Talking Tech With Mark Mahaney: AI’s New Frontier

Amanda Heckman: Hello, and welcome back to our monthly tech roundup, the video series that takes a deep dive into the tech sector with two of the brightest minds in the business. I'm Amanda Heckman, Editorial Director here at Manward, and I'm joined today by Andy Snyder, the founder of all things Manward. Hi, Andy.

Andy Snyder: Nice to see you, and I appreciate you calling me one of the, I don't know if it's true, but one of the best minds in the business. I know Mark is, but we'll see about me.

Amanda: Got to give you credit. And with us again is Wall Street tech veteran Mark Mahaney. Hi Mark, thanks for joining us.

Mark Mahaney: Hey, Amanda. Hey, Andy.

Amanda: Let me start by saying this is a real conversation being had by actual humans because today we're going to dive into artificial intelligence or AI. It's been around for a few decades, but it's really making waves right now. ChatGPT has set the internet abuzz with its ability to generate conversations, write research reports, lyrics, poetry, organized data, you name it, and now both Microsoft and Google are in a race to be top dog in AI search. They both shared demos of how their AI search functions could work, to mixed results. It's an imperfect system to say the least right now.

Mark, what did you think of the market's reaction this week to these developments?

Mark: The market seemed to have taken a binary approach to this. Microsoft stock has traded up very nicely since they demoed it.

I think they're now calling it Sydney. Their Bing search offering that is based on what's called LLMs, large language models, as if we need another acronym, but it's important, whereas Google did a very quick sort of hasty demo of Bard, and what was quickly reported on was the error that was in one of the Bard responses about the things that can be seen from the Hubble telescope. It turns out that the Bing demo also had several errors in it.

Amanda, your statement was that these are imperfect, and I think that's true. It's just very hard to do. If we step back just a little bit, this is an application of AI, large language models, and what they do is they work to predict the next word in a string of responses. When they do that, then they do the next word after that. That's how they then create the sentences after that and the paragraphs and the pages and the whole responses.

It's predictive responses, and it's very computationally complex. I think the workloads, the data point I picked up this last week, the workloads required to process large language models are doubling in every three and a half months. It's very computationally complex. This is the cutting edge, and it's one application of artificial intelligence.

Since you asked about the stock market reaction, I'll just note real quickly, Google's shares traded off 10, 11%. The wonderful headline, it's not wonderful. The dramatic headline was they lost a hundred billion in market cap. Stepping back and thinking about it a little bit, I thought that was just excessive. Google has not been one of my top picks for the year, but as it stays under pressure because of this, this is something where I just think it's a false trade-off.

One, this is going to be complex. There are going to be just as many errors and maybe probably more with the Microsoft offering than there are with the Google offering. I don't think anybody's really ahead of Google on this.

Secondly, I find it very hard to believe that anybody's going to be taking share away from Google in search because Google is such a strong brand associated with search. This isn't Pepsi/Coke, this is like, soda/water. I'm exaggerating by that, but I'm making a point, too. They have this dramatic advantage in terms of distribution. They are the default search engine on the iOS devices and on Android, and they have this Chrome browser.

The economics, they've got so many more advertisers than anybody else does, which means that search revenue is generated through an auction. Marketers bid against each other to figure out who wants to pay the most to be the top dog, the top listing, as long as they meet each certain quality constraints. That's kind of the black box that is Google search, but whoever has the most advertisers generates the most revenue. It's going to be very hard to see Microsoft economically outbid or outbid for any sort of distribution deal because they won't have the advertiser liquidity. I just don't think you're going to see any mar market share shift, and then the market's concern though is on the margins.

If something that's dramatically more computationally complex, does that mean that we're going to have a problem with Google's margins? Because Google search is like, a 90% incremental margin business. Andy, you and I is going to be a hard time thinking about any business that's got margins as high as search, although it requires a heck of a lot of capex to get started, but the incremental margins are 90%. The concern is, out in the market is that, well, I'm going to take my Google shares down or I'm going to short it because that great, wonderful margins they've had in the past are at risk, and I think there's two mistakes with that.

One is, I think the vast majority of search queries just don't require a large language model response. The ones that they make revenue off of, 49ers tickets, Valentine's Day flowers. You don't need a couple of paragraphs of responses, and what you really want are links to the most relevant commercial transactional sites, so that's one.

And then the other thing is, I just think that Google's invested ...

This is my son, Noah, heading off to school. The other one is that with large language models, Google has been aggressively investing in artificial intelligence since about 2018, probably earlier, but there was a major CapEx cycle associated with artificial intelligence, and you could see it when their CapEx ramped from 15 billion to 25 billion, and they've been running at that level. They're now at 30 billion a year, so it's not like all of a sudden large language models require this dramatic increase in CapEx by Google. They've been doing it for five or six years, so I think what we're going to look at, two years from now, we're going to look back. We're not going to see a dramatic change in Google's market share, nor a dramatic change in Google's unit economics, and this correction in Google's stock is one that we're going to look back and say, "That's what I probably should have aggressively bought." That's a longer answer than you wanted, Amanda, but I've been thinking about this a lot this week and hopefully that helps.

Andy: I like your approach on this. I guess my question, you're the internet analyst. You're the guy. When we look at that, that's a basically 23, 25 year spectrum looking back on the dawn of the internet to where we are today, and right, now we're kind of looking at AI in the lens the last two weeks or two months. Do you think the market's being fairly shortsighted here, trying to figure all this out in the span of February 2023 and what we're looking at is something that could be transformative for the next decade or more? Do you think investors would be wise to be a little patient here?

Mark Mahaney:

Mark: Your setup is right. I think that's likely to be the case. I also think the investments have been happening, we've seen AI improvements from the largest companies. There's two things about AI. It's very computationally complex, so it requires a lot of money to start with, and in order to be successful, you need access to a lot of data. What it probably does is it probably even steepens the moats around some of the largest tech companies out there. Google obviously, but Microsoft, too, maybe Meta, maybe Amazon, and one way to invest in this is, of course, is to play the infrastructure companies, the chip companies that really focus in AI, in NVIDIAs. It's not a formal recommendation, I don't cover NVIDIA, but that's oftentimes cited as a great way to play the AI wave.

Andy: I think there's a lot of excitement out there, but certainly, the long view and what we're looking at Manward is all the various uses of this. ChatGPT is certainly interesting, but there's so much more to that. I guess with that with, we'll hand it back to Amanda.

Amanda: Sure thing. I think this is a really exciting topic that we're going to come back to again and again, particularly in our publications, but also, in these conversations, so thank you both for getting together today and talking through it, and thank you to everyone who's watching, and we'll see you all next month.