cases, uh, this is something that, uh, can quite simply provide that, uh, difference. And again, you know, provide what can be characterized as peace of mind, because there, there is that risk that, you know, if AI is falling into or being manipulated, uh, to such a degree, that it can end up, you know, producing hindenberg like, uh, scenarios.

And obviously that needs to be avoided. You know, we don't want this, you know, to end up like the denberg a dramatic crash of, you know, uh, say, uh, an entire mobile network, uh, because AI was, uh, being used the wrong way in the wrong hands. And here's something that can help transparency, that can help for very specific instances where it will prove a vital.

And so that was looking at images and, uh, you know, the, uh, that aspect of it. But there's also another important aspect here that's being impacted, and that is how transparent AI can actually play a role in the application of audio product development. So now we're talking about the audio dimension here, and what, uh, caught our attention was that Nokia recently, uh, presented its approach to this and how they're building their products, uh, basically using, uh, transparent ai.

So products such as their, uh, OZO audio and immersive voice capabilities are assuring that mobile devices come with the sounds of immersion, clarity and focus, again, cutting down the potential confusion and manipulation. Now, uh, this of course, requires extensive exploration and also analysis and testing of every step of the development process. And so, as such, Nokia is using machine learning models, uh, to analyze the output and tweaks of the algorithm algorithms so that the Nokia team can get the desired results that it's aiming for.

Also, Nokia is using, you know, software tools, uh, practices and ML ops, which I think has been a bit unsung, but it's certainly going to be integral to how AI itself evolves. But, but it's also, uh, understood as AI slash ml when it comes to product development operations, so many other key aspects that aren't necessarily Gen AI only. Now, with that in mind, what is going to, uh, come out of that is this, there's more trustworthiness that can come from the data that's used for developing not only their products, but also how, you know, audio capabilities are integrated and optimized.

And so as a result, uh, Nokia is basically using a configuration management tool, Hydra, uh, that I think will gain more recognition here, uh, over the course of this year and beyond. But, uh, what it's, uh, actually enabling is, uh, the use of pie torch lighting, which is a framework that helps to speed up development work and basically unify the tools that are being used. And as such, it can, uh, ensure that Nokia and whoever else uses these capabilities can reproduce its results every time.

So we definitely need that consistency and reliability built in. And to, uh, basically tie off of this, um, introduction by me, there's a new acronym I wanna share with folks out there. It's lumm.

Now, I know many folks have heard of it, but not necessarily ev everybody, and it's simple Linux Utility for resource Management. Now, why is s LM important here? Well, it's a tool that NoQ is using to make sure that what is being shown is that the user is running a specific job, which you touched on Olivier and with the resources that are required.

And as a result, Nokia is, uh, uh, having s lm, uh, work with its container system, uh, tainer, uh, to allow the transfer of its code in a transparent way between allocated resources, uh, across the datas data centers. Now, this way, Nokia can transfer and manage its jobs without the challenges that have typically arisen when different versions of the same software are used. So we know there's gonna be a lot of software updating going on here, and thus, version alignment is not going to be possible every single time.

But that shouldn't matter. It doesn't matter which version's being used. You have to be able to tap into something like S SLM to assure that you, to get these, you know, better outputs, uh, these improved business outcomes and so forth.

And with that, uh, Libya, what are you seeing in terms of, you know, not only Nokia's, um, initiative here, but what's going on in terms of enabling, you know, audio optimization, smart audio, and so forth? Yeah, so for, for starters, s SLM is just a wonderful acronym. I love it so much.

Um, so beyond that, no, it's, I mean, it's brilliant. Um, beyond that, no, it's the, I think the, the, the connection between ai, uh, whether it's generative AI or just kind of, you know, assistance. Um, and so the application of, of, uh, you know, more advanced AI and generative AI in our daily lives and audio, uh, and I is, I think still understated and underappreciated.

Um, I feel that, um, as, as we move forward to a world that's much more AI forward, that where AI becomes much more ubiquitous, we can be basically in any room and just talk and, you know, prompt an AI assistance to help us with a query, with a search, with a process, with whatever it is, um, voice is still the most natural interface. It's not so much a keyboard, it's not tapping a screen. Um, it's not, you know, gestures and, and pinching and all these things help, and, and they're gonna continue to be part of that ecosystem.

But I think that the natural voice interface back and forth between an AI who can respond to us in our natural language, uh, and that understands our natural language, um, is, is critical moving forward. And we're starting to see it with, uh, smart glasses. Mm-Hmm, which are not technically, I mean, they're still an XR product, they're not augmented reality, they're not mixed reality, they're just smart glasses.

But, uh, the smart glasses that were put out by meta, um, in collaboration with, uh, RayBan last year, which by the way uses, uh, Qualcomm's own XR platform is a really good example of what's to come, which is basically telling the camera, Hey, take a picture, start filming, doing things like that, make a call, call Ron. Uh, and then automatically we'll do that. And then the relationship between the user and the, uh, the, the device becomes even more important with, with regards to voice.

Um, because now you have to have smart speakers, uh, that, that if you're wearing glasses, direct the sound into your ears and optimizes the sound to be able to sort of cancel out noises outside of, of your listening pleasure. Um, but focus the, the sound directly to your ears. So we're seeing that technology move that way.

We also need noise cancellation to enable microphones, whether they're microphones around a room or microphones in a device, whether it's handheld or, or worn on your face, to be able to, uh, cancel out noise, whether it's a crowd, um, or, uh, you know, wind if you're in a car, for instance, or any kind of ambient sound to focus on the voice and be able to understand it better to process that language in order to be able to query the ai. And so that that relationship between, um, voice interfaces, voice capabilities, or sounds and, um, and AI is, is really, really critical. And what's interesting on the device level, especially on the software and silicon level, um, is that it's, it's a feed, it's a, it's a performance feedback loop where AI needs voice and sound in order to be more, uh, to create more utility for users.

But also AI can help fine tune that, that audio, uh, and sound capability by, um, using noise cancellation, by using voice enhancement and sound enhancement, choosing the right channels, um, and doing all of this, this very complex, intelligent, fine tuning to get the best, um, voice capture and the best sound feedback. So, um, so it's, it's, it's pretty great. I think that, um, up until now, I've noticed that sound has been sort of the last thing on a lot of tech companies platform, um, uh, introductions.

And I've, I've seen this with Qualcomm, I've seen it with others where it's kind of like, oh, yeah, by the way, let's talk about sound as well because we have all this cool stuff, right? But, um, I suspect that that connection to AI is going to bring sound quality, sound intelligence, um, and the type of, of, uh, the type of project and, and endeavor that you were talking about with Nokia, uh, and bring that to the forefront of those experiences because it's, um, it's, it's not just an afterthought anymore. It's not just a, oh, and we have sound as well.

It's, it's critical, uh, to the future of AI or AI interfaces rather. Yeah, I, I think those are just spot on Olivier in terms of, you know, why audio is going to play such a major role in how, you know, gen ai, uh, as well as, you know, all the, uh, related applications can be successful because natural language processing, yes, uh, you know, uh, you can type it and so forth. But there are definitely gonna be contexts and scenarios where, uh, the audio aspect is going to be vital.

That is, you know, talking into a prompt, being able to, you know, have prompts that are clearly understood. And also the explainability, if it has to be done on an audio basis, is, uh, just doing that. It's, uh, allowing the user or the groups of people involved to, you know, know what's going on.

And we already touched on, you know, uh, uh, context, like, you know, mission critical, um, applications, but also public safety. And there's just so many ways, uh, why this is, uh, actually going to be just a, a major piece of the puzzle. And, uh, we anticipate that will become increasingly, uh, emphasized in terms of portfolio development.

So it's not like this, oh, by the way, aspect. And so that's why, yeah, that Nokia, uh, blog definitely leaps out at us. And so at that, uh, my, uh, concluding thought is long live s slm, but also thank you, Olivier, for joining us today.

I hope you have a great, uh, weekend and, uh, week ahead of you. I hope so too. Thank you.

The same to you. And, and same to everyone watching Next, Next weekend, if you're, yeah, if you're catching us next week, there's still a weekend coming, so you're, you're good Hanging there. Thumbs up and yeah.

So, uh, uh, to our, uh, viewing audience and listening audience, yeah, please, uh, keep in mind, uh, the 5G Factor can be bookmarked also, you know, reserve, uh, that, uh, ability to, you know, get the heads up when our next, uh, recording and our next, uh, you know, interaction is being, you know, broadcast out there. And so, uh, with that, everybody have a wonderful 5G and gin AI day. Thank you.