Name: Brainstorm AI 2024: Google’s Strategies For Search
Uploaded: 2024-12-10
Duration: 13 min 8 s
Description: Liz Reid, Vice President, Search, Google Interviewer: Jeremy Kahn, Fortune

Liz Reid, Vice President, Search, Google Interviewer: Jeremy Kahn, Fortune

Brainstorm AI 2024: Google’s Strategies For Search

Conferences

Liz. Thanks for being here with us. Thank you for having me. It's a great room. I can't see you at the back. But um I want to warn everyone, I'm gonna go to questions pretty early in the session. So please think of your questions. Um But first I want to ask you was, there's been a lot of reporting out there in the last few weeks that young people in particular are starting to Google things less. They're not going to traditional Google search. They're using some kind of A I chatbot or they're using these new A I enabled search engines, things like perplexity. What, how are you viewing this trend at Google? And how do you sort of want to trying to combat that potential loss of market share? Yeah. So I mean, there's always been different options with search and we welcome that. Um I think what we're really focused on with search is the user um and really helping that user find information. Um The truth is that there are not just billions of questions that people come to Google for. There's billions of questions people don't even ask and they don't ask because it's too hard, they don't ask because they're not confident they can get that answer. Um And so we really think about and search, how do we make it so that you can really ask anything you're curious about? And then it really feels effortless that it's so easy to do that. You don't think twice before you do that. And we think there's a huge opportunity ahead on that. Interesting. You mentioned, you know, people come to search because they need answers and they want quality information. And there's a lot of concern right now about what's happening to the information ecosystem on the web. There was already a problem sort of even before ja I exploded about search engine Optim A and whether that was degrading the quality of information that was available on the internet, then there's a, you know, a doubling of that concern with generative A I and the fear that's producing a lot of, you know, a I quote unquote swap that's out there. And how are you, how are you sort of seeking to address this problem at Google? Is it, is that our concern now that you're OK? You know, we have a great search tool, but it's got to be able to surface information that's, that's of value to people. So ensuring that we surface high quality information is, is definitely paramount and it's a huge reason that people have trusted search. Um This, this challenge has existed since the beginning of search, right? Um People have constantly created amazing content for the web and they've created poor quality um content in many ways. It's a, it's a cat and mouse game with those who wish to create um spam across. Um I would say A I generated content isn't by itself intrinsically bad, right? You can use A I um to really help people write and create and produce amazing information. Um But there's also this challenge of scale content abuse. Um And this is something we take very seriously. It's something we've continued to work on for many years. That's the level of spam and search is very low. Um And we'll stay focused on it because it's absolutely paramount. Um You've now rolled out these A I interviews, sorry A I overviews pretty widely that's created a lot of concern on publishers, the part of people and other businesses that really depend on search traffic coming from Google, you know, to their own websites where they seek to monetize that often through advertising. What, what are you saying to publishers about where this technology is heading? Because there's a lot of concern in the industry about, about what the effect of this is going to be? Yeah. So I think we have a long history in search and having a value exchange with the ecosystem and we've sent billions of clicks for a long time. Um But one of the things that there's sort of this common misconception is that people either want answers or they want links. Um That's not actually what we find from research. Um People wanna get started, but they actually really want to hear from other people. They're interested in their perspective, they're interested in their understanding of the space. And so we really tried to build A I overviews and continue to iterate them on a way that enables people to get started, but then continue to connect to the web. Um And we see that that's one of the things people love um about A I overviews is this ability to have that hybrid to understand and then to continue on in your journey. And I think that's only going to continue. We see that people, especially younger users really want to hear from the authentic voices, they want to hear from the people they trust and that's part of what makes it special, connecting to the web. And so that's at the heart of how we think about in the A I transformation still enabling that mission of connecting people to the best of the web. And how can actually A I make that even better in allowing people to connect to truly the best of the web? Are people still clicking links with A I overviews? We do see people click both within the A I overviews as well as on the rest of the page. And in fact, when we see A I overviews, we see that the clicks that go out to the web are of higher quality. People spend more time on the sites they go because these are really giving the information that they want. All right questions from the audience who's got a question right here. This gentleman has a question that hand went up very quickly. Um Wait till the mic gets to you if you can let us know who you are. Yeah, Josh Legasse with insignia. I didn't even know that was a thing. It just started popping up and I love using this. I'm glad to hear that. That's great to hear. Uh agree with the clicking back and forth. My question is, is it learning from me and feeding back answers to me or is that learning from everyone and feeding back answers to me or is there something else that's happening? Um So I think there's different levels of, of your question, there's sort of um what is the level of contextualization that we think about in these spaces? Um And there's some understanding of like, how do we interpret the query that can be more contextual. Um But in general, we use um throughout search, not just in A I reviews but broadly, um we try and learn from very privacy safe way um giving people controls, but we can learn a lot. This is how people learn these are the best sites to go to, right, is often based on user feedback. In addition to our high quality signals. And so we try to do it in a way that's very clear that gives people agency. Um but that we can actually improve based on the usage of the product. Other questions from the audience. Oh How about this woman back here? There we go, wait for the panel. Hi, I'm Julia Martinson with service. Now, um I mean, Google search has obviously been, you know, the best for a long time in terms of search and, and also providing the right results, you know, you know, ads kinda maybe, you know, for some more cumbersome, for more, some more enjoyable. Um I'm wondering with seeing how LL MS went also into more domain specific small language models. I wonder if um maybe, you know, a road map question, would it be ever, could we ever see a domain specific research basically where you're also kinda going down that level and, and making it more relevant for that specific domain you're in. Um I think we've continued to look at um uh different domains in different ways. I mean, if you come to search today, we have um a mode that enables you to do images, right? If you really want to go deep in images, um we have the ability to go deeper on shopping or on local or on others. Um And I think um that will continue, there will be cases where what really people want is a specialized um understanding of information, I also think with A I the ability to just um personalize the, the responses um evolves in ways that we don't traditionally think a lot of people previously with personalization with ranking, think of it as like, OK, I'm gonna re rank just the site. OK. Um But actually there's a lot of different ways um in which personalization can be useful. OK? Um it starts with basics like some people love video results and some people really do not. OK. Um Can you understand that? Um what's the level of understanding of what you want and responses? Are you an expert in that domain? And we should assume so or are you, you know, sort of trying to help your five year old with a concept that you don't actually know very much and explain it to her? And how can you do that? Um Are you on the middle of a long journey or on a shorter journey? Do you like to scan information quickly? Are you an audio learner? And then we should make it easier to understand information in that form? Um And so I think besides getting better within domains, there's just a huge opportunity to start to understand like what does it really mean to make information effortless if you go back to the mission? Um Most people like if they repeat the Google mission kind of start stop halfway through, they go organize the world's information and that's kind of where they stop. But there's this part at the end about universally accessible and useful. Like if we show you information that's organized, but it is not useful to you. It's not really that useful, right overall. And so I think there's an opportunity with A I to rethink what it means to give information in a way that's truly useful to you that truly feels effortless to consume and interact with. There's a lot of people really like the A I overviews and a lot of business is use Google search either on their websites to allow customers to search for things or they use it some internally on their own information. Is there the ability to kind of take the A I overview kind of technology and have that work on your own site or on your own information? Yeah. Um So there's a bunch of different offerings that Google Cloud has. Um uh And I will confess that I'm not the expert in Google Cloud. So you may want to get additional information on this. Um But at a high level, um we both have services that allow you to have your own data um basically ground a response in the data on your site. Um There's also a service that allows grounding with search so that you can get the best of sort of search his understanding of what's high quality information in your product. Um And so cloud offers a variety of different surfaces so that we can actually take some of the, the key underlying LL MS and understanding about how to get high quality information and make it available for others. I know there was at least one more audience question this gentleman right there. Let's get him in. Um, and then maybe I'll have time for one more of my own. But hi, my name is Matt. So I'm involved in running Ja I at mcdonald's, I'm curious, Google's famously brought search together in like one bar that you can do everything and, and yet now you're sort of splintering it again. You've got your Gemini product, you've got your Gemini app. Where do you see this going? Do you see it sort of coming back to one interface or do you see sort of the chat interface and the search interface? That's a great question. I swear. I didn't set you up for that because that was the other question I was going to ask is are the products going to blur together? You have Gemini as a separate product from search is that, are those going to be kind of recombined? I don't think we honestly know what the future has. We're pretty early in the journey. I do think that one of the things I think about a lot is how do you meet users where they are? And so, so if they come to search, can we give them the information the form they want if they choose to use Gemini. Can they do that? Um If you actually go back? And like long enough ago, people ask this question about youtube and Google's video search. They asked this question about Google Maps and um local search on search. Like, wait, where should I search for a restaurant on search or maps? Should we just merge search and maps together? Really? What you want to do is make it e like whatever is easiest for the user to come forth. Um The other thing that I would add in that space is that I think um you'd mentioned the keyword box, OK? Actually, search is evolving. So the keyword box is only a component, right? We introduce voice, we introduce hopefully all of you know what lens is. If not, you should come find me afterwards, that ability to search for what you see. And I think the space of information is going much more multimodality in its input. And the reason is because, you know, we're humans that existed a long, long time before technology and the way that we understand the world and we think about the world is based on a lot of different senses, right? It's based on what we see and what we hear and how we talk across. And so if you want to make information effortless, we're going to continue and search to think about what is the most natural way you would express if you had no constraints. If you assume tech was not a barrier. And you just asked your friend as easy as possible, how can you have that experience with search? So, so the tech really comes to you instead of you having to translate your need to the tech. Well, you've got this Google has, as everyone knows this department of justice case that's out there and there's just been this remedy phase and some of the suggested remedies involve the separation of search from other products. Are you guys actually sort of planning internally for, if you had to implement one of those remedies, what that would look like? So I don't have more really that I can comment on regards to this. What I will say about search is that we have consistently found that users choose search because they really find it the most helpful. And I think we're going to continue to really focus from, from my point of view on how do we help users? Got it. There's a lot of talk about sort of agentic A I and people may be using other agents increasingly, you might have people coming to Google search that are not actually human. It's, you know, you have an agent coming. How are you thinking about that? The idea that the user may not actually be a person, I think in general, you know, our, our goal at the end of the day is to help the user um as people interface or as as agents interface with search in different ways. Um I think we will always try and evolve it to the question about like, OK, how do you help the agent? So the agent helps the user. But at the end of the day, we're really focused on helping people. Um and we'll continue to innovate in that space, right? And can you give us any sense of, you know, road map is that we're going to see more multimodality? What are we going to see from search going forward? Um I think you will definitely see more multimodality. Um I think in general you will continue to see this evolution such that it feels like accessing information, feels just much easier, it feels more effortless, it feels like the question that you actually have in your mind. Um Is the question you get to ask. You don't have to break it down. You don't have to translate. You don't have to worry about what language you speak. You don't have to worry about how you learn best. Um And that search just becomes much more accessible in a new and exciting way. Great. Fantastic. That's all the time we have for uh for it today. But uh thank you to Liz and thank you all for listening.