Rishi Jaluria
“Companies that don't truly embrace generative AI will see their multiple compressed by 50% over the next five years. Every company needs to have a generative AI strategy, or they will get left behind…”
Mark Odendahl
Welcome to Industries in Motion podcast from RBC Capital Markets. This is where we explore what's new and what's next in today's fast moving markets and industries. Please listen to the end of this podcast for important disclaimers.
Mark Odendahl
My name is Mark Odendahl and I'm Head of US Equity Research here at RBC Capital Markets. I'm joined today with two important and senior members of our tech research team. Rishi Jaluria who covers software, primarily application software. And Brad Erickson, who is our internet analyst. Rishi, Brad and the entire tech team work together to recently published a report software and internet implications of generative AI and Chat GPT. This topic is very fresh in the marketplace and we're very proud and excited to have two of our tech leaders on this podcast to help us understand generative AI and the implications from it. Guys, welcome to the podcast
Rishi Jaluria
Thanks for having us.
Brad Erickson
Thanks for having us, Odie.
Mark Odendahl
Rishi, in the beginning of the report, you talk about generative AI is kind of the fourth capacity unlock for technology over the last couple of decades. You know, internet, cloud mobility. Now, AI, could you talk us through that?
Rishi Jaluria
Yeah, absolutely. You know, I think that framework has been really helpful for us to just state how important the rise of generative AI and especially you know ChatGPT is kind of a watershed moment is, as we think about all those technological unlocks, there was a big watershed moment. With the Internet it was Netscape. With the cloud, it was both Salesforce and AWS. With mobility, it was the release of Apple's iPhone. And now with generate with AI, it is ChatGPT. Right, which is the fastest growing internet service in history, it took two months to hit 100 million users, which I think is absolutely incredible. And the reason we think this is so important is because AI prior to this was either really powerful but required data scientists and a lot of customization to use, or it was very simple consumer grade interfaces, like Apple Siri or Amazon's Alexa. And now with ChatGPT, you're combining the best of both worlds. And that's why we think this is so important. And honestly, this has huge implications, not just for software and internet, as the report talks about, but really for society as a whole.
Mark Odendahl
Brad, do you want to expand on that?
Brad Erickson
Back to your the comment on the fourth sort of gen of the capacity unlock, you know, each of those three first things just took human beings and the tech that we interface with and dramatically expanded them across new vectors essentially. And so not to call this the, the fourth dimension, so to speak, but it could literally be that powerful where you're going to give companies and people and organizations this extraordinary ability to find efficacy and all sorts of the workflows and things that they have to do. And that's everything from, you know, simple content creation to obviously software development, auditing giant, giant datasets, in different ways and exceeding any performance, frankly, that a human could do along those lines previously, so it has transformative power, we think going forward
Mark Odendahl
Brad, in the report, I think that you discussed that there'd be two primary outputs for Generative AI. And you had a fun title with that moats and boats. Can you give us some detail on that?
Brad Erickson
Yeah, sure. This is a common recurring theme across technology innovation for the last 30 years. But anytime you have new, disruptive tech, you're always going to have some level of usually concentrated incumbent players who have the resources to invest and theoretically can have a play on the new tech. And then of course, you also have, you know, you'll have venture capitalists and a rash of startups, also approaching the space very quickly. And I think you're gonna you're you've already seen that. Certainly in the case of large language models, where, you know, Google has been Google wrote the paper on the, on the original Transformers, which is the T and GPT.
And then Amazon's been working on this for cloud applications longer than anybody. And then, of course, Microsoft hooked up with open AI, and they brought chat GPT to market. So with those LLMs, obviously, those are the large language models. Those are the incumbents I'm speaking of, right? Those are the moats. And then the boats are the 20 plus other startups who are also creating their own large language models and raising capital to train these because it's hundreds of millions of dollars and 10s of 1000s of GPUs, to do so.
And so the question is, is, you know, are those incumbents going to be able to maintain the moat, maybe even expand it right, as they lock people into using their LLM, potentially, with their cloud service? Or are some of the startup models going to commoditize these things and then actually, you know, use disruptive pricing to drive adoption from again, companies that can leverage it, whether it's software companies for internal efficiencies, or internet companies that are you know, created using their own one PII data to create outside or differentiated user experiences? So I think that how that plays out is very unclear. But I think certainly you're seeing the incumbents apply very, very large dollars to this because they represent how important and existential this is for the next you know, five ten fifteen years.
Mark Odendahl
Great. Let's, let's start also digging into the implications for the tech sector. Brad, how do you think generative AI or ChatGPT is going to impact search going forward? Or the internet going forward?
Brad Erickson
Certainly for search, what it introduces is this much more interactive capability, right? Historically, you go into search with a with a pretty high level of intent of what you're looking for.
In the case of the interaction with a ChatGPT, for example, or a Bard in, in Google's case, you know, you can, you can kind of leverage this amazingly powerful tool that is effectively accessing the entire internet, to reason through things, and or create feedback that you otherwise would have had to put together yourself and could take hours and hours, days and days, weeks and months, potentially.
In our work for this report, what we were coming to the conclusion on was that, you know, for a lot of the different commercial searches that people leverage Google search, for example, today. In many cases, generative AI may not be a full replacement, it may supplement in some cases, but it probably provides very little if any disruption to the monetization model that search operates on. On the other hand, there's a vast majority of Google searches are actually for informational purposes only meaning they're not monetized. 80% is roughly the number. And that's where things could really change. And I think the big question, of course, everyone's looking at is whether, you know, Bing is going to be able to leverage having chat GPT and a perception of an early lead to be able to drive share shift. And we'll see what happens there.
In the case of internet companies, what it's going to come down to is you are theoretically going to be able to leverage your differentiated first party data layered on top of a large language model to produce a superior and frankly, unique user experience. And so, you know, the companies that obviously possess the majority of that data, or the best data out there should be the best position ones to benefit from this.
Mark Odendahl
Rishi, implications for the software industry, both positively and potentially negatively...
Rishi Jaluria
Yeah, absolutely. And much like Brad, maybe I'll bifurcate this into two separate questions. The first, which is maybe a little easier, is what does this mean for the cloud. And we think this is going to be really additive to the cloud. You know, really be it because you have net new workloads being built on platforms like open AI. And you know, we've made the kind of bold prediction that just like you had a trillion dollar mobile app economy built with the iPhone, and later Android, you had a trillion dollar cloud economy, once AWS launched. We think the same thing is going to happen with generative AI. And this is coming at a very important point when investors are really worried about cloud saturation.
Now, if we think about software a little bit more broadly, I think there's two ways in which software companies can really benefit. One is on the back end, which is you use generative AI to make your software development more efficient and come out with new features and iterate faster. You know, marketers can be significantly more effective in what they do. So you get way more operational efficiency, you can come to market faster, you can create software that's more custom tailored for the customer and their use cases in their own internal data.
But also there's the productizing of it. And we're starting to see this already where software companies are integrating generative AI and ChatGPT into their solutions. Uh, you know There's ways to, you know, really make the data that these companies and these market leaders have significantly more valuable. We also think, you know, this is a huge shift in the way we interact with software.
You know, we're all of us on this podcast are old enough to remember when we move from the traditional command line interface of Microsoft DOS to the GUI or graphical user interface that we saw with Microsoft Windows. And I think with generative AI that may not happen right away but shift from the point and click interface that we that has dominated enterprise software for the past 30 years into a chat first functionality of the way we interact with software, we think that's going to be a huge step function change. So we think this is generally going to benefit software across the board, especially market leaders.
Mark Odendahl
Now I'm gonna move on to a fun topic. So we have Rishi, our Microsoft analyst, Brad, or Google analyst. So I'll ask you both. We'll start with Rishi. Microsoft got the head start in the space by integrating ChatGPT into its Bing search. Though Google obviously has since launched its own initiative with Bard. So how do you envision this playing out?
Rishi Jaluria
I think Microsoft has many ways in which they benefit from this, right? I mean, search is what's occupying a lot of the headlines. And I think it's really impressive that Microsoft's been able to embrace ChatGPT and integrate that technology into the new Bing, which, which we've used actually, throughout the report. And I think very impressive. But remember, their investment in open AI, they own less than half the business, it's kind of an arm's length thing. So open AI can iterate and innovate on their own, without it necessarily being a direct reputational impact to Microsoft. But I think beyond search, and even cloud. But I would say I think it's really impressive how Microsoft is able to integrate ChatGPT and generative AI throughout the product suite. Right, you know, already, there's GitHub copilot, which can help developers iterate on code faster, and test code faster. So they can come out with new products, and you can get developers to become significantly more efficient. We're already seeing integration with teams, where if you join a video meeting late, you have real time transcription, that'll catch you up on the key bullet points of the meeting.
Throughout the Office365 suite, you have a lot of use cases where it can help you started document without having, you know, help you avoid writer's block, right? What is kind of the format we need? If I have a resume? How do I make this sharper and crisper? If I have bullet points that I want to turn into a PowerPoint presentation? ChatGPT with that office 365 integration can do that. So I think there's benefits throughout the suite of Microsoft products. It's not just search, it's not just cloud. It's really everything under the sun for Microsoft.
Mark Odendahl
Your thoughts Brad as it relates to Google?
Brad Erickson
For Google, I think there's several vectors here, which I think can be kind of a longer term benefit. Number one, they obviously have created their own LLMs. And I think it's important to remember that they're the only player in the space that's doing this on a vertically integrated basis, top to bottom, which is to save all the way from the silicon to the models to the application outputs. And I think, you know, presumably, Google has had a roadmap for generative AI probably in place for as much as five years or more.
The difference, though, is that, you know, there never was this challenge, or like what ChatGPT has become, in order to push them any faster. And there is the well-known risks with all of generative AI for companies is compute costs. And so the risk, of course, with Google is that they have to significantly ramp up CAPEX, as they accelerate more into this space to keep up with at least the perception of where ChatGPT is within Bing search, for example.
I think the day before Microsoft came out with the copilot announcement, Google quietly slipped a blog post in there about how they were rolling out a model that was about 10x more powerful than Bard for their workspace solutions effectually effectively the exact same thing as copilot.
So I do think you're going to continue to see them, roll out these solutions and use them effectively, the single biggest existential question, ultimately for Google is, are people going to be compelled to use Bing more because of ChatGPT and ultimately change their search habits given that Google has over 90% of the usage on search on a global basis excluding China and Russia… That question remains very much an open question. To date, nothing's changed. But we'll kind of see.
I think, you know, on the cloud side, and I think Microsoft has the exact same tailwind, frankly, that Rishi spoke about earlier is that Google has one of the largest developer networks in the world. And so naturally, you have an API out there now, where developers can build applications on top of a model that's called Palm, which is over 600 billion parameters versus Bard, which is under 100 billion. Developers are creating applications on that right now.
If you look at the time from when chat GPT was launched, to the time that major companies are bringing it out with, you know, end use applications. That was about three months. Well, this was announced about a month ago. So we'll, we'll see what happens there. But I do think that, you know, again, as developers are compelled to build on top of that applicant, using that Application Builder, you're going to tie in a lot of incremental cloud activity associated with that to Google Cloud.
Mark Odendahl
So Rishi, could you give us an example of, you know, maybe the most enlightening use case that you've witnessed as it relates to AI?
Rishi Jaluria
There's a lot of great like headline grabbing, use cases out there. You've seen it can score the 99th percentile on the LSAT. It's gotten into law school, it's gotten to business school. You've seen there's a drug that's clinic currently under clinical trials that was developed entirely by generative AI. I think those are all impressive, but those are kind of out of grasp the one that I think, you know, personally, I'm using that I think is really interesting. That's also a lot of fun for me, is I'm currently working with ChatGPT for to write me a movie script. It's written me the budget and the business proposition, give me the full cast and why they'd be a great fit. Who should be in charge of the movie? So I think this is incredibly impressive that it can really come up with these sorts of brand new content, just like that with simple prompting.
Mark Odendahl
And Rishi, you guys spend a lot of time talking about the potential societal impact of, of AI as well as any risks? Could you give us some thoughts there?
Rishi Jaluria
Prior industrial revolutions have primarily automated away blue collar jobs and manual labor. This is the first time that on a large scale, you have an industrial revolution that can replace white collar workers and knowledge workers.
I think we need to think ahead and say how do we think about reskilling and upskilling? Those workers that might be disrupted from this and you know, that could be a pretty big tailwind for the Ed Tech sector.
Now, if we think about potential risks, as we think about the societal implications, as well as data, from our perspective, it feels like governments are going to get increasingly involved around regulation and scrutinizing how these systems can be used.
Already, we saw Italy banned ChatGPT, really around concerns around a data privacy, right? And that starts to get to the question of who owns the data, what is really able to be used, what should be more walled off and proprietary. And no one knows the answer to how this is going to play out. I think it's a very delicate balancing act that we want to see this technology evolve that we think that the net benefit to society can outweigh the detriments from job losses and job reshuffling. But it's just on you know, governments really have to figure out how to tow that line properly and regulated without killing innovation.
Mark Odendahl
Brad, any view from the risks side of things on your part?
Brad Erickson 24:36
So maybe just from a I mean, business risk, you know, there are companies in internet that that are a part of the content creation. In fact, that's a key source of their value. And I think you can argue that, you know, the capabilities at least conceptually at this point of generative AI could theoretically replace signal efficacy portions of those businesses. So there are companies that could absolutely, you know, they could just see their sheer business model blown up by this, frankly.
I think this partially speaks to Google in particular tentativeness around rolling this out more quickly. I think in Microsoft's case, it makes a ton of sense to really sort of throw spaghetti at the wall and see what sticks, which is why they're being so aggressive in both rolling it out as well as marketing it.
ChatGPT doesn't yet silo their data from an enterprise perspective, they're actually leveraging all the inputs into the core language model. Bart and Palm already silo that because Google has, I think they're anticipating what the obvious pushback would be from enterprises leveraging, ChatGPT’s LLM.
So you know, it's, it's those sorts of things that make the technology to some degree feel quite underbaked from a broader commercial use case and regulatory standpoint.
It's pretty scary to me, that the creators, or people that have invested in the creation of this technology are already fully acknowledging that A, the innovation curve has probably gone far beyond what they expected and B, they don't know where the end exists and they say that with a very ominous tone, meaning we don't know what harm this could drive in the world.
I don't think there's a technology that's ever sort of done that.
We joke you, most maybe some listening, have seen the joke around, people looking to hire the kill switch engineer. It's a joke today. It might not be in five or 10 years.
Mark Odendahl
And, you know, why don't we conclude where I'll try to pin you guys down on one prediction, either medium or longer term as it relates to the growth of generative AI? Brad, why don't I start with you?
Brad Erickson
For me, I would say it lies in the advertising vertical. And social media in particular. There's a huge industry, particularly as short form video, mediums like tick tock and reels have sprung up. There's a there's a huge industry in and around the content creation and the creative that drives engagement, and hopefully advertisements and conversion of new customers and lead generation things of that nature, in that space.
Given that, that these companies like Meta, for example, and Google are already leveraging significant AI in and around recommendation engines and the ability to attribute conversions and so forth, I think what you're going to see is that, the AI is going to learn from what visually or audio wise or copy resonates with users that drives conversion, and the AI is just going to constantly iteratively learn and I think what's going to happen ultimately, is it's gonna significantly marginalize the agency creative model that has existed for 50 years, the Mad Men model so to speak. And you'll see that play out over time.
Mark Odendahl
Rishi, one prediction?
Rishi Jaluria
Companies that don't truly embrace generative AI, will see their multiple compressed by 50% over the next five years.
Every company needs to have a generative AI strategy, or they will get left behind and you know, constantly draw the analogy to the cloud, where you had companies that really embraced the cloud and saw their stocks take off. And on the other hand, you had companies that maybe paid lip service to the cloud, or were dismissive of it, that saw massive rewriting of their multiple and a lot of shareholder value destruction. And we think the same thing is going to happen here with generative AI, we're paying very close attention, not just to the marketing and the announcements, but really, to what extent are companies re architecting and changing their workflows changing the way they do things for this new world, in general of AI and becoming really AI first companies, because those companies that don't those companies that are dismissive and saying, “Hey, AI is a new thing. It's just like crypto back in the day.” Those companies we think will be far left behind. And we really want to pay attention to the reality behind the marketing.
Mark Odendahl
Well, guys, this has been a great conversation. And I'm expecting this just be the first of probably many generative AI podcasts that we'll be doing throughout the next couple of years. I just want to put in a special thanks to you, the rest of the team, your associate teams for writing this Imagine report on this very important topic and spending the time to dive into some of these important themes out of generative AI so thank you very much for your time today. And thank you for your time on the report.
Rishi Jaluria
Thanks, Mark
Brad Erickson
Thanks!
Mark Odendahl
What else lies ahead in today's ever evolving markets and industries? We'll be keeping track right here on Industries in Motion.
Until then, thank you for joining us on this episode recorded on April 21, 2023. Make sure you subscribe to Industries in Motion wherever you listen to your podcasts. If you'd like to continue this conversation, or you’re interested in more information, please contact your RBC representative directly or visit our website, www.rbccm.com/industriesinmotion for further insights.