This blog post was originally published at Qualcomm’s website. It is reprinted here with the permission of Qualcomm.
An interview with Qualcomm’s CMO on the future of AI, generative AI and on-device AI
Just as Qualcomm was critical to the wireless and mobile revolution, so too is it part of the artificial intelligence (AI) revolution. Don McGuire, SVP and CMO at Qualcomm, was recently interviewed by TIRIAS Principal Analyst Jim McGregor regarding the future of AI and generative AI, and what this means for Qualcomm, the industry and society. The following is a transcript of the interview.
Editor’s Note: The following interview was transcribed and edited for readability.
Jim McGregor (JM): Don, the world is excited about generative AI, and we’re just starting to learn what it can do and what it’s going to offer the world. Can you give us Qualcomm’s vision — or your vision — of what generative AI looks like in 5, 10, even 15 years?
Don McGuire (DM): Obviously, we’re in a hype cycle. But it’s a hype cycle that is actually quite sustainable.
AI has been running in the background doing things for us across our technology portfolios for many years.
We’ve invested in research in AI for 10 to 15 years now, and it’s been on our platforms for quite some time. It just hasn’t been recognizable or up in front of people’s faces, as it is with the advent of generative AI, large language models and the advent of things like ChatGPT.
Because of this, the novelty has come to AI and quite frankly, the use cases. So, I think that generative AI is here to stay. Obviously, there’s standards, there’s responsibilities, there’s things that have to come with any new technology, but I’m excited about the future, and what generative AI can actually bring across a myriad of different vectors.
JM: You mentioned all the different types of applications that are using generative AI today. Battery management, network connectivity, computer vision — all these are already running on a smartphone, right?
DM: Absolutely! From your camera experience taking pictures to improving audio, anything with training and algorithms can create better experiences and allow people to be more productive, more creative or move through their lives better — and in a more efficient way. People have to opt into that help, but once they do, the sky’s the limit.
JM: Now we’re doing text-to-text, text-to-speech and text-to-image but we’re not even touching video, gaming or metaverses. Yet, generative AI — like ChatGPT — is already putting a tremendous strain on cloud resources. How are we going to be able to scale to video, to gaming and to metaverses?
DM: I think in order for generative AI to scale, there has to be a hybrid approach. There needs to be a certain amount of AI work and workloads that happen on device and at the edge, all the way up through the cloud.
And if we can think about that for a second, that’s really the only way it’s going to scale because everything can’t go to the cloud and come back down. There are latency issues. There are power consumption issues. And if everything has to go to the datacenter, there’s simply not enough space, enough power, enough water or resources to be able to scale generative AI.
You’re already seeing examples of this when people go to ChatGPT today and ChatGPT says “Come back later, can’t give you an answer right now.” So, this hybrid approach to AI allows you to split the workloads and spread the workload out.
There’s so much that can be done on device. That’s where — at Qualcomm — we have our core DNA. We’re already running billions of parameters on device, where that work can start and doesn’t have to go to the cloud. At some point, the entire workload can be done on device.
So, there’s going to be this spectrum, but it’s going to bridge out to this hybrid approach — to AI in the cloud, AI at the edge and on device — in order to truly scale generative AI.
JM: Is this also sending a sustainability message?
DM: Absolutely! Servers have to use more water, and water is a precious resource. Power is a precious resource.
So, there is a strong sustainability message that goes along with the whole hybrid AI approach, purely from a resource consumption perspective. The hype cycle around AI as a cloud function only will be short lived. We’re already seeing evidence of datacenter operators’ and hyperscalers’ power consumption going through the roof, and water consumption rising by 30–50% per month at their data center sites.
That is something everyone should be concerned with and really where on-device AI — and moving some of these workloads onto our powerful platforms that are high-performance, low-power platforms — is going to help bring down that resource load, and those requirements of additional power and stress on precious resources.
JM: Our (TIRIAS Research) study looked at it and said by 2028, it was going to be a cost — in terms of operational costs — around $76 billion. That’s not including the brick and mortar, just the operation and the equipment. And that’s not sustainable because that factors in the power and water and everything that you mentioned. So, as we look at AI/generative AI, we’ve been using traditional AI for a lot of specific functions. But generative AI breaks that mold and starts giving us the ability to have something that’s cognitive to a certain extent, or at least contextual, with the information that’s around it.
So how do you think this is going to change businesses? How’s this going to change Qualcomm?
DM: I can give you some examples of how we’re already incorporating AI into the marketing function. It is helping to make some of the initiatives that we would embark upon — that took a lot longer and were more manual — faster.
With the advent of generative AI, we’re developing copy a lot faster, we are doing product naming a lot faster, and we’re starting to utilize and experiment with these tools to raise the productivity level, time-to-market, efficiency and speed.
It’s really not about replacing people; it’s about replacing tasks. And I know there is a lot of worry and concern about AI causing huge job losses. I think with every new technology, there is always that concern.
When new technologies come to market, there is a lot of fear. It’s been that way since the advent of the elevator or the wheel. “Oh my gosh, there’s a wheel, why do I have a horse? I don’t want to get off my horse and get in something with wheels!” So, you can track back through time, new technology fear, but once you get over the fear cycle, it starts to be practical, and people start deploying it.
So, look at it practically as replacing tasks, making people more efficient, but also the creation of new skills and new jobs.
You and I were talking earlier about the advent of computer-generated imagery (CGI) in the movie business and how that replaced the need to build manual sets. Those set designers and set builders have had to morph into graphic designers, or graphic artists or computer graphics specialists, but it’s a cycle.
And with every new technology, new skills and new jobs are created. And there are new ways for people to learn and to apply those skills. So, I’m not too concerned about that over the long term with regards to job loss versus task improvement. That’s been my experience so far and, I think, a point of view that is more pragmatic.
JM: Companies like Qualcomm have invested in core technologies that are going to be critical to AI. Qualcomm has invested heavily in connectivity, low-power processing, as well as some of its own AI research. You’ve also invested in a lot of different platforms that are probably going to benefit from AI: XR, the handset, the PCs, etc. So how do you see that morphing devices as we go forward?
DM: There is this whole device ecosystem storyline as a part of rolling out and scaling AI in general. But with generative AI specifically, we’re already showing experiences running, for example, on connected cars or the new software defined vehicle, and how you can utilize generative AI while you’re driving or even sitting in your car waiting for something to happen, such as picking up your kid from school. You can actually be having a conversation with a car and be accomplishing tasks.
So, from the vehicle all the way through to the PC, back to the smartphone, where the ChatGPT craze is rooted, we see generative AI working across devices. As you move through your life, your digital life and your physical life, and you’re engaging with different types of technology tools, you’ll be able to thread that AI horizontally through how you interact with those tools going forward. And it should be helpful in making things easier, faster, more efficient, more productive, more creative, or whatever adjective that you want to throw behind it. So, we’re really excited about this; the horizontal scalability of the technology as it relates to our platforms.
I know you’re going to be talking to some of our tech experts as part of the series (of interviews), but there is AI throughout the tech world right now. And within the ecosystem, in the industry, there’s this debate brewing about how AI can be deployed and can utilize CPU versus GPU versus NPU. We really believe in a combination, but where we really shine, where we really accelerate and have a differentiated mindset towards how this gets built out in a holistic manner is really that NPU, which we’ve brought to the table over generations of our Snapdragon platform. So, we’re excited about what that will give us from a competitive advantage in bringing these new experiences and new use cases to the world.
JM: This is all great. I really think that generative AI is going to change how we learn, how we work, how we play — it’s going to change so many things about our life. What excites you the most?
DM: From my vocation, my role in the organization as the CMO, I’m really excited about the tool sets that we’re going to be able to use to create and to do marketing better, such as the soft science of marketing.
Personally, I’m also excited about the fact that you’ll be able to fill in the downtime or the gaps by using these tools to actually move through your day in a more efficient manner. We mentioned the car earlier but imagine being able while you’re driving, or even waiting to pick somebody up from somewhere, and you can accomplish three or four things off your daily list of things to do simply by having a conversation with the car — using generative AI and large language models — to then go and complete those tasks. I think that’s pretty exciting as well.
And there is this responsibility and sustainability that all has to be taken into consideration. Because technology, after all, is just technology until you put it to use. So, everyone has to have a role to play in the responsible rollout, in this responsible scalability of AI.
But I’m also really excited about the fact that our platform — specifically the Snapdragon platform — plays across these device categories. The more we evolve our platforms, the more complicated workloads we’re able to handle on device and at the edge, which is going to play a more meaningful role in the AI landscape.
JM: I have to say, just playing with the Stable Diffusion demo at Mobile World Congress was like playing a game. It really was exciting. So, I look forward to seeing all those applications come down from the cloud to these devices. And obviously, it’s coming there very quickly.
DM: Yes, and we’re going to move past novelty to practicality. And then I think adoption curves will start to go up just like with all new technologies, so I’m excited to see where things go.
JM: You mentioned we’re in the hype cycle right now. So, how long before we go from that to the practicality stage?
DM: I think it’s moving at an accelerated pace.
I’ve been in tech for all of my career and been a tech marketer for most of my career, and I have to say, I’ve never seen something move at such an accelerated pace as I’m seeing with AI, specifically generative AI.
Most adoption curves are a lot slower. And even the evolution of the technology and how fast it’s moving, which I think is also scaring people as well. If you feel like it’s happening so fast, you feel like you’re out of control. So, it’s impressive. It’s a little scary. But it is fascinating to see how fast it’s accelerating.
Pat Lawlor
Director, Technical Marketing, Qualcomm Technologies, Inc.
Jerry Chang
Senior Manager, Marketing, Qualcomm Technologies