00:00:00
Andru Edwards: We are here with Google's Robby Stein.
00:00:03
Andru Edwards: Robby, what do you do for Google?
00:00:05
Andru Edwards: VP of product, Google Search, working across many of our product experiences,
00:00:08
Andru Edwards: including a lot of the new AI search experiences you heard about today,
00:00:12
Andru Edwards: AI overviews, AI mode, lens, and more.
00:00:14
Andru Edwards: All right, all right.
00:00:15
Andru Edwards: Let's just jump right in.
00:00:17
Speaker 2: All right, let's go.
00:00:18
Andru Edwards: Tell me about the big news, at least under your purview,
00:00:21
Andru Edwards: that was announced today
00:00:22
Speaker 2: as
00:00:23
Andru Edwards: it pertains to things like AI mode.
00:00:24
Andru Edwards: Yeah, biggest news today by far is that we were bringing AI mode out of labs,
00:00:28
Andru Edwards: is rolling out two people in the US starting today.
00:00:32
Andru Edwards: So you'll start to see this show up as a tab
00:00:35
Andru Edwards: in the core Google Search Experience
00:00:36
Andru Edwards: and within all of the Google apps
00:00:38
Andru Edwards: as a quick way to get access to information.
00:00:40
Andru Edwards: And how it works is you can ask anything on your mind now,
00:00:43
Andru Edwards: type a normal question.
00:00:44
Andru Edwards: You can take a photo, ask a question about what you see,
00:00:47
Andru Edwards: and Google configure it out using advanced,
00:00:50
Andru Edwards: state-of-the-art modeling, and all of the understanding
00:00:52
Andru Edwards: of the open web and Google's knowledge systems
00:00:55
Andru Edwards: across shopping, finance, sports, and everything else.
00:00:58
Andru Edwards: Okay, AI mode to me, when I saw it at least, feels like a radical redesign of Google search,
00:01:04
Andru Edwards: possibly the most radical redesign ever since the beginning of the one box with the search,
00:01:09
Andru Edwards: and I'm feeling lucky, right?
00:01:11
Andru Edwards: Which I think is fascinating.
00:01:12
Andru Edwards: I'd love to know some of the backstory behind what sparked this change.
00:01:17
Andru Edwards: Was there a risk, for example, of Google kind of being disrupted by AI as people started turning to AI chatbots?
00:01:23
Andru Edwards: I think the big learning for us is that we started working on AI overview.
00:01:28
Andru Edwards: use, which was the summary at the top of search for questions that are pretty specific.
00:01:33
Andru Edwards: So if you were to ask, how do I get ketchup out of a couch, I have two young toddlers,
00:01:37
Andru Edwards: you're pretty predictably see a little AI snippet now at the top of the page that has helpful
00:01:42
Andru Edwards: information and links to dive in.
00:01:44
Andru Edwards: And what we found was that people wanted more of that.
00:01:46
Andru Edwards: And they actually wanted more control over getting it.
00:01:48
Andru Edwards: And they wanted to ask follow-up questions.
00:01:49
Andru Edwards: They wanted deeper experiences based on that.
00:01:51
Andru Edwards: In fact, people were actually adding AI to the end of their queries because they wanted to generate
00:01:56
Andru Edwards: these little AI blurbs.
00:01:57
Andru Edwards: Did that work?
00:01:58
Andru Edwards: No.
00:01:59
Andru Edwards: OK.
00:01:59
Andru Edwards: Well, it will now.
00:02:02
Andru Edwards: Not that exact thing.
00:02:03
Andru Edwards: But we basically decided that, well, this is--
00:02:06
Andru Edwards: the people are coming to Google with such different needs now.
00:02:09
Andru Edwards: Why can't we give an experience that really
00:02:12
Andru Edwards: lets them have full control?
00:02:13
Andru Edwards: And for many questions where AI we think is most useful,
00:02:16
Andru Edwards: these are typically the harder, more advanced,
00:02:17
Andru Edwards: more specific questions people have in mind.
00:02:19
Andru Edwards: You can just go right to AI mode now and use it.
00:02:22
Andru Edwards: And it does represent to us, like the frontier of what
00:02:24
Andru Edwards: Google Search can do, and the most advanced version
00:02:26
Andru Edwards: of what AI in search.
00:02:27
Andru Edwards: you know, can do for you.
00:02:30
Andru Edwards: One thing about Google search,
00:02:32
Andru Edwards: the traditional Google search, very powerful
00:02:34
Andru Edwards: if you knew how to put in
00:02:37
Andru Edwards: the different search operators, right?
00:02:40
Andru Edwards: And AI mode is really more about natural language.
00:02:43
Andru Edwards: How do you see casual users
00:02:46
Andru Edwards: versus power users, for lack of
00:02:47
Speaker 2: a better term,
00:02:48
Andru Edwards: using AI mode differently?
00:02:50
Speaker 2: Yeah,
Andru Edwards: so I mean, I think with AI overviews,
00:02:52
Andru Edwards: people are starting to learn
00:02:54
Andru Edwards: that you can ask more natural language questions
00:02:57
Andru Edwards: and get to AI at Google, without AI mode even.
00:03:00
Andru Edwards: So, we are seeing more and longer questions
00:03:03
Speaker 2: just
00:03:04
Andru Edwards: in search
00:03:05
Andru Edwards: because it shows AI overviews a large percentage of the time.
00:03:08
Andru Edwards: So I think there's a learning effect that we're seeing
00:03:09
Andru Edwards: that's driving actually the growth that Sundar talked about today
00:03:12
Andru Edwards: in AI overviews, which is for these kind of specific questions
00:03:15
Andru Edwards: people have in their mind where AI overviews typically send up,
00:03:18
Andru Edwards: particularly show up.
00:03:19
Andru Edwards: There's a 10% increase in searches like that this year
00:03:24
Andru Edwards: after the release of AI overviews,
00:03:26
Andru Edwards: which is one of the fastest growing products launched this decade.
00:03:29
Andru Edwards: Bringing and coming into AI mode,
00:03:31
Andru Edwards: I think there's already a belief now
00:03:32
Andru Edwards: that there's some ability to use these longer questions,
00:03:35
Andru Edwards: but people can take that to a much larger degree.
00:03:37
Andru Edwards: People are copying and pasting in whole blocks of code,
00:03:41
Andru Edwards: of homework questions,
00:03:43
Andru Edwards: of medical information that they're copying from some report.
00:03:46
Andru Edwards: They're asking questions about it.
00:03:47
Andru Edwards: And because you can have a back and forth
00:03:49
Andru Edwards: and you guarantee this AI experience
00:03:50
Andru Edwards: is unlocking a lot more needs that can be,
00:03:53
Andru Edwards: solved by Google now.
00:03:55
Andru Edwards: What does adding AI
00:03:57
Andru Edwards: into the search experience
00:04:00
Andru Edwards: allow people to do now that they
00:04:01
Andru Edwards: weren't able to do before? How is that really
00:04:03
Andru Edwards: changing? Yeah, I'll give you a few examples.
00:04:05
Andru Edwards: I had a
00:04:08
Andru Edwards: discoloration of some tile in my
00:04:09
Andru Edwards: bathroom. It was getting really weird in there.
00:04:11
Andru Edwards: I didn't know what it was. I know what to do with it.
00:04:13
Andru Edwards: It was very peculiar color.
00:04:15
Andru Edwards: And so I took a picture of it using lens.
00:04:17
Andru Edwards: And then I asked, what is this and how do I get
00:04:19
Andru Edwards: rid of it, basically, in natural language.
00:04:21
Andru Edwards: And then the model was an AI mode
00:04:23
Andru Edwards: multimodal question with image.
00:04:25
Andru Edwards: The model did segmentation
00:04:27
Andru Edwards: using lens context. It figured
00:04:29
Andru Edwards: out what the thing was, like some mold
00:04:31
Andru Edwards: of some kind. It explained different ways that I could
00:04:33
Andru Edwards: clean it myself and then also found vendors
00:04:35
Andru Edwards: in my area with phone numbers of who
00:04:37
Andru Edwards: could professionally have it cleaned. That's a
00:04:39
Andru Edwards: pretty tremendous leap
00:04:41
Andru Edwards: in what's possible
00:04:43
Andru Edwards: on search now.
00:04:45
Andru Edwards: For example, I was comparing
00:04:48
Andru Edwards: products that I was researching
00:04:49
Andru Edwards: for my daughter, who's
00:04:51
Andru Edwards: having a birthday coming up. And I wanted to put a bunch
00:04:53
Andru Edwards: of products in. It's not clear anyone's ever compared these products before, but because an AI
00:04:57
Andru Edwards: is kind of the ability to, from first principles, look at each underlying product, look up the
00:05:03
Andru Edwards: product information of each of them within Google's shopping graph, and then it can make comparisons
00:05:09
Andru Edwards: and analyze it and then produce a table that shows the differences. It's pretty incredible.
00:05:14
Andru Edwards: So those are all these really profoundly new things you can do, and that's why we really
00:05:18
Andru Edwards: believe that in this new era, it's more, it's less about, it's this evolution from
00:05:23
Andru Edwards: you know, information to intelligence.
00:05:26
Andru Edwards: And it's really about being able to ask anything.
00:05:28
Andru Edwards: And it's like the user is actually the biggest constraint.
00:05:32
Andru Edwards: Like you don't realize you can ask such hard questions.
00:05:34
Andru Edwards: Right.
00:05:35
Andru Edwards: And so the challenge to you and everyone watching is to put really, you know,
00:05:39
Andru Edwards: specific and hard questions into Google and see what happens.
00:05:41
Andru Edwards: And speaking of which, it kind of led me to this next thing that I thought was incredibly
00:05:44
Andru Edwards: impressive.
00:05:46
Andru Edwards: I think it's called Query Fan Out?
00:05:47
Andru Edwards: Yeah.
00:05:48
Andru Edwards: Okay.
00:05:49
Andru Edwards: Tell me more about how Query Fan Out works.
00:05:50
Andru Edwards: As they were presenting it today, it almost felt like.
00:05:53
Andru Edwards: Like, you have dozens of little employees going out or doing searches for you, and then
00:06:01
Andru Edwards: you have someone at the tail end who can kind of take everything in, collated together, and
00:06:06
Andru Edwards: give you a presentation based on all these searches.
00:06:09
Andru Edwards: Like, that's my
00:06:10
Speaker 2: simplified version.
00:06:11
Speaker 2: That's exactly, that's exactly what happens.
00:06:13
Andru Edwards: So it happens to me you put a query into AI mode.
00:06:15
Andru Edwards: First, a model starts to understand the query, and it makes a plan.
00:06:19
Andru Edwards: What's this person trying to know?
00:06:20
Andru Edwards: And what are the things it needs to research in order to know more?
00:06:23
Andru Edwards: and it will generate potentially dozens of other questions,
00:06:26
Andru Edwards: and it literally uses Google search as a tool.
00:06:28
Andru Edwards: So it's Googling, like in the background,
00:06:30
Andru Edwards: all these questions that it thinks it needs to know about.
00:06:33
Andru Edwards: And then each Google query is connected
00:06:35
Andru Edwards: to the vastness of the web and Google's knowledge bases.
00:06:38
Andru Edwards: So it's 50 billion products in the shopping, graph,
00:06:41
Andru Edwards: 250 million location places information,
00:06:45
Andru Edwards: real live time data for stocks, for weather, for finance,
00:06:48
Andru Edwards: all available into this model.
00:06:51
Andru Edwards: And it can reason and put attention on any piece of that
00:06:53
Andru Edwards: information that things is useful. So you're booking tickets to the game. It could look up weather.
00:06:56
Andru Edwards: It could look up reports asking what good seats are, what the best seats are at that game.
00:07:00
Andru Edwards: And it could look up a number of things in order to generate this response. And then the last
00:07:04
Andru Edwards: thing it does is it takes all that info in. It generates response, but even further, it renders
00:07:09
Andru Edwards: the design itself and outputs the visualization and all the way down to the HTML of how
00:07:17
Andru Edwards: it shows up on the page. So if it's data, it could decide to visualize it using a charting
00:07:22
Andru Edwards: library or if it's a product it can make a grid and it is like a tray or a grid of images
00:07:28
Andru Edwards: so you can just browse through these like you know products so it's not a predetermined
00:07:32
Andru Edwards: template for the results it's generating the visual look as well
00:07:37
Speaker 2: many of them are generated
00:07:39
Speaker 2: on the block yeah
00:07:40
Andru Edwards: okay that's right what changes with the query fan out process when you're doing
00:07:45
Andru Edwards: deep research it's very little it's a very similar overall process but it has deeper thinking
00:07:52
Andru Edwards: involved and the model is given much more time to ask questions, do research, and have a multi-step
00:07:58
Andru Edwards: experience. Because instead of waiting a few seconds, people might be willing to wait a few
00:08:01
Andru Edwards: minutes in that case to get an expert length, thorough report. So it might do research,
00:08:07
Andru Edwards: it might think about what it's seen. And traditionally in AI mode, it does this too. It could do
00:08:11
Andru Edwards: follow-up questions to learn if it feels like there's a gap in its knowledge. But if you have
00:08:15
Andru Edwards: minutes to think, you could do that numerous times and you can maybe search hundreds of things
00:08:20
Andru Edwards: during that time, and then bring a very comprehensive, very deep report for someone.
00:08:24
Andru Edwards: Okay.
Andru Edwards: I switch over to live search.
00:08:26
Andru Edwards: Yeah.
00:08:27
Andru Edwards: There were so many things in this keynote that blew my mind, guys.
00:08:30
Andru Edwards: So, live search.
00:08:31
Andru Edwards: You can take your phone, point it at an issue, for example, a broken bike, and have Google
00:08:37
Andru Edwards: walk you through, listen to you, and kind of be your partner in figuring something out.
00:08:44
Andru Edwards: Again, mind-blowing, including going through an instruction manual, including.
00:08:50
Andru Edwards: including going through YouTube and finding the proper video.
00:08:53
Andru Edwards: I guess my question is, how close are we to real-time AI becoming your side-by-side assistant,
00:09:01
Andru Edwards: or are we already there?
00:09:02
Andru Edwards: Yeah, I think in many ways we are already there.
00:09:04
Andru Edwards: I think you saw the video that showed some of this Astra capabilities that we are bringing
00:09:09
Andru Edwards: into search.
00:09:10
Andru Edwards: And really what this means for most people, how to think about it is you have this kind of world
00:09:16
Andru Edwards: knowledge AI that knows how to research, how to use Google search.
00:09:20
Andru Edwards: search and how to access all the information within Google Search's knowledge bases.
00:09:25
Andru Edwards: That model and that knowledge base can be used in different ways.
00:09:30
Andru Edwards: One, it can spend a lot of time thinking and generate a really long report, or you could
00:09:35
Andru Edwards: use that same information and just talk to it live and interact with it just like you would
00:09:39
Andru Edwards: a phone call.
00:09:40
Andru Edwards: And in that case, the model has slight adaptation because it knows it's live, so it'll be
00:09:44
Andru Edwards: pithier with its response, it'll be able to be interrupted, it'll obviously have a voice
00:09:48
Andru Edwards: response by default.
00:09:49
Andru Edwards: But it's the same thing under the hood.
00:09:51
Andru Edwards: And so when you see an example of someone doing a science project,
00:09:54
Andru Edwards: and they're asking, hey, what do I do between these chemicals?
00:09:56
Andru Edwards: How do I mix this with that?
00:09:57
Andru Edwards: It's processing that information through the camera live,
00:09:59
Andru Edwards: taking the content from context of the user in,
00:10:02
Andru Edwards: looking up information on the web and within its knowledge base,
00:10:05
Andru Edwards: and then giving information back to the user all synchronously.
00:10:09
Andru Edwards: And we think, again, it's an expansionary moment with AI.
00:10:13
Andru Edwards: Now you can ask all kinds of questions to get help with things using Google
00:10:16
Andru Edwards: that you couldn't possibly foresee as possible.
00:10:19
Andru Edwards: a couple years ago.
00:10:21
Andru Edwards: There was a moment in the keynote where someone was using, I believe it was live search,
00:10:26
Andru Edwards: or using the camera,
00:10:28
Speaker 2: and she was like,
00:10:28
Andru Edwards: I forget what car it was, like, referred to as a sports car,
00:10:31
Andru Edwards: and it was a garbage truck.
00:10:32
Speaker 2: Right,
00:10:32
Andru Edwards: yeah.
Andru Edwards: Right.
00:10:33
Andru Edwards: And it corrected her.
00:10:35
Andru Edwards: Yeah.
Andru Edwards: How does live search decide or understand what it's seeing and how confident is it when it's
00:10:43
Andru Edwards: explaining things back to you based on what it's seen?
00:10:45
Andru Edwards: Well, the models are developed with multimodality pretty natively.
00:10:48
Andru Edwards: So as they're trained and developed, it sees images and understands them pretty intuitively.
00:10:54
Andru Edwards: And so I think in that example, the model just had a strong kind of prediction of what that thing was and felt confident about it, and so made a correction there.
00:11:03
Andru Edwards: And that's true generally for world information, which is being enhanced even further with the latest research from Gemini on the modeling side.
00:11:12
Andru Edwards: But I think the other part of, which is really special, is that within Google Search, you know, we have Lens.
00:11:17
Andru Edwards: And Lens has this pretty amazing ability
00:11:19
Andru Edwards: to understand objects within an image.
00:11:21
Andru Edwards: So if you were to take a photo of this space
00:11:24
Andru Edwards: and say, I really like the pattern on this floor,
00:11:28
Andru Edwards: there'd be segmentation of this experience,
00:11:32
Andru Edwards: or maybe like this carpet, it would segment it,
00:11:35
Andru Edwards: it would be able to pull out,
00:11:37
Andru Edwards: make a distinction between the carpet
00:11:38
Andru Edwards: and the floor and the wall,
00:11:40
Andru Edwards: and then would focus on that element
00:11:43
Andru Edwards: knowing that they can be segmented out.
00:11:46
Andru Edwards: And then it can ask qualified questions.
00:11:47
Andru Edwards: about that thing using the information in the photo itself.
00:11:51
Andru Edwards: So it could find similar carpets based off of the digital representation of this
00:11:56
Speaker 2: carpet.
00:11:57
Andru Edwards: So you can do pretty powerful visual understanding now, and that's just going to be part
00:12:01
Andru Edwards: of everyday life with AI.
00:12:02
Andru Edwards: Insane.
00:12:04
Andru Edwards: You also announced personal contact in AI mode, where it can reference your past searches
00:12:09
Andru Edwards: and other Google apps starting with Gmail to help answer questions.
00:12:12
Andru Edwards: We saw Apple famously or maybe infamously crash and burn around the
00:12:17
Andru Edwards: this similar feature.
00:12:19
Andru Edwards: You guys said it's coming this summer.
00:12:21
Andru Edwards: How will this work in ad mode and how does Google make sure you get if it's right as it pertains
00:12:26
Andru Edwards: to finding the right information and also I think what a lot of people might be wondering
00:12:31
Andru Edwards: also is about privacy.
00:12:32
Andru Edwards: Yeah.
00:12:32
Andru Edwards: So first of all, the experience is really focused on helping with these class of questions
00:12:37
Andru Edwards: that are about taste, personal preferences and interest.
00:12:40
Andru Edwards: So whether it's about what movie should I watch, what shoes should I buy, fashion,
00:12:44
Andru Edwards: decision cards to purchase, things to do.
00:12:46
Andru Edwards: really occupy this kind of area around personal taste and interest.
00:12:52
Andru Edwards: It's not, and we don't believe you need to personalize every piece of information,
00:12:56
Andru Edwards: particularly facts and key pieces of information.
00:12:57
Andru Edwards: So that's the first thing.
00:12:58
Andru Edwards: And then the approach is to use the information that's most useful from your search history.
00:13:03
Andru Edwards: So whether that's queries or from past AI mode interactions.
00:13:09
Andru Edwards: And then the idea is that, you know, if you're searching for sushi restaurants all the time,
00:13:14
Andru Edwards: the model can learn pretty quickly, okay, this person really like,
00:13:16
Andru Edwards: sushi. And when you're applying it to the scope of recommendations, that's a pretty confident
00:13:24
Andru Edwards: thing that you love sushi. The model knows that we're really focused on recommendations to
00:13:28
Andru Edwards: personalize to start, and it's going to, if you ask for where to go for a date night, you might
00:13:33
Andru Edwards: see sushi at the top,
00:13:34
Speaker 2: and you'll
00:13:34
Andru Edwards: feel like that's pretty magical. Whereas like, let's say
00:13:36
Andru Edwards: you're just learning about some new kind of restaurant, and you just search in, you know,
00:13:41
Andru Edwards: you Google something like, I don't know, Georgian food.
00:13:46
Andru Edwards: you know and like you did it once
00:13:50
Andru Edwards: like the model should then realize okay like a one-off search
00:13:53
Andru Edwards: isn't mean just because technically it's about food
00:13:56
Andru Edwards: that that's like all you should ever see
00:13:57
Speaker 2: again right
00:13:58
Speaker 2: so it
00:13:59
Andru Edwards: kind of learns these nuances and learns when to apply these nuances
00:14:02
Andru Edwards: and then with Gmail that's an opt-in experience to connect Gmail to make them even more
00:14:06
Andru Edwards: enhanced because obviously if Google search had better familiarity
00:14:10
Andru Edwards: with the places you actually book and your itineraries and where you go
00:14:13
Andru Edwards: it could be even more helpful yes and for all of this it can all be
00:14:16
Andru Edwards: turned off, it's all subject to the standard personalization and web history, policies and controls
00:14:23
Andru Edwards: and settings that people have.
00:14:25
Andru Edwards: Okay.
00:14:26
Andru Edwards: Then we have shopping assistants.
00:14:27
Andru Edwards: Yes.
00:14:28
Andru Edwards: Which includes trying on clothing while shopping with just a photo, which sounds wild.
00:14:33
Andru Edwards: I mean, we saw a demo when it worked, but sounds absolutely wild.
00:14:36
Andru Edwards: Tell me more about this and how accurate it is.
00:14:38
Andru Edwards: You know, people have different skin tones, people have different body types.
00:14:42
Andru Edwards: It's just worked across the board.
00:14:43
Andru Edwards: Yeah.
00:14:44
Andru Edwards: I mean, it was designed with multiple body types.
00:14:46
Andru Edwards: in mind for this exact purpose.
00:14:50
Andru Edwards: And I think the evolution was that there were a bunch of launches
00:14:53
Andru Edwards: around models that you could pick that were like you.
00:14:54
Andru Edwards: So a lot of
Speaker 2: the
00:14:55
Andru Edwards: technology was being created to fix,
00:14:58
Andru Edwards: take a 2D or representation
00:15:01
Speaker 2: of clothing
00:15:02
Andru Edwards: and put it onto a person.
00:15:04
Andru Edwards: And then the next innovation was doing it for you specifically.
00:15:07
Andru Edwards: And really the focus is around the kind of just general look
00:15:11
Andru Edwards: of it.
00:15:12
Andru Edwards: But I think you can look for a lot more enhancements
00:15:15
Andru Edwards: to come from that team.
00:15:16
Andru Edwards: Okay.
00:15:17
Andru Edwards: A-M-O can do tasks, interpret visuals, shop, as we talked about, even make reservations.
00:15:23
Andru Edwards: Why launch this in search as opposed to, say, the Gemini app?
00:15:28
Andru Edwards: Why launch it in search
00:15:28
Speaker 2: itself?
00:15:29
Andru Edwards: Is search kind of becoming the new OS in a way?
00:15:32
Andru Edwards: Well, first of all, a lot of these are very underlying capabilities that have been breakthroughs
00:15:36
Andru Edwards: from Gemini and the Gemini modeling team.
00:15:39
Andru Edwards: So for example, we talked about Project Mariner, which was browser actions last year, Project
00:15:44
Andru Edwards: Astra, Deep Research.
00:15:46
Andru Edwards: These are all foundational.
00:15:47
Andru Edwards: And so the way we think about it is there are two different kinds of products.
00:15:49
Andru Edwards: One is highly an AI search experience.
00:15:52
Andru Edwards: The other is going to become a universal assistant.
00:15:55
Andru Edwards: But there's some shared technologies underneath.
00:15:57
Andru Edwards: I think how they're applied will be different.
00:16:01
Andru Edwards: And I think that's really how to think about it overall.
00:16:04
Andru Edwards: I think for us, for search, I mean, there's billions of people use search and rely on it.
00:16:10
Andru Edwards: And so the intent is just to make it easy for them to use AI to get the tasks done
00:16:14
Andru Edwards: that they're already coming to search for today.
00:16:17
Andru Edwards: And that's really why we built AI mode and AI overviews.
00:16:19
Andru Edwards: One of Google's original mantras was follow the user.
00:16:23
Andru Edwards: Yes.
00:16:24
Andru Edwards: How does this, what I think, and I think many people agree,
00:16:27
Andru Edwards: is a monumental shift in search align with that mission?
00:16:31
Andru Edwards: Well, if you think about what we're trying to do here,
00:16:33
Andru Edwards: it's really a reaction to what we're seeing growing on search.
00:16:37
Andru Edwards: So people want to ask questions naturally.
00:16:39
Andru Edwards: They want to be able to communicate in ways they feel really natural with them.
00:16:42
Andru Edwards: They're used to communicating and asking questions over multimodality, over voice, over photos, over videos, over text.
00:16:50
Andru Edwards: And there's an expectation that information can be fairly immediate.
00:16:55
Andru Edwards: And so I think those are the principles in what we see in the usage, in the data, and when we talk to people, we're building products.
00:17:01
Andru Edwards: I think that's why you're seeing a whole product experience in AI mode and with AI overviews as a quick kind of preview, becoming a huge part of what we do in search.
00:17:12
Andru Edwards: And so it's in that exact mentality.
00:17:14
Andru Edwards: And then figuring out, hey, later, if this works,
00:17:17
Andru Edwards: we think advertising, we think other things can be great parts of these experiences.
00:17:21
Andru Edwards: We wouldn't start there.
00:17:22
Andru Edwards: We start with the user.
00:17:23
Andru Edwards: Yep. Okay.
00:17:24
Andru Edwards: Obviously, Big Day, Google I.O., you revealed a bunch of things
00:17:28
Andru Edwards: that are under the AI mode feature set.
00:17:32
Andru Edwards: For you personally, what were you most excited for people to either
00:17:36
Andru Edwards: learn about or see as a new feature?
00:17:38
Andru Edwards: And I know I'm making you choose between all your darlings here,
00:17:41
Andru Edwards: But
Speaker 2: I think for
00:17:43
Andru Edwards: most people, it's probably just AI mode.
00:17:45
Andru Edwards: I know it's a little bit of a cop-out answer.
00:17:47
Andru Edwards: But I think for the average person, they probably didn't follow the labs news and the
00:17:50
Speaker 2: information.
00:17:50
Andru Edwards: And the fact that you can just ask any question in search, have follow-ups with it,
00:17:55
Andru Edwards: and get an AI-led experience with the most powerful models using all of Google Search's knowledge of the web.
00:18:02
Andru Edwards: It's pretty awesome.
00:18:03
Andru Edwards: And I'm just excited to see people, like I do, you know, even on social media before seeing people's reactions.
00:18:08
Andru Edwards: It's just, that's been very special just to see that you could do this.
00:18:11
Andru Edwards: Wow.
Andru Edwards: I'd say the next one is just how it integrates all of the rich information available on the
00:18:18
Speaker 2: web
00:18:19
Andru Edwards: and in Google into One View.
00:18:21
Andru Edwards: So I think it's really cool that you can search for products.
00:18:24
Andru Edwards: If you look for headphones, you'll actually see a tray of headphones and you can click on a headphone
00:18:28
Andru Edwards: and have a panel come over it right in line where you can browse all of the images, change the colors,
00:18:33
Andru Edwards: look at the prices, see where to buy it at every merchant, see reviews, and then links to go purchase it.
00:18:38
Andru Edwards: All in one AI experience.
00:18:39
Andru Edwards: That is enormously helpful.
00:18:42
Andru Edwards: And I think many people don't realize that that's available today.
00:18:46
Andru Edwards: And I'm excited for people to start discovering those kinds of things.
00:18:49
Andru Edwards: One for the future, I think.
00:18:52
Andru Edwards: I'm pretty excited about what this agentic future holds.
00:18:59
Andru Edwards: Because really being able to take actions for people and save people a ton of time
00:19:03
Andru Edwards: I think is a really unique thing that we can do.
00:19:06
Andru Edwards: Because many people go to Google with, you know, many people go to Google with,
00:19:10
Andru Edwards: they're not looking for information,
00:19:11
Andru Edwards: they're looking to get something done.
00:19:13
Speaker 2: You're looking
00:19:13
Andru Edwards: up times because you ultimately want to book the thing.
00:19:16
Andru Edwards: And so I think there's an opportunity is to be a lot more helpful for people.
00:19:20
Andru Edwards: And I think people can start to see when you combine a model that can plan and think
00:19:25
Andru Edwards: with information that it can research,
00:19:27
Andru Edwards: with then the doing at the end,
00:19:29
Andru Edwards: that's pretty incredible that
00:19:31
Speaker 2: an AI system
00:19:31
Andru Edwards: could basically do that whole journey,
00:19:34
Andru Edwards: which is something that a person would have to spend hours potentially doing just a year ago.
00:19:38
Andru Edwards: Right, right.
00:19:39
Andru Edwards: Okay.
Andru Edwards: Based on everything we just talked about, is the traditional search box irrelevant or does that go away?
00:19:46
Andru Edwards: So the search box, it's interesting.
00:19:47
Andru Edwards: People come to search for a huge variety of things.
00:19:52
Andru Edwards: And we don't see that changing.
00:19:54
Speaker 2: So like
Andru Edwards: someone just wants to type in a closing business time.
00:19:57
Andru Edwards: They want to type in a new artist.
00:19:58
Andru Edwards: Just a single name.
00:20:00
Andru Edwards: You heard a celebrity name, let's say.
00:20:01
Andru Edwards: Oh, yeah, they're a great actor.
00:20:02
Andru Edwards: You just type in the person's name.
00:20:04
Andru Edwards: In many cases, what we show in search is absolutely the best thing to show for that question.
00:20:08
Andru Edwards: You get a visualization.
00:20:10
Andru Edwards: You see photos of the person.
00:20:11
Andru Edwards: You get to see recent posts from that person.
00:20:14
Andru Edwards: You get to see news articles from them.
00:20:16
Andru Edwards: You see their official website.
00:20:17
Andru Edwards: You can look at their music.
00:20:19
Andru Edwards: I think where AI mode really shines is when people are looking for these really specific things.
00:20:25
Andru Edwards: And that's where we've seen a lot of the growth just in search.
00:20:29
Andru Edwards: Because you can do so much more for people that that allows them to ask all these crazy questions.
00:20:34
Andru Edwards: And that product is really designed for that purpose.
00:20:37
Andru Edwards: I think people are going to use them in tandem.
00:20:38
Andru Edwards: and we're going to obviously bring that into search as we can.
00:20:42
Andru Edwards: So, like, user don't have to know which of these to use.
00:20:44
Andru Edwards: They'll just put something into Google.
00:20:45
Andru Edwards: If there's helpful AI that can help you, you'll see it, show up.
00:20:49
Andru Edwards: And if there isn't, you'll get this visual page.
00:20:51
Andru Edwards: You know, I don't know if you've seen, like, if you search for movies, TV
00:20:53
Speaker 2: shows, entity, place.
00:20:54
Andru Edwards: It's like a big, rich experience.
00:20:56
Andru Edwards: Like, that's actually a great experience.
00:20:58
Andru Edwards: If I just type a celebrity's name into AI mode and you get a single sentence description,
00:21:02
Andru Edwards: that's not necessarily the best
00:21:04
Speaker 2: experience.
00:21:05
Speaker 2: Yes.
00:21:06
Andru Edwards: And so I think you realize just how much people come back.
00:21:08
Andru Edwards: to Google for.
00:21:10
Andru Edwards: And that's why we view this as more of the frontier of search for, for, for, for
00:21:14
Andru Edwards: to, to, to answer some of the hardest things, um, versus a full on replacement.
00:21:18
Andru Edwards: And that is it for this edition of Geared Up.
00:21:21
Andru Edwards: Thank you so much for listening.
00:21:22
Andru Edwards: Of course, you can catch John and I on YouTube.
00:21:25
Andru Edwards: I'm at YouTube.com slash gear live.
00:21:29
Andru Edwards: And John is at YouTube.com slash John for Lakers.
00:21:33
Andru Edwards: Feel free to head over and subscribe to our channel to stay up to date on all the latest tech.
00:21:38
Andru Edwards: Speaking of subscribing, you can subscribe to Geard Up in your favorite podcast app if you haven't done so already.
00:21:43
Andru Edwards: Just search Geared Up.
00:21:44
Andru Edwards: That's two words, not one, in Apple Podcasts, Spotify, Pocketcast, Overcast, or really, wherever you choose to listen.
00:21:52
Andru Edwards: If you like what we do, please consider leaving us a rating and review.
00:21:56
Andru Edwards: It really helps other people find the show.
00:21:58
Andru Edwards: Geared Up is a Gear Live podcast, and you can see more from us at GearLive.com.
00:22:03
Andru Edwards: Thank you so much for listening.
00:22:05
Andru Edwards: Or John Rettinger.
00:22:06
Andru Edwards: I'm Andrew Edwards, and we'll catch you in the next episode.
