YouSaid · the spoken record

Sridhar Ramaswamy

lines on the record: 80
first: 2020-06-29
most recent: 2020-06-29
sittings or episodes: 1
sources: podcast

Every line below is reproduced as it was said and linked to the record it came from. Nothing here is summarised or generated. Directory · Search · Corrections

“And the thing that people don't realize about machine learning is that it is these systems are fallible. If you tune them aggressively towards fighting bad content, you take out a lot of good content. If on the other hand, you try not to take out good content, a lot of bad content slips through. This inherent computer science, this is the precision recall trade-off. This is unavoidable. And so you need a lot of training data. You need a lot of sophistication before it gets to a point where you can actually make good decisions. Again, to the point of small decisions leading to difficult outcomes, you know, my experience with YouTube was watching Jimmy Fallon videos. That's sort of my thing. And then you realize, oh, wait, like all of humanity is on YouTube. And then you go, oh, wait, all of humanity is on YouTube.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“It becomes a very hard problem because at the scale of YouTube, what was the last stat that Susan quoted? Something like 400 hours of video uploaded per minute to YouTube, you essentially have to rely on automated means to figure out what is okay to monetize versus what is not okay to monetize.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“Why unsolvable? Explain this is putting ejectal content like child pornography and it gets monetized or almost anything. There was just a controversy around a YouTuber who was doing something to a young, you know, making a joke about a young girl, things like that, and to demonetize or monetize people has been a big debate. Why is it unsolvable or do you think it is unsolvable?”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“At one level the reason was that I had been there for very long. 15 years is a very long time to spend at any place, especially for someone like me that tries to see the positives and negatives of situations. I wanted a change. But in many, many ways, the issues that we faced with YouTube monetization in 2017, starting first with the presence of objectionable content, and then a set of Essentially unsolvable problems at the time around what content was okay to monetize was what made me decide that I just needed a clean break and I needed to start something over.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“I do think these things are hard for people to understand just how I think about the world today. Compared to how I would have thought about the world three years ago Just different. And especially if you're someone who has seen growth all along the way from when you are an underdog to when you're this massive company, I think you tend to vastly underestimate your power. I honestly think that human nature. And my answer to some of this is we need diversity of thought. We need diversity of opinion. We need a diversity of product for especially important functions. So this again was one of the reasons why I deliberately set out to create NEVA with a dramatically different set of principles. And there's reason for these principles and we'll get into that in a minute. But to me, this diversity of thinking is really critical for our success as”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“I'll never forget it. We're good people. We would never do anything wrong. It's the ability to abuse people's lives because of this personal search history, which I want to get into what you're doing at Neva, is a tremendous power over lots of different ways. And we had this long talk and he said, well, we're good people. And he kept saying, we're good people. And I said, what if good people didn't run Google? What then? Like, it was sort of like they never occurred within the company that possibly this could be abused in a way that is problematic for humanity, you know what I mean? So when you're in there, when you're making decisions, do you think you all think how powerful you are? Because one of the things I had the takeaway from was that you all don't understand how powerful you are. You don't appreciate it or something like that. I'm not really clear what was why that was nothing I would worry about.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“Right. So, this is the last question I want to ask about you being there was when you're doing this years ago, when you were, I don't know if you remember, you were probably around when they were deciding whether to take over Yahoo search, for example, if you remember those days, and then they were stopped by the government. One of the pieces I wrote was that they couldn't have 90-some percent of the search market. It would have wanted them into a number in the U.S. that was so massive. It was already massive, but this would be massive, I guess. And I wrote something like, at least Microsoft, it was kind of an offhand remark. It was meant to be obnoxious, I think. I meant it to be obnoxious, which was at least Microsoft knew they were thugs. You know what I mean? And I called them thugs. And so I got a call from, I think it was Larry. It probably was because Sergei never cared. And he said, we're not thugs. Like, what are you talking about? How could you call us that? And I said, I don't believe you even think you could even try to do this. Like, what goes into your brain that you try to do this? And he said, we're good people.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“I mean, you know, it's the leadership team. It's the think tank, the senior engineers have a lot of say. The senior product managers have a lot of say in how these decisions get made. And people try to make them very, very thoughtfully. For example, for some of the policy decisions about what ads to accept and what ads to not accept, as you can imagine, things are not at all easy. With political ads, you can say maybe one answer is that there are no political ads, but at perfectly reasonable people will say, wait, you are denying a voice for the underdog. And so these truly tend to be difficult questions that go through a lot of soul searching, certainly within Google, within my team.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“All right, when you think about right outcomes, who decides that? And you're going to get to decide this at NEVA, which we'll talk about in the next section. But when these debates happen, do competitive pressures always win within these companies, who decides this? How does it get to a decision?”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“But there are also competitive pressures. Facebook was doing this very aggressively, that plays into the calculus that goes on. But I would say the larger effect here is the thing that I mentioned earlier of you can make perfectly good decisions and still end up at a place where you can conclude that this is not the right outcome.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“So there were a lot of us who felt that search history, for example, was incredibly personal. And you and I will totally agree. It's one of the founding theses of NIVA. It's a very personal product. It's a very important product. It is a mirror onto ourselves in a way that's quite unimaginable. And so we were very particular that this search history not be used for ads targeting, say, on the display network. And so this was a big and important debate. But there are always competitive pressures. Another example of a product that we spent a lot of time discussing was remarketing ads. Should your behavior off of Google, the fact that you went to some site be used to tailor ads for you on Google? This went through a lot an incredible amount of debate.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“Well, it's important to know that in search ads, on an ads in general, there are a lot of incredibly thoughtful people that worked on these problems, many of whom are still there at Google, many of whom have left. And that is very much an element of self-restraint and sort of doing right for users that has gone into a lot of decisions. We walked away completely from porn advertising about six years ago because we were like, this is not a great business for us to be in. Similarly, when there are issues around rehab ads, I was the one that pulled the trigger and said, you know, we should not accept any rehab ads until we know who the good players are and who the scamsters are in the business. And we also had really long and tortured and important discussions, for example, around should Google data ever be used for serving ads off of Google. Some of this has been written about, but these were all.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“So, when you talk about this sort of step into things, does it not, one of the arguments being made by a lot of people, including regulators, is that the way it has progressed, which is, you know, constant information grabbing from people or whatever or any activity, it's the business plan and it has no other way to go. This should have been anticipated. And of course, it begs the question was, she'd argue you were there. You know what I mean? You didn't, you knew what you were doing kind of thing. And I think everybody else is. It's an issue when arguing with Facebook people all the time. It's like, you know what you're doing. You know what's happening? And they have sort of reasonable explanations throughout. So talk a little bit like what it's like to be inside of a company when that's happening where you know what you're doing and you know where it's led and what do you do because that is the business”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“Perfectly reasonable concepts to start with, but then you drive a million people and a million decisions off of these, you sort of end up in a situation where our information is slashing around everywhere. You go to many reputable sites on the internet today and you're dumbfounded by the number of people that want to look at what you're doing. This whole concept of, say, like you're reading a book in your living room, which you can reasonably say no one else has a business to know that you're doing it, such a thing doesn't exist on the internet anymore. And we got to this place making a set of decisions, each of which were perfectly fine. And to me, a lot of why I wanted to start at NIVA were to start with a different set of principles that hopefully would lead to a very different outcome.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“To me, advertising and data is a classic example of this. At first, advertisers on even search ads, for example, would send back conversion information to Google. What I mean by that is you clicked on an ad for, I don't know, a standing desk or a headphone, and then you went and bought it on the site. That information was reported back to Google with the first intention of just creating better reports. Hey, wouldn't it be nice to have a report where you saw the number of clicks, the amount of money you spent on it, and how many conversions you've had? Similarly with third-party cookies, they intend very much was that if you know more about what a user did, then we would be able to serve them more useful ads.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“Yeah, a recurrent theme here, and you've made this in many of your other podcasts. Is that decisions, small decisions. That looks perfectly fine at the time you make them can still lead to outcomes that are quite a bit different from how you imagine them”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“And to serve them either on or off Google, that's the other concept that it was anywhere else. So when you think about the idea of what was happening as you moved into the ad area, one of the issues is, of course, collecting of data and the use of data, which is what goes into what's happening at Neva. How do you think that transitioned at Google? I don't want to like Lodi, a lot of people, I want to get into the idea of privacy and what you're trying to push there at Neva, but what was the mentality about privacy? I recall it was like any piece of information we'll take. And it didn't matter what it was. It was sort of a initially people all are thinking it's sort of a rapacious information morgue, but actually it was a very different conceptual ideas that you had to have it to make it better. Like it was, it sort of more morphed into something else. But the concept was the more information you have, the better, the better you could serve, the better products you could make, et cetera, et cetera.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“And the belief was that we would be creating a team and system that could sort of look at advertiser needs holistically and be able to serve them.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“So the rough thinking at the time was that, yes, search advertising was big and it was growing a lot. But there was this other world of display advertising, both on the internet and obviously advertising on TV. And with things like YouTube, there was clearly a hypothesis that more and more of our video consumption was going to go online. And so DoubleClick was really a big effort on Google's part to provide advertising off Google. And it started with websites. It started with display advertising. And there are a number of efforts that went into other things like audio ads, radio ads, print ads. So there's a lot of experimentation back then. But the idea was to be wherever advertisers were in terms of how they were spending dollars.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“And so when you started to get that, one of the biggest purchases obviously was double click. I think that was 2007, but that was somewhat later. How did the thinking evolve of what Google's business was? Because I remember in the early days also wandering around with Larry Page and he was very interested in search on television. He had a room full of televisions that he was. there you know before they got into anything else that they're of the many many things they were doing”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“Me, that was the brilliance of Google, that was page rank, that was really using the entire power of the web to decide what was relevant for a particular query. It was a revolutionary idea then, I think. It's a revolutionary idea now that quality is sort of determined by this kind of popularity. And as back then also had a lot of focus on the user. We made sure that the formats were unobstrusive and that they were very relevant to the search queries. So early Google was very simple. Early Google was very user-focused. And clearly it paid off in a big way.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“And then add words after that. So when you think about early Google, the concept was really just search. It was not necessarily a business attached to it. It was the idea to search the whole web, essentially. And it had been dominated at the time by Yahoo, really, and other companies that were not really doing it in an algorithmic way. They had actual searchers. They had people that would place people in a direct companies and sites and things like that in a directory. So it was a very different conceptual idea. It was to actually find things.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“It might have been applied semantics. I could be getting the name wrong. That's right. That's right. That was what launched AdSense again in 2003.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“That's right. So when I joined, I think the ads team had less than 50 people. The backend infrastructure team and quality teams, which are just beginning to get going, were about 20 people. So it was a pretty tiny team. People don't remember that Google went through a lot of changes, even when it comes to advertising. They tried essentially the equivalent of selling search ads like display ads for a very long time. And then this model adverts of pay-per-click auction was just beginning to get going in earnest when I joined. And this brilliant combination of both advertisers deciding how much they wanted to pay and Google then looking at ads and saying what's the likely value that these are going to give for users, this combination of revenue and quality was the magic sauce that really propelled Google Ads revenue. And it was”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“All right, let me talk a little bit about the early days of Google. Can you give a sense of people? I remember many of your headquarters. There's one, oh, I forget the addresses, but they were small, different places. And I went to all kinds of things. Talk a little bit about the early days of search. I think it's really pertinent to what you're talking about here, what Google was thinking about. Talk about the way it was. I mean, I remember it being crazy. I remember it being fast moving, but at the same time, it was moving into a territory that wasn't, people didn't think would make a lot of money and things changed, obviously, when ad search happened. So talk a little bit about that. When they bought several companies, but it's sort of a touch and go situation at the very beginning, despite the growth.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“more and more amazing opportunities along the way got on Larry's staff in 2013 running ads together with Susan Wajiski and then was running it by myself in 2014 when Susan went over to run YouTube spent five more years there before deciding that I needed to reinvent myself that's when I left and in some months after that started Neva”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“The name of startup was Epiphany. Yeah, yeah, I did customer relationship management software. It was a hard drive and it was very clear that growing that business was going to be hard. I had some friends who were at Google and even then it was clear that this was going to be an amazing company. And if you look back, Google made $1.5 billion in revenue in the fifth year of its existence. That was remarkable. So, anyway, I joined the Search Ads team and there was crazy growth. So I got these incredible opportunities in like four or five years. I was leading Search Ads, which made most of the revenue for Google.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“was the research arm of what was originally AT&T. When I joined, it had split off and become loosened. This was the fabled place where the transistor was invented, where Unix came to be. A lot of things that we take for granted in tech today happened in Well Labs. I was there for two years, moved to the Valley in 1999, was part of a startup that went through a true.com boom and bust, joined Google in 03 as an engineer. Funny story. I was sent to the ads team because my first boss, this guy called Mike Frumkin, found the word database in Metresume and said, aha, ads have databases. Sridhar should go work in ads. It was an amazing opportunity.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“Cool, I consider myself a computer scientist and software engineer first. I have an academic background, got a PhD from Brown out on the East Coast, spent five years in research labs in New Jersey, including the fabled Bell Labs.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT
“So, we've known each other a very long time. We've talked many, many times over the years about Google and everything else. So what I think would be great is if people understand your background, because you're sort of like one of the techies of Silicon Valley. But give people an idea of how you got to Google, and then we'll get into Neva and what you've been doing there.”
2020-06-29 · Decoder with Nilay Patel · Recode Decode: Sridhar Ramaswamy · IDENTIFIED FROM THE TRANSCRIPT