Trust is the most important thing. It's the bedrock of everything we do. There's... There... You know, search is nothing without trust.
Nick Fox:But that's what's become just a core part of Google's DNA is embrace any technology shift that comes and build on it because those are the things that are gonna be the fuel for the next wave of growth. It's less like, is the idea good or not? It's more, is now the right time for the idea? The ideas are good, but sometimes, you know, people aren't ready for it or the technology is not ready for it. That's a core part of what we all need to do as as technologists is is deeply understand that.
Nick Fox:And then once the circumstances are right, you know, you try again.
Christoph Magnussen:We are back at Google for an episode that actually spans like twenty three years of technology. Nick Fox, he's the SVP for Knowledge & Information. He spans products like ads, but also Google Maps for example, anything that's related to commerce and more importantly search, which is one of the products from Google and which has changed the biggest in the past twenty five years now at the moment with AI. But we actually took a journey back to the earlier days as well when it comes to product development, new technologies, stuff like AJAX that I quite frequently talk about because there was a big change in day and what we can learn for the AI journey today. So this is truly an AI to the DNA episode that I've been looking forward to and you can be very excited about the conversation with Nick.
Christoph Magnussen:So we're here at the Google headquarters in Berlin actually and second time we meet. Nick, welcome to the podcast.
Nick Fox:Thank you for having me. It's good to see you again. I think it's been about a year since...
Christoph Magnussen:Roughly a year exactly. And there is something special about you because you're twenty three years with Google. We just chatted about it. Like, I was at university in 2003. You started at Google, and there are not very many people who span such a, like, vast amount of time.
Christoph Magnussen:At the moment, you are SVP Knowledge & Information. And the funny part is in Germany, you would say K&I. We say K&I as well. You say K&I, and there are some Germans mixing up AI with KI because the Germans say KA, so there's a lot of confusion about it. But it's...
Christoph Magnussen:It roughly, as I get it right, it spans the ads. It spans commerce. Yep. That's... And search
Nick Fox:and our and our geo products. So Maps, Waze.
Christoph Magnussen:Maps, for example. So that's huge already. And... Many of the OG...
Nick Fox:We would say OG, original Google Alright.
Christoph Magnussen:Alright. I guess.
Nick Fox:Of them.
Christoph Magnussen:Yeah. There you Okay. And one of the biggest changes, obviously, is the search product at the moment and that was... I was curious about because you can tell from the early days, I think, when you started on ads, for example, which I still say is one of the greatest invention when it comes to business models of all time. That's my personal opinion.
Christoph Magnussen:It's not a sales opinion. It's my opinion. It's a strong business model. But you can say how that changed, and now you face this current situation where the world is pretty much new Mhmm. About that.
Christoph Magnussen:And I think that's when we... Last time when we met.
Nick Fox:Yeah.
Christoph Magnussen:So your very easy take on it. How will the Google search look in a couple of years from now? Do you have a feeling already for that?
Nick Fox:You know, it's always hard to predict the future. And so I'll I'll start by saying that. But, you know, our vision with search is that you should be able to ask whatever question is on your mind. And that's been our... I would say that's been our vision for really the history of search, and it remains our vision.
Nick Fox:And one of the things that's so exciting about working on a product like search is you never feel like it's done. Right? It's it's... So the way I think about it is, you know, you or I probably have tens, hundreds, maybe thousands of questions that come into our head on any given day. I certainly do.
Nick Fox:I'm a very curious person, but I think a lot of people do as well. And our vision for search is you don't you don't need to think about it. Just if if a question pops into your head, you can ask that question. Some of those questions will be serious questions. They'll be long term projects.
Nick Fox:It could be, hey, I'm thinking about going to university. Which one should I go to? What do... You know, you can imagine quite significant meaningful questions that someone might might want to ask. Some questions are, you know, quite simple.
Nick Fox:Right? You know, is it gonna rain tomorrow? But across that range, search should be able to answer any of those questions. And, you know, what what... So what's exciting is search isn't there yet.
Nick Fox:And and and and that is the opportunity for search over time. You should be able to ask a question by pointing your camera at it. You should be able to ask your question by asking it. Search should know what types of things you're interested in in a personalized kind of way, and it should be able to sort of anticipate. So that's where search is going.
Nick Fox:But it's very exciting for us because I think we're still, you know, you could say that search is a product that's been around for twenty, twenty five years, I think we're still just scratching the surface of what Search can do. That's what makes it so exciting.
Christoph Magnussen:I can still remember a time when people... Especially in Germany, where people said like, oh, you cannot do that because of data privacy and so on until they understand there was a difference between like, data privacy for for example, and a strategy you do in terms of data security. But I remember a time when they weren't, like, allowed in companies to Google in the early days and you were selling these boxes. We were also chatting about them in in the early days and people can remember that. And that's when I realized how important it is.
Christoph Magnussen:It becomes a habit. What did you see when this habit changed with the ChatGPT launch, for example, and then you presented the search. And I'm very aware how difficult it is to change a product that has a legacy like search and even if you still have the insights of people. And now this behavior changed. I give you my personal behavior.
Christoph Magnussen:I know when, for example, I go to the Google search and use the AI button. I do that on purpose. Yeah. And I'm always very curious to look over shoulders and see, like, what do people do?
Christoph Magnussen:How do they behave? How do they behave? What is the difference in behavior at the moment? Do So you
Nick Fox:I, you know, I'd say one of the... I think this isn't... This is not necessarily well understood. One of the toughest challenges for us is not necessarily know... Is not so much a question of what should the product be, but rather how do you get the product from where it is to where you know you want it to go.
Nick Fox:Mhmm. So if you can... You know, we've been going through a transition with AI. Right? The the, you know, over the course of the past two, three years, we've introduced AI Overviews, we've introduced AI Mode.
Nick Fox:The product is fundamentally different from what it was just two or three years ago. But it wasn't an overnight change. And the reason it wasn't an overnight change is we have billions of users across the world using search in in one way. And that was generally you ask searches with keywords. You ask a small number of words.
Nick Fox:You don't tend to do follow ups. That was that was sort of the old way of searching. The current way of searching, the new way of searching is much longer searches. Many many more words in the search. Follow-up questions could be two, three, five, 10, you know, questions in a...
Nick Fox:In in more of a session, more of a task. And and and so what was... What's been most interesting for me to see is there's been this there's been this learning effect. It wasn't clear to me that people actually would change how they search. Right?
Nick Fox:It's because to your point, a lot of these habits are ingrained. The way of searching, all of that is ingrained. And what's been fascinating and really encouraging to see is as we've launched these new experiences, actually behavior has changed in a in a really dramatic in a in a really dramatic way.
Christoph Magnussen:When you take that, I mean, to me in the early days, it was always very clear. Google is very good in organizing information since this is the mission, the core mission. But how do you see it different when you say, like, from the older days organizing information? You you have... Let's make it easy.
Christoph Magnussen:You have the the list of blue links in the very early days. That's organizing the information. But now AI gives out the information, also has maybe sometimes an opinion. And that's a very, I would say, difficult path to, like, use and design the product. That's a big challenge.
Christoph Magnussen:Like, what do you do in order to get there? Like, what could you share from from an internal process since this is strategic product development?
Nick Fox:I would say over... There's been a number of shifts in in Google search over the years. Right? You you talk about the 10 blue, you know, 10 blue links was probably the first two, three, four years of of of search. Even by the very early two thousands, we were showing, you know, we were showing we were showing shopping results.
Nick Fox:We would show news results. We would show image results. And so, you know, one thing I would just say is search in general is a story of constant constant iteration and innovation. And it really comes back to what I was saying before. Right?
Nick Fox:We want people to be able to ask whatever's question on... Whatever questions on their mind. If they're asking a weather question, we should we should we should show weather results. We should actually show what the weather is. If you're asking what's four plus four, we should show eight instead of, you know, I don't even know what what the what the sort of blue link web result for for for the weather would be.
Nick Fox:So I'd say in general, there's been this this constant evolution towards more deeply understanding what a user is looking for, and then providing the right response and the right format for that. The Knowledge Graph was actually a meaningful step along that way. If you asked, you know, how tall is the Eiffel Tower? Or or or a question like this, we'll tell you the, you know, we'll tell you the height of the Eiffel Tower and actually show that. And so I would say search search search's evolution has been one of...
Nick Fox:Let's understand what the user is asking for, and let's increasingly provide the response in in in the format and the style of the user is looking for. You you asked about sort of opinions. I don't generally think of of of searches providing opinions quite in that way. Even our AI responses don't quite don't quite provide an opinion. If you ask, you know, what's the best pair of sneakers for me?
Nick Fox:We'll we'll provide a response that way. Is that an opinion? Is that, you know, is is that more, you know, that that that that's a that's a question. But, you know, for example, we... If if if you come to search and you ask who should I vote for?
Nick Fox:We are not gonna vote
Christoph Magnussen:That's that's what I that's what I meant by I I I know technically how difficult that path is to not do that because people have this trust and say... That's why, for example, I open Google search instead of anything else. They're like, I know when I go there. And
Nick Fox:It's super it's super important that users... So you mentioned trust. Trust is trust is the most important thing. It's the bedrock of everything we do. There's there, you know, search is nothing without trust.
Nick Fox:And so, actually everything we everything we do starts starts from that foundation of of of trust. It's it's kinda why I called out opinions in this way. If if if users are concerned that we have that we have an opinion, that might that might erode trust. Right? And so we we're we're super careful about those types of things.
Nick Fox:Again, you know, what's a, you know, some of the things that Google can answer now are, know, my wife and I have a couple of hours to kill what what restaurant should we go to right now. In that case, you know, some good recommendations for a restaurant. But that's... I would say, again, that's a little bit different from an opinion.
Christoph Magnussen:Of course. It's also more direction you wanna have and and to build in. And how would you say, like, what what do we at the moment get wrong about, like, that's more my my curiosity about product development. And when I think about future products and what to, like, what to look at, I often get into the wrong direction and sometimes I'm super right And then I say something that is quite far ahead and people are like, cannot imagine. Later on, said, I told you.
Christoph Magnussen:There is there something happening. But sometimes I'm so super wrong next to that. And I would be curious to to learn a little bit more about that process behind the scenes when it comes to how people use AI because Google was very early, is still very early also in the research field. And bringing that research to a product is a totally different story. And on the way to that, people need to learn a lot.
Christoph Magnussen:Is AI strongly used internally for that learning journey, for the product journey, and how do you leverage that? So what could we learn from that process that everyone could do or you think everyone should do?
Nick Fox:Yeah. So maybe I'll just talk a little bit about what our what our development process looks like, because I think you're exactly right. You know, people people like to make guesses, that people like to prognosticate about the future. We kind of do it. And you know, it's important as part of the product development process, but you're gonna be right as right off as often as you're wrong, or maybe you're probably, we're all probably wrong more often than we're right.
Nick Fox:Google takes a very experimental product development process. So if you look at how we've brought AI into search, and this is quite interesting to talk through, because I think it's an interesting development approach. The very early days of AI Overviews, it was a Labs experience. And so you had to opt into that experience. This was a Labs experience actually in in The US initially.
Nick Fox:You need... As a user, you need to opt into the experience. Once you opted in, would get a AI Overviews on a small portion of queries. That enabled us to understand, is this useful? Is it not useful?
Nick Fox:What types of queries is it useful on? What types of queries is it not useful on? How do we tune for information quality? How do we tune for accuracy? How do we tune for latency?
Nick Fox:All these types of things. So there's a lot of development of the product in a... In this in this kind of small sandbox. There was this opt in Labs experience. Okay.
Nick Fox:At that point we got, we had developed the product to become better. By the way, if you remember that those kind of early days, these models hallucinated a bit. They were slow, kind of all of that. So a lot of that we developed in this Labs experience. We got to the point where we said, okay, this is working well.
Nick Fox:It wasn't perfect, you know, by any stretch, but it was working well. And we had, I would say it was enough to have conviction that this was the right direction to go. We then said, okay, let's start enabling this for a broader set of users. And so we launched it more fully. We initially launched it to a fraction of users.
Nick Fox:So we have we have an experimental approach where it's a portion of users that get it first on a portion of queries. Right? And so, you know, sort of a percent, you know, a certain percent times a certain percent. Again, to move slowly and to build confidence because if you if you do something in the Labs environment, well, maybe you're optimizing for that certain set of users that opt in, but you're not optimizing for the for the general population. And so so that was the next step was a portion of users, portion of their queries.
Nick Fox:We then continue to sort of grow from there, larger portion of users, larger portion of queries. Each step of the way, we were building confidence, but also building the technical capability to do it. And also interestingly enough, users were learning, you know, I was talking about the the user learning effects before, users were learning through that process too. Right? So users were understanding, okay, there's this AI Overview thing on the page.
Nick Fox:What is it? What do I use it for? How do I use it? We actually started to see even in the early days that users were users were doing more queries. And so they were even in that early stage learning, hey, search can do something it couldn't do before.
Nick Fox:I I can do queries that I couldn't do before. And so again, that gave us some more confidence. We then went from there and said, okay, that that's within the context of an AI response within the search results page. What if we much more radically change what search actually is, and that's what took us to AI Mode. But then we took a similar process of starting the Labs environment, then to a fraction of queries, and then a sort of fraction of users, and then and then and then grew it from there.
Nick Fox:I think the the key thing from a sort of how do we make it work point of view was we were expanding as the technology capability was expanding, and as our user readiness was expanding as well. And and that's that's what brought us to where we are today.
Christoph Magnussen:If you, let's say, like a German Mittelstand company knowing, okay, we cannot do exactly the same, but there are many processes, I think, that are very familiar to you after all these years and you say, well, that's very normal. That's how we do it. You mentioned Labs, for example. You mentioned how these first ideas start. How much insight can you give in, hey, you know what?
Christoph Magnussen:We also stumble upon products instead of we know they're coming. Like, what is the the the fraction of these both sides? You know what I mean? Like, sometimes... I mean, I'm invested.
Christoph Magnussen:I mentioned it already in the podcast couple of times, long time before ChatGPT and AI. Very early. Talked on the YouTube channel about it. No way on earth would I have thought that a chatbot would be a thing. Like, no way on earth.
Christoph Magnussen:And we all now know the story what what came around the corner. And we all know it's only a fraction of the truth when it comes to AI. So there are certain areas we just stumble upon, and then there are things we strategically work on. Like, how would you say in... If you could put numbers on it and what is more important?
Nick Fox:You know, Google's famous about having talked... Talking about 80/20, sort of 80, 80% or or actually, what was it? Was seventy twenty ten was was was a framework that that we that we've talked about in the past.
Christoph Magnussen:Seventy twenty ten.
Nick Fox:Yeah. So 70% is sort of the the the stuff you have, you know, quite a bit of confidence in. 20, 20% is kind of further out there. 10% is is is kind of the extreme. I wouldn't I wouldn't
Christoph Magnussen:For like when it comes to ideas or is like things you try?
Nick Fox:I would... I think a concept around that is amount of effort you put in.
Christoph Magnussen:Okay.
Nick Fox:You know, sort of a division of effort. I haven't looked at it recently to see whether Yeah. But I
Christoph Magnussen:heard I heard the number before you
Nick Fox:Actually actually do that, and it's it's probably shifted over time. I would say I would say these days, we... There's there's a set of, you know, there's a set of work which is sort of iterating on the core. There's a set of work which is you have strong conviction. You you know, as over the course of the past two to three years, we've had very strong conviction that AI in search is the future of search.
Nick Fox:And and so a lot of our effort, it's hard to quantify it exactly. A lot of our effort is there. But it's not all of it. Right? There's still work on core, you know, core ranking, core spam protection, core, you know, core information quality in the blue links, in in the in the knowledge panels we've talked about.
Nick Fox:So it's it's not sort of like all the work is AI. It's also the core. And then I would say there's there's some fraction of work which is kind of exploring the frontier. And you know, you don't wanna put... You for sure don't you don't wanna put kind of all your eggs in that basket.
Nick Fox:Right? That you you need to, you know, cabin that to to some extent. And really what you wanna do is you wanna do enough work to figure out is there something there or not? Or in a lot of cases, it's less like, is the idea good or not? It's more, is now the right time for the idea?
Nick Fox:You mentioned chatbots, right? You know, we've been working on, you know, we had the Google Assistant in the late, maybe sort of 2017, 2018 kind of time frame, maybe even before that. The technology wasn't right for it. The the vision was there. The idea of the product was there.
Nick Fox:The technology wasn't there. And so
Christoph Magnussen:remember collaborative products in very early days, Google Wave, I think it was called. Yeah. And I saw that and I'm like, okay, what is going on? This is so cool. But it was super early days.
Christoph Magnussen:When was that? 2006 or something.
Nick Fox:Yeah. And and so it's it's, you know, it's it's funny. You're you're in tech for long enough. Right? And you and you see you see many of the same ideas come around because the ideas have, you know, the best ideas have some core insight in human behavior or human-computer interaction behavior.
Nick Fox:The ideas are good, but sometimes, you know, people aren't ready for it or the technology is not ready for it. And so, you know, a big part of the question is, what is it for an idea that failed? Was it the idea was a bad idea? Was it the idea was a good idea, but the technology wasn't ready? Was the idea a good idea, but the people, you know, the world at large wasn't ready for it?
Nick Fox:It's often hard to disambiguate between those things. But I think that that's that's a core part of what we all need to do as as technologists is is deeply understand that. And then once the circumstances are right, you you know, you try again.
Christoph Magnussen:How did internal training and learning... Or maybe also we we can... Like, I wanna I wanna talk about the topic of learning around AI no matter if it's internal learning or even, let's say, with kids. How how is that changing? How is this behavior changing?
Christoph Magnussen:I mean, like, I I see a YouTube logo standing there in the corner and this is one of my main sources that I use quite a lot for podcasts or anything. Yeah. And then I dig deeper into that, for example.
Nick Fox:Yeah.
Christoph Magnussen:I see my kids do usually much broader approach jumping around, and then sometimes they stick to one channel, which is not good because they... Their opinions are not not completely grown up yet in terms of, okay, like, what is a good information, whatnot. How do you think this needs to change in the future where it becomes more difficult technically to understand what happens? Because some people are like, I don't get what happens, so I take the first thing I get.
Nick Fox:Yeah. I think it's a it's a it's a great question. It's one we we wrestle with quite quite a bit. There's off... There's often an assumption that there's one right answer to a question, or there's one right link, or one right video, or one right podcast.
Nick Fox:And it generally isn't... That's generally not the case. Right? You generally want, you know, for any number of things. It could be that you're it could be that you're learning about, you know, some issue and you wanna see multiple sides of the issue.
Nick Fox:If there's something in the news. You know, one of the one of the things that's always been a principle for us, and has and has for sure been a principle as we've brought AI into search is, let's say you do a news a news question. We should have links within within the response. Why do we want links within the response? Well, part of it is users wanna go deeper and understand something more deeply.
Nick Fox:Part of it is we wanna make sure we're supporting the ecosystem with clicks to the ecosystem. Part of it though is there's multiple... There's there's often different context you get from different sources. Maybe each source has a certain perspective that they that they want want to highlight. And so, you know, one of the huge values of search is it can it can surface a number of points of view on a topic, and it really should it should it should surface number of points of view so that so that the so the searcher can be fully informed, and then they can make sort of their own decisions about, you know, about the thing now.
Nick Fox:That can apply to a news topic. That can apply to a policy topic. That could apply to something you're looking to buy. If you're looking to buy a pair of sneakers, there's not just one pair of sneakers for you. There's, you know, maybe you want white ones.
Nick Fox:Maybe you want black ones. Maybe you want, you know, one brand. Maybe you want a different brand. And so it's... Again, it's it's it's a it's a core principle within search that we should show, you know, multiple options.
Nick Fox:And and that applies really across the range of things that people search for.
Christoph Magnussen:What do you what would you say, like, if you if you look at AI, kids, and learning, what are you more leaning towards? Some people say, like, okay, go all in. There are some... Even some schools around that that have been built, like leveraging AI for just, like, studying. And then there are some people say, like, they they need a long time to stay away from it.
Christoph Magnussen:Like, how would you see that? What would be your idea here?
Nick Fox:I'm incredibly optimistic about the opportunity to improve learning with AI. And it's it's clear to me that there's an enormous opportunity there. There are for sure risks, and the and the risks need to be worked through. And and that and that relates to, you know, it's it's a balance of that opportunity versus the risks that I think relates to how quickly should school should schools be adopting. But let me just paint the opportunity for a sec.
Christoph Magnussen:Mhmm.
Nick Fox:The opportunity of of AI with learning is for highly highly personalized learning. Every student, and I mean personalized in two different ways. One is every student's at a different level in their understanding of a topic. You know, in a certain physics course, you know, one person might understand something very very well. Another person might under...
Nick Fox:Might be struggling with that concept. And then on some other concept, they're they're flipped. Right? And so it makes sense for the person that's strong in one subject and and and weak in a in a different concept to focus on what they're weak on. For the other person that's, you know, where that's flipped, you know, that they should they should flip where they're focused.
Nick Fox:AI can do that. Right? It it can, you know, the the... I would say the biggest opportunity is a user can... A student can upload their their course syllabus.
Nick Fox:They can upload their textbook. They can upload their class notes. They can upload their, you know, you know, videos from the class. They can upload a study guide for the test. They can, you know, all all of this.
Nick Fox:And then they can say, hey, help me study for this test. And it will then focus on where are you strong? Where where are you weaker? And then, you know, really deepen the understanding where where where the user is weaker, and and kind of reinforce where it's strong. That's one flavor of personalized learning.
Nick Fox:The other flavor of personalized learning is, you know, different people are excited about different things. Right? We... You know, there was there was... You know, we're spending a lot of time talking to educators, and some of the stories that we hear from educators that really get us get us excited about this is, there was one educator that was saying, you know, I was I was trying to explain some math concepts to to a set of students.
Nick Fox:It wasn't sinking in. And then some of those students were very very motivated by sports. And so the AI was able to to translate the math concepts to examples related to sports, and suddenly everyone everyone lit up, everyone's hand went up in the class. And and because it was able to translate set of concepts they were struggling with to a topic area that they're actually very excited about, and and enabled it to sink in and and resonate in in a much much much much deeper way. So it's very clear that...
Nick Fox:It's very clear to me at least that AI will enable this personalized learning experience that I think will for sure foster much much deeper learning, much much much deeper understanding. The risks we need to be worried about are the risks of cheating, the risks of of students taking shortcuts, the risks of, you know, instead of a student writing a paper themselves, the AI writes the paper. And, you know, we need to manage our way... You know, we need to manage that. But I would say dynamic where we're overly focused on the risks versus the the opportunities is is missing missing out on that opportunity.
Christoph Magnussen:How much of that idea of also super individualized learning do you also transfer internally? Because we talked about last week, I think, we were... We had a session for for our companies. Learning is at the very core. Like, you need to train, you need to stay ahead of these things.
Christoph Magnussen:And became quite a challenge for some people saying, like, okay, this is quite intense having every week, so sometimes every day a new concept to understand and to, like, to find something where to put it. I mean, people can, like, keep seven things in their mind at the same time and then, like, you're you're, like, bombarded with all that stuff. How much do you use internally of that concept? And I would be more curious, how did that change from the early days? Because this is something I know Google shares quite a lot from from internal processes.
Christoph Magnussen:I have another topic I I wanna stress with you later when it comes to, like, knowledge management internally, but learning really interests me when it comes to the speed of learning that is necessary at the moment. Yeah.
Nick Fox:I think everyone's struggling with that. You know, to be honest, it's, you know, new models out every week, new products out every week, multiple times a week. Right? It's it's it's it's it's kinda dizzying at the moment. Think I think a lot of people are struggling with it.
Nick Fox:The approach that we've taken as it relates to this, particularly for AI internally, is is very consciously carving out space for people to learn. The biggest challenge we have is everyone's running really really fast. And do our do our people have the the time and the space in their in their busy schedule to actually actually learn? And so we've created, you know, we've done sort of, what do we call them? Sort of like AI learning weeks, AI learning days to to actually to carve out the space for people to...
Nick Fox:And then it's, I would say, it's similar actually to a lot of the concepts I was talking about for students more broadly. A lot of that is learning by doing. Right? It's it's hey, what's an what's an idea that you wanna try to implement? You can do this now with with with our with our latest models, you know, for us it's with Antigravity internally.
Nick Fox:And and actually, you know, give that a try. Mhmm. And and so that that's been that's been a really big part of it. A big part of it has also been sort of sharing good ideas across. And so it's less kind of, I would say it's it's it's not really course material as much as it's, here's this thing that's working really well for me.
Nick Fox:Let me share it with let me share it with others. You know, they in turn will share, and then it's then it's really trying it out and and sort of giving your giving your spell... Giving yourself the space to to explore. To to your point about how it's different versus before, I think that just the need for it's gone up. Right?
Nick Fox:And and and and so, you know, it's it's not just learning new technologies. It's also learning new workflows for how to use those technologies. And so all of that is a... It's a much it's a much bigger step change much more quickly. And so it's it's it's that much more important.
Christoph Magnussen:I mean, 2003 was three years pre YouTube. Two and a half years pre YouTube. Gmail was out there.
Nick Fox:Yeah. Yeah. Gmail, I think, was 2004, I wanna say. 04/01/2004, I think.
Christoph Magnussen:Must have been something like that. Because, like, in the early days of university, I used a version of Outlook for the Mac. I can remember that.
Nick Fox:Before you had one gig of storage? Yeah.
Christoph Magnussen:Exactly. And then I'm like, okay. Gmail, fastest program. And and, of course, like, in that time frame, you had technology adopting and changing fast, and I always use one example that when...
Nick Fox:But it feels faster now. Right? Maybe not
Christoph Magnussen:much faster now. Okay. And and I always try to get my head around because I'm quite good in adapting and learning fast. I was like, back in the day, had 2003, I had a Salesforce account. There were two years on the market.
Christoph Magnussen:And what I remember back then how the product changed when AJAX came out. So the the the the framework that enabled it that you could click a button—
Nick Fox:You're dating both of us.
Christoph Magnussen:And the thing is, your products, like, if you take Maps, for example, it couldn't ex... I I couldn't imagine a world where Maps exist without AJAX, like Yeah. This interaction. And, like, for the people saying, like, what is that? Like, in the end, you you don't click this blue link anymore and the whole page loaders, but a part of the page.
Christoph Magnussen:YouTube wouldn't be possible without that. Mhmm. Not at all. So Web 2.0 was a thing that was enabled by this technology. If you take it now with AI and agents and you give more autonomy, so now it's search search with AI, but many of the tasks that I see coming that I use every day are longer running tasks.
Christoph Magnussen:Yeah. Like, right before you came, I closed my laptop. I'm like, okay. I I switch off the agents for now, but they were working, like, three hours on the way here. Yeah.
Christoph Magnussen:And some of them were not finished and it feels like already the older days when it took longer to download something. What do you see internally that you can share where you say, okay, this is accelerating even more because Google's job is to to, like, organize all that stuff and all this information coming up. And to me, it's like a pile of knowledge and a pile of information, sometimes useless stuff that I produce Mhmm. Sometimes very useful stuff that I produce. Yeah.
Christoph Magnussen:And even within one company to, like, bring it back in. It's so hard. Very, very hard. Yeah. That's a tsunami of stuff coming.
Christoph Magnussen:Yeah. Do you guys have an idea of what to do?
Nick Fox:Well, first of all, actually wanna go back to what you said before because I love that you brought up AJAX by the way.
Christoph Magnussen:I I remember that day I called a friend of mine, I'm like, do know what you can do now? You can play a video in the browser. And he said like, what should that be useful for?
Nick Fox:What's... What was the reason I wanted to go back to that is, you know, I often get the question kind of were were you guys, you know, afraid by LLMs? Where... You know, how did you think about that that technology shift? And I often actually go back to AJAX Mhmm.
Nick Fox:And say, if you look over the course of time of of Google, it's been these technology shifts that have been the fuel of of of the next wave of product growth. And if you just look across the industry, it's the same thing. A technology comes along, a technology shift happens, and it enables an entire new interaction paradigm. And and that interaction paradigm enables entirely new products. And so AJAX was, okay, let people like like, what is this?
Nick Fox:This is some technology, whatever thing. But it enabled, you know, fundamentally enabled. You don't have to refresh a whole page and you can interact with something. And that's what and that's what led to Google Maps. It's what led to YouTube.
Nick Fox:It's what led to all of these things. But that's what's become just a core part of Google's DNA is embrace any technology shift that comes, and build on it. Because those are the things that are gonna be the fuel for the next wave of growth. I think that's a great... I mean, I think that's the...
Nick Fox:If you look at, you know, something like Uber that came along because apps are available. Right? I think I think the more that people embrace that as an insight to be on the cutting edge of adopting new technology, and therefore what what that can what that what that can do is is is is the right way to approach it. To your point just just now about, you know, I do think peep... I I think people are getting...
Nick Fox:I think users are getting inundated at the moment. And AI can generate a lot of content. Your your agents can can generate a lot. And so I do think that's the next, you know, what exactly is the right answer to this? We'll see.
Nick Fox:But I think that's the next the next phase of this is there's been sort of this Cambrian explosion of of of capability and output and all of these tasks running. I think people wanna get, you know, get that under control to some extent. Our approach within searches that you should be able to, you know, kick off an... You know, kick off a a restaurant booking within search, you come back to it. Right?
Nick Fox:And and so, you know, a lot of that can happen within the context of search. But exactly how that happens sort of across the industry, I think, is still something we need to work through.
Christoph Magnussen:Yeah. There's some some inspiration I took from you guys actually 2015 or something when you published an article about why Google is one big repository. I think it's it's a very long article that explains why you use the Monorepo technology, what tools you built for yourself to manage that for whoever is not technical. In the end, it's like a set of massive folders with a big difference is that all the apps and everything you provide runs from these folders. And I think there are large tech companies out there using a different approach, but I...
Christoph Magnussen:My personal opinion was was they do it as an excuse because they couldn't organize their shit. That's how I see it. Because, of course, there are advantages of both worlds. So I'm I'm a bit pointing back and white here. But using this approach helps to find stuff faster, provide stuff faster, ship stuff faster.
Christoph Magnussen:And I now see many of the new tech companies coming up using that. And I know, not for sure, but pretty sure that Uber, Meta, and a lot of these switch to that approach as well. And I found it very interesting because when agents produce stuff, they produce a lot of files that you then, like, commit as a PR to this Monorepo. But for end users, this is not a product yet. And what I see at the moment is something there that needs to bring these worlds together technological wise.
Christoph Magnussen:I mean, like in the early cloud days when people said, oh, it's just a computer of someone else, but, you know, that it's not. It's much, much more difficult than that. And you guys, I think you came up with the Kubernetes technology ten years before everyone else actually used that. So this organizing information, it feels like it becomes even more important now. How close are end users to a product that they can roll out and ship?
Nick Fox:You know, I think it's interesting for for users, our approach has been, and I think will continue to be, the user should not have to worry about organization. Right? The the... You mentioned Gmail. The insight of Gmail initially was, you know, Gmail initially didn't have folders.
Christoph Magnussen:I don't But you have to search
Nick Fox:and It was... And the idea was you can just search.
Christoph Magnussen:Yeah.
Nick Fox:And, you know you know, in Chrome, most people don't really use bookmarks. Right? Like it's it's an an... Certainly don't spend a lot of time organizing bookmarks. The initial insight of search is you don't need to know where something is, you know, just straight up Google web search.
Nick Fox:You don't need to know where something is. You you can just you can just type something in. I think assuming that users need to go through some organization step is generally incorrect. Mhmm. Because users don't wanna do that.
Nick Fox:That that's that's that's a lot of time. That's a lot of effort. That's a lot of that's a lot of resource involved in in in forcing someone to go through some organization step when the right thing is you should just dump every... You sort of dump everything out there and then and then then just use, you know, sort of search and retrieval to find it. I expect that's how how things will continue to work.
Nick Fox:If you have to kind of spend too long setting up, it's it's it's generally it's just generally not what works.
Christoph Magnussen:Yeah. And that's what it feels like in the moment. Like, you... And I, well, it's my job or part of my job to, like, head my head around and say, like, okay, what... Which way do we do it and how do we organize it?
Christoph Magnussen:And I had a very strong opinion early days also about using Google Cloud tools and then when Google Workspace came up, I said like, search is then solved. Like, you don't need folders or anything for the files and let's go with that. And of course, there are many people saying, no, but I want the folder. I want the structure.
Nick Fox:Some people love it.
Christoph Magnussen:They love it. I know.
Nick Fox:And then everyone... Everyone's different and and...
Christoph Magnussen:Yeah. Yeah. But I... Just on the way here, I had a good example. I like...
Christoph Magnussen:I I I was working on our Monorepo and one of the agents, I think it was Claude, which is usually very opinionated, said, oh, no problem. I deleted 100 gig of files. No worry. Because, like, I'm like, okay. I kind of knew these were the worktrees of the Monorepo and it...
Christoph Magnussen:It'll be fine. And I'm like, that's very opinionated. And this is exactly one of the moments that brought me to like, okay, why do I use the tools that work in the cloud and I can go back and forth? But technologically, it's so much different when so many agents push stuff because they work in such a different way that the infrastructure the infrastructure becomes more important. The small stuff becomes more important.
Christoph Magnussen:Like, I'm kind of looking. Where's the AJAX of AI that really makes this agent use as fun as clicking a button in the old days.
Nick Fox:Yeah.
Christoph Magnussen:That's what I'm looking for.
Nick Fox:Well, part of also what you're getting at is, like, what are the permissions and and and what access should an agent have. And I I think actually figuring that out is gonna be a big part. Like, do people want to connect their email to an agent or not? Yeah. Do you want your agent to be able to send email on your behalf or not?
Nick Fox:You know, we've we've for sure taken an approach which is much more opt in, much more user control around it. And we think that that level of building trust is important there. And so... Mean, I think there's a number of these pieces that need to come together to get it right. But ultimately, user permission, user control around it, and and then the the knowledge that your agent isn't gonna go do something that you feel sort of out of control over is is is gonna be an important part of this to get right.
Christoph Magnussen:So since I know you're you're tight on time today, I have one last question that really came to my mind when I knew we were meeting. How do we make sure with the time span that you have and that you know about technology, how do we make sure that we stay human in the lead
Nick Fox:Yeah.
Christoph Magnussen:Not AI in the lead? Yeah. Or do we need to... I think we... Yeah.
Christoph Magnussen:How do we make that sure?
Nick Fox:I think this is essential. This... I mean, this is this is the most important thing. This is... I mean, is...
Nick Fox:At a model level, this is why we invest so heavily in alignment, making sure that, you know, alignment when it comes down to it is making sure the model does what the user asked it to do. And, you know, alignment in some ways is a complicated word for just make sure that make sure the product works. Right? If if you go ask if you go ask your agent to, you know, go book a restaurant, books a different restaurant or doesn't book the restaurant or book, you know, books at the wrong time, you know, it hasn't it hasn't worked well. So I think that's that's the fundamental thing is we need to ensure the models do what they're saying they're gonna do.
Nick Fox:And there's also transparency about it. Right? There's transparency about what they were doing, what they were thinking, you know, along along the way. So so that can all be introspected making making sure that, you know, that's an... That's...
Nick Fox:You need to have the transparency to actually then inspect. Did did the model go down a path that that it was intended to? But I think this is incumbent on all of us that are building those of us that are building models, those those of us that are building products on top of models to ensure that it is the human in control. I think sometimes the conversation around, you know, I think sometimes the conversation around the products that people are building takes humans out of the loop a little bit. I'll I'll give you a simple example.
Nick Fox:If I'm if I'm looking to book... If if I'm looking to plan some travel, that process of, you know, there's there's... I think there's kind of a dystopian future, which is, hey, agent, go book me a trip to, you you know, hey, agent, go book me a trip. You know, in theory, there's some version of of the world where an agent can understand your preferences and understand all this kind of stuff and and go do that on your behalf. But I think as people, we like planning that out.
Nick Fox:We like saying, okay, wanna stay in this hotel and not this hotel. I want these flights and not these flights. There's there's drudgery in there too. Right? Which is do I wanna fill out the...
Nick Fox:You know, fill out every single form and put in my email address each time and put out my... Put in my name each time and my passport number each time. And I think there's there's drudgery and there's there's toil that we can that we can remove from it. But I think as we're building products, building products that put human at the center and give humans agency over how the how the product works, the decision making works is is the right path forward. So that that's the approach that we're taking, and I think that's what users want.
Christoph Magnussen:And we meet again when you guys or someone else found the AJAX of AI, and then we jump into that. Thank you so much.
Nick Fox:Thank you. It's great to be here.
