Marissa Mayer: AI to Make the Mundane Magical

0:05Hey, everyone, and welcome to Generative Now. I am Michael Magnano. I'm a partner at Lightspeed And this week on the podcast, I spoke to Marissa Mayer. Marissa is the former CEO of Yahoo and was one of the original 20 employees at Google, where she led Google Search, Gmail, Google Maps, and many other products that shaped the digital world we still know today. And now she's the founder and CEO of Sunshine, a company that's dedicated to making the mundane magical through a bunch of great AI-first products, Specifically, Shine, which is meant to make photo sharing easy and intuitive and magical.

0:42We talked about her career, AI, what she sees moving forward for AI, and of course, Sunshine, and why she and her team chose to deploy AI to make our lives easier and a little more magical. So take a listen to this conversation with Marissa Mayer. Hey, Marissa, thanks for doing this. Thank you very much for having me. Yeah, so I've been really, really excited to talk to you. Um, obviously you are somebody who has been in and around AI for a very long time. I think you studied it, uh, back in the day in your undergrad. And I think graduate degrees, obviously you worked at Google, you worked at Yahoo, you're CEO of Yahoo, and now you have a startup that's very AI focused.

1:23So you've been thinking about AI for much, much longer, uh, than any of us have. And so maybe like first and foremost, I would just love to ask you, did you see all this coming? Are you surprised by what is happening right now? I've been fascinated by AI for a long time. And there's obviously people who have longer 10 years in it than I do. What I will say is, you know, when I was at Stanford, there was this interesting major that ended up being my major called symbolic systems, which combined cognitive psychology, philosophy, linguistics, and computer science. And the idea was cognitive psychology is how people learn.

2:01Philosophy is how people reason. Linguistics is how they express themselves. And computer science, can you create something that can learn, reason, and express itself? And for years in AI, there was a lot of focus on the learning. And there was a lot of focus on the reasoning in the models, inferences, and things like that. And for me, it's an oversimplification, but I have to say that I often was somewhat dismissive of the linguistics piece of it, being able to express itself. But it was really fascinating to watch as OpenAI released ChatCPT and people could actually chat back and forth and watch the artificial intelligence actually express itself in language that sounded human.

2:45I think that was really interesting to see that, you know, you really have to work on all three of those pillars as you're building AI. But it was really the expressiveness that helped capture people's imagination and really show people the potential of what generative AI could do. Yeah, it feels like that word express, you know, this act of expression is kind of has actually been the magic of generative AI, right? You talked about ChatGPT and the chatbot expressing itself through words. Obviously, also, we've seen this through images and videos and music. It really can express itself to the point that I think most people would define as creativity.

3:22Obviously, you also mentioned learning. I think that's definitely been an aspect of it that me, somebody that hasn't been in AI that long, even years ago, felt like it could do. We think about things like the YouTube algorithm or Spotify, Discover Weekly. These things can learn our preferences. But what about the second one you mentioned, reasoning? That one still seems like these models are figuring out how to reason. But is this something in your mind that it's been able to do for a while? I think it's been able to do it for a while. Because to me, reasoning is just the abstraction of learning.

3:59And so, yes, I think that now we certainly see today that these models are reasoning quite well. But I think they've been reasoning quite well for a while. So if you look back at other examples that were in many ways some of the underpinnings of what we see happening in AI today, translation, right? Pulling in all the language from the web and understanding this is the, you know, this web page in English, this is this in Spanish, and therefore beginning to make inferences. if someone types a new wholly unseen English phrase, how would you translate that into Spanish? Things like that. Similarly with facial recognition, the fact that it can have me be tagged in one photo, but then in another photo where the angle, the lighting, the shading on my face, things like that are different, yet it still recognizes me.

4:49That's, in my view, these models really reasoning. It takes a set of learnings, very specific learnings, and abstracts them. So you can actually reason and say, well, if that was Michael, that must be Michael as well. Right. So is the difference then now between this reasoning that sounds like it's been doing for a while and kind of what's coming next is like this multi-step reasoning, like this ability to like reason and then go through like this chain of thought on multiple levels and come to like sort of complex answers and tasks? Is that kind of the difference? I think so. And I also just think that what you're seeing is a scale law playing out.

5:27So what was, you know, what looked like groundbreaking work in 2010 in translation, when you play that 15 years forward and you've got something that's rapidly evolving and scaling, it starts to look a lot more sophisticated. And it's not just doing rote translation from one language into another anymore. It's doing complicated reasoning across, you know, hundreds of thousands of web pages on a topic and putting together a synthesized answer. And so, yeah, there's very complicated chains of reasoning that are at play now. Yeah. You mentioned scaling. I mean, like one of the things that a lot of people have been talking about over definitely over the past couple of months, but definitely over the past couple of weeks is like, you know, the story of the past five years has been scaling and scaling laws constantly being proved right.

6:16Now there's a sense that we may be sort of like reaching the upper limits of the existing scaling laws, especially as it relates to maybe transformers. Do you feel like we're hitting some sort of ceiling or plateau or are we about to like blow right through this thing? I don't know that we're actually approaching a scaling law. And I think you have to be working pretty directly with these models to really observe whether or not that's what's happening. One of the things I do suspect, though, is we've gotten really good at collecting information. And so I think that a lot of what you're seeing with a lot of the LLMs in particular is that the world of words online is vast, but it is in fact finite.

6:59So once you actually crawl all that information and you ingest it, yes, there will be differences between the different engines in terms of how they process and use that information, but they'll probably be more alike than they are different. And I think that that's one of the things that you're seeing is that that consumption of new information to ingest and synthesize is in many ways, you know, we're getting to the 80-20 rule or maybe even the 90-10. And so that element of it, do we have new creative thought, new creative content for these LLMs to ingest and learn from that wasn't created by another LLM, which has its own difficulties?

7:42I think that that is, it's clear that they can consume information faster than humans can generate it in large part. And so I think that we might be seeing some element of a plateau there, which I also think will lead to some amount of commoditization across LLMs, where in many cases they may be more alike than different. Right. And when we get there, it's probably then the value is going to shift to other aspects, right, around maybe distribution or capability of, or, you know, specialization of one model versus another. Obviously, you've you've been involved with some iconic companies and products.

8:19AI, at least from my perspective, like seems to be sort of resetting the playing field a little bit. It's not only like creating a new type of product opportunity to be built. It's also creating new opportunities inside of organizations. How do we staff teams? How do we, how do we like build processes and actually get these products made? Maybe starting at like, at like the team level, like what do you, how do you think teams are sort of going to evolve as a result of AI? Like, will we value engineering more? Will we value product management more? If product managers can just generate code through natural language, like how are sort of teams and team building strategies going to change in the coming years?

9:00Well, I think that what's happening on a product level with AI is often the most powerful technologies change end user behavior and they change end user expectations. And we saw that in the early day of search, right? All the search engines were kind of the same. They could all kind of get things done. They weren't incredibly efficient. And then Google came along and had just better answers, caused people to search more on Google, caused them to search way more. There's way more internet search being done today than there was before Google existed. And, you know, it basically caused people to ultimately change their behavior in terms of how often they searched, how they searched, what they searched for, and also change their expectation in terms of what they got.

9:45And I think that what you're seeing now across many forms of generative AI is whether it's chatting to get a better understanding of a problem, get editing help, get visual help in visualizing something and creating an image or a graphic. The way that you can do that, the speed with which you can do it, and the ease with which you can do it, is going to cause users' behaviors to change and their expectation to change. So I think that ultimately, the groups that will be valued most in organizations as a result of this change will be the ones that can adapt, be most insightful about those user behavior changes and expectations, and adapt most quickly and respond.

10:31So I don't necessarily know that it will be a functional revaluing across companies. But I definitely think the companies that respond to these user behavior and expectation changes are the ones that are going to win. That makes me think of an area of business right now that to me on the surface looks potentially highly sort of disruptible. And that is consulting, right? Where you've got these huge teams of people that are being paid and billed hourly to do this type of manual work that potentially could be replicated by agents. I'm certainly very curious to see if large consulting firms sort of adapt and embrace the way you're talking about, or sort of try to hold on to existing business models and structures.

11:14Yeah, I'm curious what your thoughts are and like the types of businesses that you think will be most quick to adapt this versus the ones that will be sort of resistant. Well, not surprisingly, I think that technology-rich industries are ones that are going to adopt this most quickly and readily. So I wouldn't be surprised if we see consulting actually, you know, they generally do use technology effectively and study it. And it's like packed on HR and office and workflows, you know, as part of what they do. So I wouldn't be surprised if they actually do well in terms of adopting it. But I think that anywhere where there's a tension between the new way of doing things and the new way your customer wants to do things and attention with that and your existing business model, I think those are the companies that have really hard choices to make.

12:05Because they can probably cling to that existing business model, a la your consulting example, and make more money in the short term. But in the long term, they're really mortgaging their future. So I think that, you know, right now, most companies need to be thinking about and, you know, most people working in those companies need to be thinking about how can AI make me better and faster at my job. And there's all kinds of people who are doing studies now in terms of, you know, who does better? You know, the consultant, the doctor, the lawyer, or the AI, right? If you go to the AI, old suggestions.

12:41And there certainly are some professions in some cases where the AI is outperforming the people. And if the people attempt to actually edit what it's suggesting, things only get worse. But, you know, it's early days and there's anecdotes right now that cut both ways. Yeah. Yeah. I saw some, I think it was a New York Times article just like two days ago that said it compared sort of doctors on their own versus doctors with AI versus just AI. And I thought it was going to be doctors and AI were the most effective. And it actually, according to this study, I have no idea how accurate it is. It was just the AI that was the most effective or I don't know, however they were measuring it, which kind of blew my mind a little bit.

13:24I've seen that study and I've seen some in a few other industries as well. And so, you know, those industries where the AI solo can outperform and the value that the people are adding could be negative, which is kind of the point that that makes. And I don't necessarily think that it was a big enough sample study to say that conclusively, but sample size in the study. But I think that it definitely is thought provoking and it does make you realize that where you should be spending your time and where we can add value versus where the AI should be adding the value is something that companies are going to have to be strategic about.

14:01Yeah. I mean, you know, as somebody who's spent a lot of time building products, you know, as like sort of a traditional PM, but also a designer, I know, obviously, you've built a lot of products like I do wonder, like, what is the future of design or product management, product management, maybe specifically, if like, like, isn't AI just going to be able to write out like the perfect product spec and make like the exact right assumptions about what a user wants? And then, boom, maybe even take out the coding part of it at some point as well and just spit out the code or not even spit out the code, just generate the product in real time.

14:33But I guess we'll see what the role of the human is. You mentioned business models. Just one more thing I want to touch on before we get to sunshine. So something I've been thinking a lot about is sort of the business model of search. and sort of the fact that kind of search and advertising has been this economic engine behind content on the web and sort of the free and open web as we've experienced it over the past couple of decades. But now through AI and through these agents, we're having the AI go out and do stuff for us and in many cases consume content for us, thus obviously disrupting sort of that that economic engine and the advertising, the attention that we we humans normally give these things.

15:18What happens like play this out? Like what happens to this sort of economic engine of advertising that sort of underpins the whole Internet when kind of we outsource everything to these answer engines? Well, I think that the advertising still will have value. And I do think one of the critical signals that happens in search is that a result's ability to pay does affect its quality. The classic example is concert tickets. For example, you know, if I type in Taylor Swift concert tickets right now, there are lots of articles talking about the fact that the Ares Tour has been, you know, such a phenomenal seller and what it is and where it's been.

16:01But if I type in that, I'm probably looking to buy tickets. And there's a lot of vendors who are very economically motivated to get in front of you to ultimately achieve that sale. And I want them to get in front of me because that's what I'm looking for. So it's counterintuitive because a lot of people say, well, look, if they have an economic motivation to be in the search results and they're willing to pay, they can pay and then there's not a quality factor. But I think that Google has really shown over time that they do have a really good handle on good quality, good quality ads and formats and really using that ability to pay as a quality signal.

16:40Not exclusively, because you can't use it exclusively to understand if it's a good match to search, but really including it. And the other key thing that I think Google got right about advertising in this classic model is the search results and the ads are congruent. They're clearly marked, you know, it's sponsored, but it's easy to absorb the information because it's in the same format as the search results around it, which means if search results fundamentally change their form, they're more synthesized, they are paragraphs that, you know, summarize content, people are going to want that. It means that when I'm doing that concert ticket search, I don't want to have to go through four different, you know, ticket vendors ultimately looking for the best price and the best seat.

17:28I'm going to want that really, you know, brought together and be given, you know, the best available seats in this section or the best available seats for this price and have the same type of synthesis applied to the paid results as you would have to pay for the results. So I think that one of the things, that notion of what's native or, you know, what's twinning with the content where you've got, you know, organic content and paid content, but they live in parallel to each other. However, if the format and it's not 10 blue links, then it can't be 10 ads, each with their own link. That has to change, too.

18:08And so if we're starting to look at things that are more paragraphs, as I said, synthesized information, I think the ad has to follow that format. That's something that we know works well for advertising. So it may be quite disruptive to the type of format of what people are expecting. Super interesting. What I'm gathering from you is that the ads model won't be disrupted or is likely not to be disrupted. In fact, it'll probably be strengthened because as a result of AI, we can probably get a better sense of the person's intent. but 10 blue links is one example is really a representation of the amount of signal we have and so we're we're sort of presenting you with array of options because we don't really know exactly what you want but if we knew exactly what we what you wanted like arius tour tickets we would just give you that one blue link hypothetically i mean i even be a blue link and i just say you want these tickets right exactly exactly surveyed and these are the best ones that are available and are most reliable and most likely to actually materialize or verified resellers and things like that.

19:12And that element could come in quite strongly. So I think that just like AI answers right now, those AI summaries at the top are in fact summarizations. The question is, do you have an ad summarization that runs there as well that summarizes, yes, you were searching for a TV. In summary, we have TVs that range in this size to this size and from this price to this price and ultimately really let you, you know, play with those parameters. Right. The company that will be most successful is probably the one that can accommodate the right format. Right. Tell us about Sunshine. It's been awesome to see.

19:52I've got, I think I've got all, there are three products right now. I think I've got them all on my phone. There's birthdays, contacts, and photos. Are you focusing on all of them? Is this sort of like an app constellation strategy or is the focus really on photos now? Well, we do have a portfolio strategy. We knew it would always be that. And we started the company to take mundane tasks and make them magical and easy and apply AI where we could to everyday tasks. And today, what we're very focused on is taking photo sharing and fixing it using AI, helping apply facial recognition. So we share our photos with the right people or the people we forgot to share them with, understanding how to cluster photos.

20:38So we can say, you know, look, you and your friends, you each took the same photo a bunch of times. This is the best one. Most eyes open, best lighting, you know, the most share worthy is a term we tend to throw around a lot. But we're very focused on how do we take photo sharing, make it easier, make it better once you add, you know, once you make it AI powered. Awesome. And how do you do that? How does the AI make it better? We think that right now, sharing photos works reasonably well when you text person to person. We also think photo sharing works really well on things like Instagram, where you're trying to get your photo, one amazing photo in front of thousands of people.

21:19But we think things are pretty broken when you're dealing with small groups. say like five or more up to a few hundred people. Maybe it's an event. Maybe it's a conference. Maybe it's a soccer team. Maybe it's a party. So we have started working in some events, basically, because we know when someone's about to have an event that they're about to take a lot of photos that matter to them. And even people want to generally share them at those types of events. But we think that that type of photo sharing is really broken. So trying to understand groups and relationships among groups, the types of photos that you want to post to that group and who should be in the group.

22:01Those are all things that we're working on and understanding people's photos and helping them share them more efficiently and effectively is just a really hard and fun and interesting problem. How did you decide on this focus to like really focus in on these sort of like mundane tasks of life and sort of breathe new life into them via AI? Like, I feel like, you know, people would look at some of these things like photos or contacts and be like, oh, those categories are baked. You know, Apple, Google, they've already got them locked up in the OS. How did you and the team decide to narrow in on that as the area to focus on?

22:39I've always worked in consumer tech and I love consumer technology. And I've generally worked on consumer tech that people use every day in really small increments. search, email, maps, stock quotes, sports scores. And so what I really love when I build a product is thinking about how people use it and how can we really simplify it and really kind of grind it down so they have a simple, beautiful experience where there could be very sophisticated technology working behind it, but yet it presents itself as extremely intuitive and straightforward to the user. That's really how I like to approach things.

23:17And coming at it from another angle, so that's why we kind of focus on mundane, everyday tasks. But as you mentioned in the beginning, I've been fascinated by AI since my college days. And when I look at the landscape, I think one of the biggest problems that we can work on today isn't necessarily the building of these models. That's very complicated. There's a huge barrier to entry. I do think to some extent there will be some commoditization across it. So when you're doing a startup, there's almost nothing worse than saying, look, there's a large barrier to entry that could be in the billions of dollars.

23:58And when you're done building it, you might end up with a commodity product. That's very scary and very daunting. But at the same time, to me, I think that humanizing AI, making it useful every day for people and having them understand how it can help them in these types of everyday tasks is one of the most important things that we have to get right in the field of AI. So AI isn't necessarily this abstract thing or this thing that I kind of play with when I want to chat or make pictures, but it actually helps me every day in something that I want to do. I think that that's something that's really important.

24:35I heard that there was a physics professor wants to like to ask people three questions. They would say, what are the biggest problems in your field today? What are you working on? And the third one is, why are those different? And so for me today, I'm very excited about where AI is going, but I have real concerns that, you know, I grew up in Wisconsin. If I went to the center of the country and I do sometimes, like I go home and like ask people, you know, what do you think of AI? They're really concerned. Is it going to take their jobs? You know, are we going to lose control of it? Will it hurt us?

25:07You know, there's a lot of fear in terms of what it is. And it feels like something you can't touch or use or understand. And so I think creating applications that can use AI in a way that makes it really obvious how helpful it can be for people is something that ultimately is really empowering and really something that can move the whole field of AI forward in a positive way. How do you, you know, you talked about these services, these sort of mundane tasks, and I mentioned earlier, like oftentimes these are things that come bundled with the OS. How do you think about distribution and sort of edging out these incumbents that already put like a contacts product on your phone when you turn it on and a photos app?

25:53Like, how do you think about that that element? What we found is we started building on contacts mostly because for what we want to do, contacts is foundational. If you're trying to understand groups and relationships, how people like to share, how they like to communicate, you really need to make sure you've got the right phone number, the right email address, the right methods of communication, social media handles for a particular individual. You know, if you send someone an email or an invitation or a photo and they don't get it because it went to the wrong place. It doesn't matter. So we started off in contacts, but and it's counterintuitive because contacts are, of course, you know, in theory, everyone, you know, but contacts are inherently not that viral and not that social.

26:39Like you've never said to your friend, hey, there's this great contact manager. You've got to try it. That's true. And it is the highest rated. And, you know, it was app. of the day in the app store. And it's a great product, but it didn't have the type of social and viral growth that we really ultimately want. And so that's one of the things that made us think, okay, well, how will some of these grow, especially that they are in spaces where there are really, you know, fearsome competitors. And so we thought, you know, what really has to happen is we've got to work on spaces that are inherently social, which is why we moved to photos and also implicitly events where we feel like the event space has not been disrupted much.

27:22And, you know, there's just a lot right now of kind of friction and organizational overhead that we think that we can make easier, both with technology and with AI in particular. How much do you think about events as its own sort of discrete space being one that is just like totally locked with photos? I mean, the event, it's a really great observation you just made. Events hasn't been innovated on. I also saw like I opened the app recently and I'd recently been in Paris with my daughter and it just like knew that it knew that I was in Paris. It organized all the photos with my daughter in Paris.