[00:00:00] Marcus Johnson: Hey gang, it's Monday, June 1st. Nate, Jacob, and listeners, welcome to Behind the Numbers, a Emarketer podcast. I'm Marcus, and joining me for today's conversation we have two AI experts who are joining the show. Uh, principal analyst on the East Coast, Nate Elliott.

[00:00:20] Nate Elliott: Hello there, Marcus.

[00:00:21] Marcus Johnson: Hello there to you too.

[00:00:23] Marcus Johnson: Of America, people probably assumed. Analyst on the west, Jacob Bourne.

[00:00:28] Jacob Bourne: Happy to be here today, Marcus.

[00:00:30] Marcus Johnson: Hey, fella. Uh, today's fact

[00:00:36] Marcus Johnson: Rocky the Mountain Lion, didn't see that coming, uh, is the Denver Nuggets basketball team mascot, and Harry the Hawk, who's the mascot for the Atlanta Hawks, each earn almost exactly the same amount as a professional basketball player on a standard two-way player contract. That means that the player is splitting their time between an NBA team and the minor league G League affiliate.

[00:01:03] Marcus Johnson: And yeah, these mascots and players on two-way contracts each make about 600 grand a year.

[00:01:10] Nate Elliott: I have so many questions. Let's do it. Are, are the mascots overpaid or are the two-way players underpaid?

[00:01:16] Marcus Johnson: I think these mascots... Well, these are the two, uh, mascots who make the most. So most mascots make, uh, uh, on, on average about 60, 6-0, 60 grand a year.

[00:01:28] Marcus Johnson: Um, and so that compared to the lowest paid NBA players, they're on standard four-year contracts. That's a $1.3 million contract for an NBA player and a 60 grand contract, uh, for a mascot. These two mascots get paid an exorbitantly, uh, high amount of money compared to the average. Benny the Bull, Chicago Bulls, 400K, the Gorilla for the Suns, 200, and the Hornets mascot, 100.

[00:01:53] Marcus Johnson: So those ones- Wait, so- -are overindex ... so

[00:01:54] Nate Elliott: why do these mascots get paid- Right ... 10 times more?

[00:01:57] Marcus Johnson: Yeah, exactly. I went digging and what I can tell is just, especially this, um, uh, the mascot, um, the Mountain Lion, is just the, the things that they can do. Um- Hmm ... the acrobatics that they can perform, uh, the performances are just much more impressive.

[00:02:12] Marcus Johnson: So they can do this, they do a blind backward half court shot from the top of a 30-foot ladder, which is insane. If you Google it- Yeah ... YouTube it, whatever, it's insane how they're able to s- to, to make that shot. Yeah. Um, they can also do trampoline dunks, and they rappel from the rafters. So I think it's based on, yeah, how, uh, how impressive the performance is.

[00:02:33] Jacob Bourne: I guess it comes down to how much the team wants to invest in their mascot probably.

[00:02:38] Marcus Johnson: Yes. That too.

[00:02:39] Nate Elliott: So you're saying if I add trampoline tricks and dunks to my portfolio, to my repertoire, then, uh- We will

[00:02:45] Marcus Johnson: pay you a lot more ...

[00:02:46] Nate Elliott: eMarketer will. Okay, good to know.

[00:02:49] Marcus Johnson: That's the takeaway. Now he's gonna learn trampolining, uh, to further his career.

[00:02:57] Marcus Johnson: Uh, anyway, today's real topic, Google reinvents search as OpenAI gets cleared for takeoff.

[00:03:08] Marcus Johnson: Ask AI or just Google it Google makes a big change to a little search box, writes John Roesch of NPR. He explains that the new intelligent search box looks similar to the old one-line text box, but it's dynamic, expanding for longer queries and more chat-based exchanges. It's also multimodal, meaning users can drop in videos, pictures, files into it.

[00:03:33] Marcus Johnson: Um, did, as Mr. Roesch put it, did Google just change what it means to Google? Uh, Nate, I'll start with you. How big of a deal is Google's new intelligent search box out of 10?

[00:03:46] Nate Elliott: I'd give it maybe a 3 or a 4 out of 10- Oh ... in importance. Okay. How come? I, I actually... I, I mean, listen, uh, there are two things that sound like they're opposing ideas.

[00:03:55] Nate Elliott: They're both true. One is this is how Google wins, right? Google wins in AI by convincing people that they don't need to go to a different place- Mm-hmm ... to get AI answers. And so the more they can take the billions and billions of people who visit Google every day, every week, every month, and get them to trust that Google will give them an in-depth AI response when they're looking for it, the less likely they are to lose their enormous lead in how people discover information to companies like OpenAI and Anthropic.

[00:04:28] Nate Elliott: Having said that, they've been doing stuff like this for years. That's where AI Overviews come from. That's where AI Mode comes from. They've been doing so many things in so many different ways to show people that they don't need to go to a standalone AI chatbot to get AI responses, that they can get those responses from Google.

[00:04:49] Nate Elliott: This feels much more evolutionary to me than revolutionary.

[00:04:52] Marcus Johnson: Hmm. Jacob, what do you think? Is it- Yeah ... 'cause it, it does feel like a bit of a Trojan horse, so to speak, that they've said, "Look, come to the same familiar place that you're used to." Mm-hmm, mm-hmm. "It's just gonna be smarter."

[00:05:02] Jacob Bourne: Yeah, I mean, it... Yeah, I agree it is an incremental change, but I think an incremental change from Google still can be massive and important just due to its- Okay

[00:05:12] Jacob Bourne: central position on the internet.

[00:05:13] Marcus Johnson: Mm-hmm.

[00:05:13] Jacob Bourne: So I'd actually give it higher. I, I give it a 7 in significance.

[00:05:17] Marcus Johnson: Okay.

[00:05:17] Jacob Bourne: Not because Google hasn't been making these changes. Just because any change that Google makes is, is really significant. Mm-hmm. So I think, uh, from Google's perspective, I think this makes the AI Mode experience more akin to the standard chatbot experience, and so I think it helps it stay competitive.

[00:05:36] Jacob Bourne: Um, I also think that it's, you know... th- these agentic capabilities that, that are coming-

[00:05:43] Nate Elliott: Mm-hmm ...

[00:05:43] Jacob Bourne: um, AI, I mean, that's gonna basically make the front door to the internet kind of like an A- an AI agent essentially. So, uh, it's gonna expand access to, to, uh, AI agents in a way that just is not the case right now.

[00:05:57] Jacob Bourne: So I think that this is, is gonna be pretty impactful.

[00:06:00] Marcus Johnson: Yeah. I wanna come to that. Let me ask you a question first before we go to the agents piece. So Liz Reid, she oversees Search at Google. She recently said, "What we've seen with AI Overviews is that people don't want either just an e-" AI or the web, they want a mix of both.

[00:06:15] Marcus Johnson: Do we agree? Do people even know? We know. I wonder how much people will notice. They'll be like, "Oh, I didn't realize that it couldn't ... the search box couldn't expand before, 'cause I've never typed in a query that's, that, that's that long."

[00:06:28] Jacob Bourne: I think- I, I think people are gonna notice that. Yeah, go ahead, Nick

[00:06:32] Nate Elliott: I, I think, you know, if you're tracking that across trillions of searches, then yes, people want both the AI response and also the open web.

[00:06:42] Nate Elliott: I think what Liz is not talking about is more interesting, which is what are the situations in which people want one versus the other? You know, when- Yeah ... they talk about this being the biggest change of the search box in 25 years, it, it makes me think about all the other times they've changed the search box in the past quarter century.

[00:07:01] Nate Elliott: And by far the most common way they've done that is by adding access to non-standard search results through that single search box, right? Right now, the decision layer on Google Search determines whether you get 10 blue links, whether or not you get an AI overview, whether you see video responses or shopping responses or a news response page, a maps page.

[00:07:25] Nate Elliott: There are so many different things Google can deliver in that search results page, and this is sort of adding another layer- Mm-hmm ... to that decision layer. Yeah. And what's really interesting is the thing that they're not gonna tell us, uh, because they don't want their competitors to know, which is what are the searches, what are the queries or prompts that people do wanna see AI overviews or have an entire conversation, and what are the ones that they just want 10 blue links- Yeah

[00:07:52] Nate Elliott: or some other traditional form of Google search?

[00:07:55] Marcus Johnson: Yeah. How much of this mo- how, how much does this move us, uh, closer to the zero-click, um, world that people have been, um, kind of nervous about, uh, happening? Sarah Perez of TechCrunch was saying links will become an afterthought with the coming changes to the search results experience.

[00:08:13] Marcus Johnson: Um, how much do you think this moves it closer to, to that world where the user types in a search and then doesn't really have to click through to anything?

[00:08:21] Jacob Bourne: Yeah. I mean, we were already barreling towards that future, but I think that this does get us closer. I mean, there's a, you know, a shortcut to Gemini baked into this upgrade, so I think that that just puts the, again, the chatbot experience more front and center than, than it was.

[00:08:38] Jacob Bourne: Um, and yeah, I think, I think this is gonna become... The conversational search experience is gonna become more habitual for people-

[00:08:48] Nate Elliott: Mm-hmm ...

[00:08:48] Jacob Bourne: on a m- on a more mass scale, uh, th- through this upgrade than it was before.

[00:08:52] Marcus Johnson: Yeah.

[00:08:53] Nate Elliott: Having said that, I, you know, people often don't recognize that almost half of searches, even before AI overviews existed, almost half of Google searches were zero-click searches.

[00:09:04] Nate Elliott: I mean, it's not like- Yes ... we were at 100% of searches ended with a click, and then AI overview showed up and now none of them do. Mm-hmm. We were already halfway down that path, and AIO, based on the data I've seen, has taken us less than another halfway down the remaining half, right? It's, you know, we're still getting clicks on more than a quarter of- Search pages.

[00:09:25] Nate Elliott: Yeah. Uh, so i, I mean, it, it's pushed us further in that direction, but it's a direction we were already headed in. And I think the prognostications of, of doom that this is, you know, the death knell for the open web are, are pretty grandly overstated, in large part because we don't know what the AI answers are going to look like.

[00:09:46] Nate Elliott: Because Google doesn't know what the AI answers are going to look like. Google and OpenAI and all the other companies in this space change everything about what they do every couple of weeks. They have new models. They have new default models, uh, when you go to their, uh, chatbots. They change the way those chatbots pump out information.

[00:10:08] Nate Elliott: Is it graphic and visual? Is it not graphic? Is it not visual? They increase the number of companies cited, they decrease the number, and of course, they have been increasing and decreasing the number and the visibility of the links in those responses on a regular basis as well. We don't know what the responses are gonna look like because we're so early on in this technology that all these companies are still experimenting, still trying to figure out what their customers actually want.

[00:10:33] Nate Elliott: And to say that more AI necessarily is a death knell for the open web, I think misses the point that even the companies making the AI don't know what the AI's going to look like in six months or two years.

[00:10:46] Marcus Johnson: Yeah Um, it's a fair point. Um, Jacob, I wanna pick up on something you said, which was the agentic, um, update here, which a, a New York Times piece pointing out that Google will now offer digital assistants or agents to automate searches and do tasks over time so that someone who may be apartment hunting can be notified of a new listing without opening a real estate site.

[00:11:06] Marcus Johnson: Um, they could ask agents to search for theater tickets at regular intervals. They could send shoppers a notification, uh, when something gets, uh, put on sale, et cetera, et cetera. Jacob, what are your tho- thoughts on, on this update?

[00:11:18] Jacob Bourne: I mean, it, it comes down to how well will this actually work. I mean, I, I think it's pretty compelling to have a, a search assistant like this.

[00:11:27] Jacob Bourne: I think the, the value proposition are, is high. Um, but, you know, it, it's, it comes down to how reliable these agents are, uh, at the end of the day.

[00:11:37] Marcus Johnson: Um, Nate, how comfortable do you think people are gonna be using these agents?

[00:11:41] Nate Elliott: Well, the people who have access to these agents will be incredibly comfortable using them because right now they're only available if you pay thousands of dollars per year in Google AI subscriptions.

[00:11:51] Nate Elliott: Mm-hmm. Uh, I mean, that's a, I think a takeaway that a lot of the, the media have missed out on so far is this is only available to, I think, the two tiers of Ultra, uh, subscribers- Mm-hmm ... which cost either $100 per month or $200 per month. So we're talking about 1,200 or $2,400 a year just to use an AI chatbot.

[00:12:11] Nate Elliott: That brings it to a very small audience that is very invested, literally very invested, as well as figuratively, in using AI. Um, clearly this is the right audience to try out something like this, but I think we're a long way from this being generally available, uh, both because Google needs to decide, uh, if they're going to make this something you have to pay for.

[00:12:35] Nate Elliott: And if that's the case, then we know that 90 or 95% of Google AI users are not paying for, for, uh, uh, the use of those technologies right now. Um, and also are people gonna be comfortable using these tools? The- Yeah ... I mean, the vast majority of AI users are not the folks who are going on it 50 times a day.

[00:12:54] Nate Elliott: It's the folks who are going on it a few times a week. Mm-hmm. Uh, and so it's a pretty big jump to go from using Gemini or AI mode as sort of a, a lazy but more detailed form of googling something to building and instructing a custom agent to do tasks for you.

[00:13:12] Marcus Johnson: Yeah.

[00:13:12] Jacob Bourne: Yeah. Yeah. Yeah.

[00:13:13] Marcus Johnson: I wanna... Sorry, go check it.

[00:13:15] Marcus Johnson: You got something?

[00:13:15] Jacob Bourne: Well, I was gonna say that- Please ... I'm, I'm glad that Nate brought up the point about the cost- Mm-hmm ... because, you know, a-agentic features is, are much more expensive for, for Google to make, uh, publicly available than just your standard generative AI features. And, and so I think that that's gonna be s- just a sticking point for agentic a- adoption, y- you know, at, on a whole, on the mass scale, is that if, you know, if it's too expensive for people to either, um, you know, pay for in terms of subscriptions or too expensive for, you know, AI platforms to make available without subsc- subscriptions, then I think it's not going to...

[00:13:49] Jacob Bourne: Adoption's gonna be really slow.

[00:13:51] Marcus Johnson: Mm-hmm. Um, yeah, Google, uh, Google's Gemini Spark, so it's a AI pers- uh, personal AI agent, uh, capable of navigating a user's digital life and acting on their behalf. Um, limited number of users, making it available to those who pay for, as Nate said, AI Ultra, um, about 100 bucks a month.

[00:14:08] Marcus Johnson: Eventually, they say eventually, eventually being integrated into Chrome browser and other, other apps. I wanna pick up on what you said there about people's comfortability using this, 'cause, um, I was wondering, will consumers feel- Um, happy, comfortable giving up this level of control. Uh, so Carolina, uh, Milanesi, the president of, of consumer tech, uh, research company Cre- Creative Strategies, saying right now, uh, what you do is you ask a question, get a bunch of answers, and feel that you're in control as to which answer that you take.

[00:14:37] Marcus Johnson: Uh, if you're looking for something, uh, which product I'm going to end up buying, um, that is going to be less so going forward if you're going to say, "I want a pair of Jordans, go find them." You're not necessarily sure what steps have been taken and whether the AI has used a source or a store that was paid for and therefore came up in the search results, or if any AI actually went out and did their due diligence and picked the best one for you as a customer.

[00:15:02] Marcus Johnson: Um, what do you guys think about that?

[00:15:04] Nate Elliott: We already see a, a pretty big gap in AI usage between wealthier consumers and less wealthy consumers, and I have to imagine this is a place where that gap will widen even further. I mean, the, the notion that you're happy to instruct an AI and let it go make a purchase decision for you implies that an incorrect purchase is not a big deal in your life.

[00:15:25] Nate Elliott: Absolutely. And so, you know, that, that means that either the, the stakes here are very, very small, right? The purchase we're talking about is, you know, a new, a new order of, uh, of kitchen roll, uh, or something else relatively small. Or it means if an AI goes and spends a couple hundred dollars incorrectly or not in the best fashion, that, uh, you're either happy to eat that cost or willing to deal with the time involved in, in fixing that problem.

[00:15:52] Nate Elliott: Um, either way, uh, I mean, it really feels like something that wealthier consumers will be more likely to adopt. And sure enough, it's, you know, it, it's the, the founder CEOs at the tech conferences I go to who are on stage talking about how they can't wait for an agentic shopping future. Uh, it's not the people I engage with in my regular life.

[00:16:12] Marcus Johnson: Mm-hmm.

[00:16:13] Jacob Bourne: Yeah.

[00:16:14] Marcus Johnson: Um-

[00:16:14] Jacob Bourne: I think that these things, you know, they start with the, you know, wealthier consumers and then eventually expand. But I think in terms of just where Gemini Spark is at right now, I think Sundar Pichai kind of explained the vision pretty well. I mean, he basically said agents need to be easy to use, super secure, and really helpful.

[00:16:34] Jacob Bourne: You know, I, I agree that those are the, the three essential elements that are the very reason why we don't have agents as, you know, pervasive on the internet yet, I think 'cause there are pretty much, there are gaps in, in all three of those elements. So I think it's, it's important to remember that Spark is still in beta mode.

[00:16:53] Jacob Bourne: So-

[00:16:53] Marcus Johnson: Mm-hmm Do we think that, um, maybe, uh, these tech executives aren't reading the room? Uh, because Americans already don't trust AI very much. Right. And we were talking- Yes ... before about they, well, do they want just search or do they want AI? Do they want a mix of both? I'm wondering if there's like this disconnect between tech executive and everyday- Yeah

[00:17:13] Marcus Johnson: consumer. There's a bunch of stats I was reading for this about the trust in AI and how that's going in the wrong direction. Last three years, the first point, in the last three years, negative views of AI rose from 34% to over 50 according to YouGov. Second point, from 2020 to 2025, the share of Americans with a great deal or quite a lot of confidence in big tech companies fell from 32 to 24 according to, percent, according to Gallup.

[00:17:38] Marcus Johnson: And then from 2015 to 2019, the percentage of Americans who said tech companies have a positive effect on the country has gone from 71 to 50% according to Pew.

[00:17:46] Jacob Bourne: Yeah. I mean, I think there's a, a widening gap between, you know, big tech's, uh, moves on AI and, and, and public sentiment on AI. Yeah. So I think that they're just kind of barreling through with the hope that people will continually get accustomed and it'll continually get more normalized in society, and we'll just accept it and use it.

[00:18:06] Jacob Bourne: And, you know, I, I think with agentic AI especially, we have to assume that it's never going to be 100% perfect. It's going to make mistakes. And so I think that those mistakes are going to lead to probably rising distrust and, um, just frustration with AI. But I think even in a sort of a hypothetical situation where you have 100% perfect agent, I still think that people are going to be

[00:18:32] Jacob Bourne: This discomfort with AI is going to persist, and so I think that's gonna continue to be a, sort of a, a challenge, a hurdle for, for, for big tech kind of getting people to, to adopt-

[00:18:44] Marcus Johnson: Yeah ...

[00:18:44] Jacob Bourne: these

[00:18:44] Marcus Johnson: technologies. Yeah. And also- Can you see, can you see current consumer sentiment actually affecting adoption?

[00:18:50] Jacob Bourne: I, I do.

[00:18:50] Jacob Bourne: Oh, sorry. I mean- Go ahead. Go ahead ... especially for agents, because I, I... The trust level that you need to put into an agent to do things for you in the background- Right ... is much higher than just chatting with a chat- Right ... chatbot.

[00:19:00] Marcus Johnson: Yeah. Yep. Fair.

[00:19:02] Nate Elliott: Things besides lack of trust will affect adoption of AI chatbots and agents as well, right?

[00:19:07] Nate Elliott: We, we talk about this lack of trust, but we also, we envision this future where these chatbots, and especially these agents, will be buying things for people. As our colleague Sarah Marzano likes to point out, people actually enjoy shopping in a lot of cases. And so- Mm-hmm ... there are things that you'd rather not have to shop for, and people will be happy, possibly, to turn over those decisions and those purchases to AI.

[00:19:30] Nate Elliott: But there are lots of things, like the Jordans you mentioned before, Marcus, where people enjoy that process. They wanna go do the shopping themselves, and I think we tend to underestimate that as an inhibitor for AI adoption as well.

[00:19:42] Marcus Johnson: Yeah. Um, yeah. I wonder how much we're gonna see this kind of bifurcation of things that people happy to buy 'cause it's enjoyable, things that they're happy to let an AI, AI agent buy 'cause it's boring, like paper

[00:19:54] Jacob Bourne: towels.

[00:19:55] Jacob Bourne: Yeah. It k- it kinda feeds into this wh- whole, you know, friction maxing and analog lifestyle too, which is kind of a backlash against AI. Mm. But it's rooted in this idea that it's actually good to go out and do physical things in the world. And so on, on top of the fact that, yeah, people just like to shop- They do

[00:20:10] Jacob Bourne: it's also kinda baked- Oh ... into the, into the backlash against AI- Yeah ... uh, in and of itself, is that it's kind of... It, it spans the whole digital attitudes towards the digital, um, sphere in, in that, you know- People feel like they should be doing things Jacob,

[00:20:26] Nate Elliott: to be clear, I didn't say I was gonna leave the house to go buy those Jordans.

[00:20:29] Nate Elliott: I'm gonna sit behind my three computer screens. I'm just gonna do it myself.

[00:20:32] Marcus Johnson: So close.

[00:20:33] Jacob Bourne: Well, it could be both, right? It could be, yeah, people continue to shop online, and they're now more incentivized to, to actually leave the front door, so.

[00:20:41] Marcus Johnson: Um, so Nate, you were talking about things that can affect adoption.

[00:20:44] Marcus Johnson: A lot of different things can affect adoption. Um, speaking of adoption, uh, Google is dethroning OpenAI as the king of consumer AI, suggests a title from the, uh, an article from The Economist, uh, looking at which company is, uh, benefiting most from the adoption of AI on their platforms and products.

[00:21:03] Marcus Johnson: Catherine, um, Blunt and Rolf Winkler of the journal writing that OpenAI's ChatGPT is still by far the most popular AI chatbot, they write, and Anthropic's Claude is regarded as one of the best models for coding. But investors widely view Google as a formidable competitor. Um, Nate, what's your take on whether Google has already dethroned OpenAI as the king of consumer AI?

[00:21:26] Marcus Johnson: On artificial intelligence

[00:21:27] Nate Elliott: I don't think it has, but my goodness, it's getting close. Um, we're publishing, uh, a report about our new AI user adoption forecast in the US this week, and what we're looking at is by the end of 2026, Google's AI chatbots will be within about one million users of OpenAI's ChatGPT.

[00:21:47] Nate Elliott: Now, I, I didn't say Gemini. I said Google's AI chatbots. Yeah. And what's really interesting, of course, is, uh, we think about Gemini as the flagship AI experience at Google. I'm not sure that's true anymore. In their earnings call last month, one of the things that Google mentioned was that, uh, they now have one billion monthly users on, uh, on AI mode- Mm-hmm

[00:22:11] Nate Elliott: which is the chatbot surface baked into the search results experience. They haven't announced that, that number yet for Gemini. Um, they seem to actually have more monthly users on AI mode than they have on Gemini. Um, whatever the balance is between those two, when you add them up and de-duplicate the people who use both, we're looking at Google's AI chatbots being just, just a, a whisker away from overtaking ChatGPT in terms of monthly users in the US in 2026, and will decisively overtake ChatGPT in early 2027- Hmm

[00:22:46] Nate Elliott: accelerating f- past, uh, ChatGPT's lead at this point.

[00:22:49] Marcus Johnson: I, I thought you were gonna say that it... And they're all really good points. I thought you were gonna say that it has absolutely overtaken it because of AI overviews now being used by two-point whatever, five billion monthly users. Yeah. Um, it depends what you're looking at, which AI thing.

[00:23:03] Marcus Johnson: Google seems to have more places that it is, um, able to put AI, and so yeah, it's getting close for a number of reasons.

[00:23:10] Nate Elliott: Um- And they really should, you know, find one name and stick with it. I don't know why it's not called Gemini Overviews and Gemini Mode, but that's a, a different podcast, I think. Yeah. The, the thing I'll say is a- absolutely Google is overwhelmingly the place that people experience AI search results.

[00:23:26] Nate Elliott: Two and a half billion-plus people seeing AI overviews every month, one billion-plus people using AI mode every month, and, um, relatively close to a billion people, uh, using Gemini every month as well. Um, it is the place that people are most likely to experience AI. Um, our forecast does differentiate between people who choose to prompt an AI chatbot and people who simply see AI summaries of things like search results and product reviews.

[00:23:55] Nate Elliott: And so- Mm-hmm ... from that perspective, we don't count people who only see AI overviews but don't go to a chatbot like Gemini, AI Mode, or ChatGPT and enter a text prompt. We don't count them as AI active users. Um- Okay ... but, but even if you are just looking at the people who prompt chatbots, Google will overtake OpenAI for monthly usage and monthly users, uh, in the first quarter of 2027.

[00:24:20] Marcus Johnson: Yeah.

[00:24:20] Jacob Bourne: Yeah. Yeah. I mean, I would echo, echo that and simply say that You know, Google is in the process of dethroning OpenAI.

[00:24:27] Nate Elliott: Mm-hmm.

[00:24:28] Jacob Bourne: And I think that Google's structural and distribution advantage was there from the very beginning, but OpenAI got its first mover advantage because as a startup, it could take a lot more risks than Google could take.

[00:24:41] Jacob Bourne: And so I think that helped it both on the innovation and commercial- commercialization fronts.

[00:24:45] Marcus Johnson: Mm-hmm.

[00:24:46] Jacob Bourne: Um, but I think we're in a situation now where AI has become normalized. It's seen as just a part of doing business. Um, it's also increasingly seen as critical infrastructure, and so I think Google feels kind of more emboldened to move ahead, um, with less caution than, than it, you know, it, it did in the early days.

[00:25:05] Marcus Johnson: It d- it does seem as though Google said, "We're gonna put AI everywhere," and it's able to just, um, suffocate, uh, OpenAI i- in that regard because it, it... I mean, Gemini is soon gonna be a staple on all of basically the most popular ph- uh, half of the phones that people in America use, um, in, in the Apple iPhones.

[00:25:25] Marcus Johnson: Um, it's, uh, well, it's ideal to, to basically power Siri. It's already included in Android's And- uh, Google's Android devices. Um, and, uh, Miss, uh, Malnessi of Creative Strategies, who I referenced before, was saying that Google, what that Google has that other A companies- AI companies lack is the cultural cachet, saying that with the Gemini model integrated into so many popular services that people rely on every day for work and play, Google has made it much more likely that folks will interact with its artificial intelligence.

[00:25:52] Marcus Johnson: She goes on- Indeed ... to say it's not about just how good the model is, but where it is- Mm-hmm ... and how easy it is for people to discover Gemini and then get use out of it.

[00:26:00] Nate Elliott: Yeah. Yeah, I mean, it is also a very good model, right? It, I mean, it, it, it took some time to catch up because it launched later than some of the other models out there, but it is a good model that people find very useful for a lot of consumer-facing applications.

[00:26:15] Marcus Johnson: Yeah.

[00:26:16] Nate Elliott: But yeah, I mean, Google has the world's number one search engine, the world's number one browser, the world's number one mobile operating system, the world's number one email And web server and image folder and, I mean, the list goes on and on and on. Um, the, the, the day that it was decided that Google was going to win in consumer AI was the day that Google scientists invented the transformer.

[00:26:41] Nate Elliott: Um, there was never a chance that they weren't going to win or at least be, like, in a very competitive top two in this- Yeah ... category because they have so many advantages on distribution. And as Jacob said earlier, the- they're structural advantages. They can build themselves into the most popular everything.

[00:27:00] Marcus Johnson: Yeah. Nate, you said it is also a good model. Um, I wonder what you make of this number then. There was a recent analysis from New York Times found that Google's AI-generated responses were correct 90% of the time. Google has disputed the study. Is it that people only really need 90% or is it that even if ChatGPT has a model that's 92%, 90 is good enough and, um, really it just matters more where it is as opposed to it being a better model?

[00:27:28] Nate Elliott: I, I don't know what the number is that that study would've assigned to ChatGPT. I do know that within the last six months, OpenAI has acknowledged that 64% of ChatGPT responses that include product information include inaccuracies. Hmm. Two-thirds of their product responses included inaccuracies in the base model when they published that data at the tail end of last year.

[00:27:50] Nate Elliott: Now, six months is a long time in AI, and maybe they've gotten the 64% error rate down to 34% or 24%. But whatever that number is, um, it's too high. Yeah. Uh, I think 10% inaccuracy, uh, Google would say is too high if they were willing to acknowledge- Yeah ... that number. Um, but I also, I, I find it really interesting that we compare all of these models against perfection, right?

[00:28:16] Nate Elliott: We, we talk about, uh, you know, what percentage of the time do the models make mistakes? And Jacob mentioned earlier that the, that the shopping agents would make mistakes and, and that that would be an inhibitor for people, and I, I absolutely agree. It will be an inhibitor for people. But by the way, I make mistakes when I'm shopping.

[00:28:34] Nate Elliott: I click on the wrong product, or I accidentally put two things in my cart. I make mistakes when I'm providing information to people in eMarketer reports or on this wonderful podcast. Um, it's fascinating to me that we compare the models against perfection- Mm-hmm ... instead of comparing them against what we ourselves can accomplish- Interesting

[00:28:53] Nate Elliott: if we're doing the same task.

[00:28:55] Marcus Johnson: Mm-hmm.

[00:28:55] Jacob Bourne: Yeah. Yeah. I mean, I think that comes down to ... Th- there's just a, a comfort level there where I think we are kind of more, maybe more lenient towards human imperfection than we are with AI imperfection- Mm ... when we can spot it. I think the other thing is we don't always 100% know when AI is being inaccurate.

[00:29:13] Jacob Bourne: Yeah. And so I would agree that that 90%- Right ... is, is not accurate enough. I also would say that users aren't super obsessed, in my opinion, with comparing these ever-moving benchmark targets. And so, you know, a percentage off here and there, give or take, is, is not gonna move the needle. I think it's more about access, like you pointed out, Marcus.

[00:29:33] Nate Elliott: Yeah. I think the people who are gonna get access to Gemini Spark Beta probably care quite a lot- Yeah. ... about those benchmarks, and- That's huge ... all the folks not paying thousands of dollars per year for AI subscriptions care, as Jacob said, probably not much at all.

[00:29:45] Marcus Johnson: So Cade Metz and Mike Isaac of The New York Times explaining that Elon Musk, uh, had a $150 billion lawsuit against OpenAI and Sam Altman.

[00:29:54] Marcus Johnson: Uh, it was quickly rejected by a federal nine-member jury in the US District Court in Oakland, California. They found that Mr. Musk had failed to file his lawsuit within a timeframe required by law. Mr. Musk had accused OpenAI, its chief executive, Mr. Altman, and its president, Greg Brockman, of stealing a charity He said, uh, by attaching a commercial company to OpenAI, which was founded as a nonprofit, and taking billions of dollars in investment from Microsoft.

[00:30:18] Marcus Johnson: Uh, Blake Montgomery of the British newspaper, The Guardian, saying, "The jury's decision, affirmed immediately by the judge's dismissal of all the charges, provide the AI firm with a stamp of approval for its for-profit plans already in motion and a clear path ahead to go public later this year at around a trillion-dollar valuation.

[00:30:33] Marcus Johnson: OpenAI's plans now seem to be all but guaranteed, given that the world's richest man couldn't put a stop to them." Jacob, any takeaways from the Musk-Altman trial?

[00:30:42] Jacob Bourne: Yeah, I mean, I think the, the trial itself probably in the, in the public eye played out more like a circus than a serious court proceeding. I think, you know, on the one hand, a lot of people really dislike Musk, and on the other hand, a lot of people are really distrustful of AI, and I think OpenAI is kind of a, a poster child for AI.

[00:31:03] Jacob Bourne: And so I think, you know, the, the verdict, of course, we know how it came down to the statute of limitations, and I think essentially it was difficult for, for Musk to get around the optics of the fact that he waited until he had a, you know, a, a competing, uh, for-profit s- AI startup, uh, to, to, to file his suit.

[00:31:22] Jacob Bourne: Um, I think it's also safe to say that neither Altman nor Musk kind of emerged from this unscathed in terms of their image. Um, also it seems to me that the image of AI probably took a hit from this as well, and I think the public- Hmm ... probably views this as essentially a power grab among tech giants.

[00:31:42] Marcus Johnson: Yeah, Nate, to what Jacob just said, David Streitfeld, uh, and N-Natalia Rocha of The New York Times writing, "AI critics dismissed the three-week trial in federal court in Oakland as a power struggle between oligarchs that was of little concern to the masses." Close quote. What do you, do you take away from the trial, if anything?

[00:32:01] Nate Elliott: Yeah, I, I mean, I agree with that assessment. Um, but I also think we're far from having heard the last of this. Um, uh, Musk doesn't tend to let things slide. Uh, I imagine he'll be back with further action. Um, I mean, the, the thing that seems true is that all the founders of the various AI companies think that AI in the wrong hands is incredibly dangerous, possibly at an existential level for humanity, and they all think that they're the only ones who can be trusted, and that all the other AI founders are the existential threat.

[00:32:38] Nate Elliott: But perhaps the wildest takeaway for me out of all that is we didn't get to find out what a jury would have decided. Um- Right ... did they steal a nonprofit? Right. Because you can make a pretty clear argument that they did, just as OpenAI can make a pretty clear argument that they didn't. And the notion that if they did do that, and again, there was no finding of fact as far as I can tell in either direction, but the notion that they may have done that, and we don't get to find out if they have done that simply because a clock expired is-

[00:33:09] Marcus Johnson: Mm-hmm

[00:33:10] Nate Elliott: just wild to me.

[00:33:11] Marcus Johnson: Yeah. Yeah, it was interesting as well 'cause OpenAI's lawyers were arguing and showing that Mr. Musk was, when he was still at the company, uh, OpenAI, 'cause he helped, uh, found it, he was- had repeatedly tried to transform the lab into a for-profit company, including an effort to fold OpenAI into his electric car company, Tesla, and they were arguing the only reason he sued is because, uh, OpenAI had become very successful, uh, after he'd left.

[00:33:34] Marcus Johnson: Um, but yeah, they basically said that you weren't able to even, even sue in time, and so that's why it got thrown out. Uh, and then with this, Professor Sarah Kreps, Director of Tech Policy Institute at Cornell University, saying, "The trial served as a reminder of how much the future of AI still depends on a remarkably small group of powerful tech figures and their personal rivalries, highlighting a broader disconnect between the people building these systems and the many, uh, people increasingly expected to live and work alongside them."

[00:34:01] Marcus Johnson: That's all we've got time for. Best,

[00:34:02] Nate Elliott: best

[00:34:02] Marcus Johnson: quote of

[00:34:02] Nate Elliott: the whole episode.

[00:34:03] Marcus Johnson: I know. Sarah, big, big- Because it came from

[00:34:05] Nate Elliott: Cornell.

[00:34:07] Marcus Johnson: Okay. Uh, um, affiliation with, uh, Cornell perhaps?

[00:34:13] Nate Elliott: Uh, yeah, small one.

[00:34:14] Marcus Johnson: Yeah, there we go. Yeah. Uh, that's what we've got time for this episode. Thank you so much to Nate.

[00:34:19] Nate Elliott: Thank you very much.

[00:34:20] Marcus Johnson: And to you, Jacob.

[00:34:22] Jacob Bourne: Thanks for having me.

[00:34:22] Marcus Johnson: Yes, indeed, and to the production crew. We've got Lance helping out with this one. To everyone for listening to Brand Names: The Marketer Podcast, uh, subscribe and follow to hear about new content. Suzy will be here Wednesday talking about the Walmart of the UK, Tesco, and I'll be back Friday discussing how brands are investing in real world experiences.