← AI + a16z
AI + a16z
AI Infrastructure, Distribution, and the Next Wave of Software
00:38:41
Sophie Buonassisi speaks with Jennifer Li, General Partner at a16z, about why infrastructure is becoming one of the most important areas in AI. They discuss how the shift to AI-native systems is reshaping everything from storage and compute to developer tooling and orchestration. The conversation explores early insights from companies like ElevenLabs, why distribution has become the defining advantage in AI, and how founders can think about product, research, and go-to-market in a rapidly evolving landscape. Jennifer also shares her perspective on creative tools, the role of AI in storytelling, and what the next phase of the ecosystem may look like. This episode originally aired on the GTMnow Podcast.
Loading summary
Transcript72 lines
- [00:00]
Max
Andreessen Horowitz raised $15 billion and $1.7 billion of which was allocated towards infrastructure.
- [00:06]
Jennifer Lee
When it comes to storage, compute and all the tooling, like, we consider that a big part of infrastructure. Again, like how do we store like memory and so on. Like, those are all the opportunities that's emerging to build infrastructure.
- [00:19]
Max
Jennifer Lee is a general partner at Andreessen Horowitz.
- [00:22]
Jennifer Lee
More than 90% of the code are being written by agents.
- [00:26]
Max
What did you see early in the voice AI space?
- [00:28]
Jennifer Lee
So when we first saw the 11 Labs demo, I remember you're using like a fond of voice to just like narrate a book or like, holy shit, this is really alike. And also has all the right pauses stressing intonation. It's super engaging. I've always been a big fan of the one piece manga. It's only every couple years for eight episodes, right?
- [00:48]
Sophie Buon Assisi
Yeah.
- [00:49]
Jennifer Lee
Make these products and tools your friends.
- [00:52]
Max
The best ideas live in the graveyard.
- [00:56]
Sophie Buon Assisi
What does it take to build the infrastructure behind AI? This episode, originally aired on the GTM now podcast, features Jennifer Lee, general partner at A16Z, in conversation with Sophie Buon Assisi. They discuss the rapid evolution of AI infrastructure and the systems powering the next generation of software, from model development to developer tooling. Jennifer explains why infrastructure is becoming the most important layer of the stack and how distribution is emerging as the key differentiator in a crowded market. All right, we're back with another fun and amazing episode of the GTM now podcast, the special edition VC bonus episodes we have with myself and my general partner, Paul Irving. What's up, Paul? How you doing?
- [01:44]
Paul Irving
Doing well, Max. How you doing? I know you've been on the road quite a bit. I'm about to hit a road trip coming up. Continuing to be pretty busy and exciting times.
- [01:51]
Sophie Buon Assisi
Road warriors indeed. It feels like there's a certain seasonality to it for sure. And this just happens to be one of those extremely busy times, so wouldn't have it any other way. It's fun and exciting. And with that, you know, what better way to kick off the pod than how fun and exciting it is? Are we in a bubble or not? I know it's nuanced. Does this feel like 1998 or 2001? You know, we just had the Allbirds pivoting to new birds AI and buy and purchasing what, 50 million worth of GPUs to rent out? Like, there's stuff like that that I read and I'm just, what, you know, we're doing that again. But then there's you know, the growth of anthropic where you know, you're looking at that and you're like no, this is real, this is a platform shift that we haven't seen the likes of since, you know, maybe mobile or the Internet in general. Right. So what's your take? Or what's your, what's your view on that?
- [02:49]
Paul Irving
I think from a high level, because if you only look at the high level aspects of it rhyme so nicely, it's really easy to make that comparison. You say, okay, if AI is going to be a shift as transformational and large as the Internet when it first launched, there is going to be a huge build out of infrastructure that exists. Companies are going to grow as fast as you've ever seen them grow from an equity value perspective at least. And a lot of money is going to be thrown at it. And a lot of that stuff from a very high level will rhyme. If you look at 1998, 99, 2000 and what the last sort of three years have been post ChatGPT launch. But you have to dive into the numbers like you just mentioned to really understand that this is completely different. Are there going to be some parallels? Maybe. But the core drivers of is there value today? What are the economic factors which will influence the success of this over the coming, you know, two, three, four, five, ten years or not? So I'm just going to list off a couple of those and would love to get your reaction to, to some of them as well. But four years after the Internet's public release, there were 70 million users globally. ChatGPT and AI apps already have 1 billion monthly active users. A completely scale and in less time. 90% of this AI buildout, so you talk about data centers, GPUs, 90% of it is pre committed. When we were running fiber in the early 2000s and late 90s, it was 3% pre committed. That's totally different. I think differentiator between the two of them, you look at them is just unused capacity versus used capacity. Every time that we bring more compute online is almost a 1 to $1 creation ratio for the frontier model companies or the infrastructure companies to have that surfaced. And then you at the constraint side of things like fiber was pretty cheap in relative terms. And you know what, over the ensuing 15, 20 years we really benefited from an over build out of infrastructure. But right now the infrastructure buildout is not simple. Energy is a big constraint. Land is a big constraint. GPUs, if you know, you dive into Nvidia's earnings every single quarter or listen to Jensen when he talks like there's still capacity constraint from a supply perspective and meaningfully. And so I think the core drivers of it, you know, is there demand, how pre committed that demand is the amount of users, the amount of value being created, the revenue growth of anthropic OpenAI and even some of the smaller private market startups that we're lucky to meet and invest in is, is night and day different from what it was like in the late 90s and early 2000s.
- [05:16]
Sophie Buon Assisi
Yeah, and I, you know, I look at this from a B2B SaaS standpoint. You just saw I think news the other day that Thoma Bravo is completely writing off the Bedalia acquisition that they had made. And you know, their model was, was largely to acquire these companies that had pretty good, you know, bones economics, you name it, juice them and then flip them for 2 or 3x in a couple years. And you know, I think a famous one of these was Vista and Marketo where they got like a 3x in just a handful of years. And you know, with the kind of demise of the traditional B2B SaaS and this usurping of these kind of native AI companies, it's been pretty wild to see companies that I thought two or three years ago even or you know, in 2021 at least though worth close to 10 billion, that are now, you know, nothing, zeros. Right. So how do you kind of reconcile that in this, you know, trend or, or bubble as part of that?
- [06:15]
Paul Irving
Yeah, it's, it's a difficult thing to reconcile in the sense that I do believe both things can be true. Where you can say there is more value being created today than maybe we've ever seen in the history of technology, investing and innovation that usually on the other edge of the sword is value destruction like entropy has to exist somewhere. And I think the, the whiplash of it all is the part that's difficult for a lot of companies and executives and operators and founders and investors to manage, which is investments that you made or product decisions if you're a founder that you made 18 months ago that you would normally have a longer Runway to prove out whether this is going to be viable or not, are re underwritten on what seems to be a week by week, month by month basis and definitely a model release by model release basis. An interesting counterpoint to this though, and I was talking to an executive and we've talked about them on the podcast a couple times. The Intercom team and the release of Fin. What's really interesting about that and I think an under discussed aspect is the core Intercom product is now re accelerating. So it took this re pivoting to say, hey, we're going all in on AI. Fin is the future of this company. Its own product, it's AI native, you sell it to all your customers. But there is still a use case for some of the traditional software that they build out over a decade. And now that part of the business is re accelerating. But you know what, that never would have happened if it wasn't for Fin. And so I think you can follow the success cases, understand that they're really hard to pull off and there's going to be a lot more value destruction and creation for companies that are trying to cross the chasm of traditional B2B software to AI. But it is possible and it is cool to see the almost traditional offering at a company like Intercom start to accelerate again.
- [07:57]
Sophie Buon Assisi
Yeah, there's definitely a handful of employees in two different buckets. Ones that like were there in the early days, had some shares to exercise irons out, they're like, oh, I don't know where this thing is going. I'm going to save my money. And then the other bucket that's like, yeah, you know what, I'll pick up a flyer on this. And then, wow, the company, you know, iron comes back, the company reaccelerates in a massive way. And then, you know, you see a competitor that I think was growing incredibly fast, but probably not at the, the scale that Fin was in qualified to get acquired by Salesforce, which I heard was upwards of a billion. So obviously Intercom doing very well. Now on the other side of that, today's guest on the show is Jennifer Lee from Andreessen Horowitz, partner at Andreessen Horowitz. We've done some deals with Jennifer Love, working with her fantastic show. What were your kind of, you know, key takeaways or things you'd like to dig at from that episode?
- [08:50]
Paul Irving
We're GTM funds. So I have to start with her distribution commentary, which I thought was spot on though. And it's something, something we talk about a ton internally, which is the speed to become a default brand has never been more important if you're an AI native company. And then the gap between sort of 1 and 2 or 1 and 2 and the rest of the field just seems to widen on a day by day basis and be as large of a chasm as ever before. And I think there's some really good examples of this and they obviously and rightfully Get I think a lot of praise for the work that they've done. But like Harvey, becoming a default application, AI native application for the legal community across top 100 law firms and the rest of the ecosystem beneath that as well. It wasn't that the product could do everything you wanted it to do in the early days, but they did an incredible job establishing the brand and then the product backfilling that. From an execution perspective, Ligora's a good example of a company not far behind and growing incredibly fast. But you start to look beyond that and it is hard to see, you know, what the rest of that market could look like, you know, lovable and Replit is a good example of this in sort of vibe coding and non technical user app creation where distribution matters. And we've said that for a long time. But in this AI native era, we're calling it the distribution era for a reason. You need to get your go to market right and your distribution right from day zero. Because you know, the fastest way you become the default brand in a new category and there's a lot of new categories and products being created on a day by day basis is you need to move fast and you need to have the foundations in place to be able to, you know, build a flywheel and build an engine.
- [10:28]
Sophie Buon Assisi
Yeah. And one you do move fast, you get rewarded for it pretty quickly because there are funds like Andreessen with 15 billion under management, I think 1.7 billion just for infrastructure alone. And they're going to invest pretty ferociously. Right. So they're, I think she did the A, B and C for 11 labs. Right. And so you're, you know, you're kind of king making on your own by growing quickly and building great distribution. But then of course, you know, the mega funds can come in and really give you this mountain of money that puts a pretty good big gap between you and the rest of the pack.
- [11:03]
Paul Irving
Yeah. And I think a lot of it also comes from some of these verticals or some of these use cases like business users knowing they need to adopt AI and C level executives and boards having a mandate to adopt AI but not always knowing where to start. And you know, like any network, anyone who's had a professional job realizes and understands this is, it's just people trading and sharing notes. What do you use, what did you try out, what worked? And there's this virality, I also think in a lot of AI product demos, but, but also in people sharing use cases and success cases where once you have the momentum behind you, it's just a snowball rolling downhill and hats off. I wanted to make sure we tipped our cap to Jen for leading the A, the B and the series C for 11 labs is just what an incredible team to work with. And not just 1, 2, but 3 unbelievable investments for their team into that company.
- [11:55]
Sophie Buon Assisi
And it shows that their strategy can work. Right. If you raise 15 billion and you, if you find the winner at A, which you kind of are starting to know who the winner is, you can put a ton of money to work in that one company. But it also shows that our strategy and kind of the seed strategy works because there were rounds done in 11 labs before they got there. Right. So, you know, some smaller firm found that company and was able to get in before, you know, an. Andreessen Horowitz, ABC what do you think about that?
- [12:27]
Paul Irving
Yeah, it's the precede and seed rounds. It's just a question. We get a lot. The precede and seed rounds that capture the TechCrunch headlines and, you know, go viral on X, LinkedIn or both are, are the mega seed rounds. Those are a really small, you know, drop in the bucket in the, in the larger, I would say, investment landscape for Pre seed and seed. And you look at a company like 11 Labs and it's a great example of that. There was a $2 million round that got done at company inception and the angels and the pre seed and seed specialist funds that invest into that particular round. I mean, not only a fund making investment, but potentially a firm making investment and that all the time. You know, Pre seed and seed is not an efficient marketplace by any stretch of the imagination. There is incredibly smart founders and builders all over the world with exceptional teams and exceptional ideas, and it's hard to capture all of those at the inception stage of the company. But, you know, firms like A16Z do a exceptional job making sure that if those companies break out, they're on their radar and, you know, they're ready to concentrate capital into them as they grow.
- [13:32]
Sophie Buon Assisi
Yeah, I mean, it's, it's certainly validating obviously to the strategy of the kind of emerging manager and seed only funds, but wow, kind of what an amazing way to deploy capital to be able to kind of wait to see who the winner is going to be and then go all in behind them. It really kind of is attestation to what Andreessen Horowitz has built from a brand standpoint over the years. And I think there's probably a few firms that could pull that off. Sequoia, Lightspeed, Excel, Index, Thrive, you know there's. But it's a short list, probably sub 10, 10 firm list. Last thing on, on this one but wanted to point out founder qualities. What are some of the things that she was looking at, you know, when sizing up founders to invest in.
- [14:17]
Paul Irving
There's a couple that stood out to me, but one in particular which was the best founders understand where the puck is going from a model capability perspective and build their product roadmap in concert with that. But the very best founders go one step further which is they know where the models are going, model capabilities are going and they'll build a patchwork version of that functionality into the application ahead of time and then get it into customers hands and then when the model capabilities are there to backfill that one quarter later, two quarter later, you are way ahead of the rest of the market. And she, she cites eleven Labs as a company that's done a great job of that and it's easy to see why and how and if people are users of the product, you've experienced it firsthand. But, but that level of understanding of where the world's going, where these AI models are going and building your product roadmap around that is I think what the best founders, you know, building AI native software and AI native applications are
- [15:20]
Sophie Buon Assisi
going yeah, and she's got a great eye. Phenomenal track record. All right, so with that let's get right into it. Today we have our SVP of marketing, Sophie Buenasisi doing the episode in person with Jennifer Lee, general partner at Andreessen Horowitz. We'll let them take it from here.
- [15:37]
Max
Jennifer, welcome to GTM now.
- [15:39]
Jennifer Lee
Thank you for having me here.
- [15:41]
Max
Absolutely. It's great to have you here and have a lot we want to cover is Andreessen Horowitz raised 15 billion and 1.7 billion of which was allocated towards infrastructure. And I believe that means that infrastructure is actually tied with apps for the largest vertical bet in the race, which is huge. Why now? Why infrastructure?
- [16:01]
Jennifer Lee
Yeah, I've been longtime enterprise investor but spend most of my time in infrastructure and as I've been looking back in the last eight years in venture like the role of infrastructure just become more and more prominent for multiple reasons. One is we're still on the end of shifting everything to cloud. But along that race we have this thing called machine learning came about around 2017, 2018 and a lot of tool chain that's built up upon that both on data, both around, you know, ML tooling or coming through and 2022, of course everybody knows the ChatGPT time. But even before then we're just seeing a lot of great like AI and research came to the market and all of that requires new infrastructure build out. And now of course we're seeing that in full fruition. But I think this is not like overnight, this has been like a gradual change that everything that we are running all these AI workloads on actually not building for the specific workload. Like sure we have GPUs that are specifically tailored for large scale inference and training, but everything else when it comes to storage, compute and all the tooling, we consider that a big part of infrastructure are actually being revamped in real time. Like what are the agents using as tools, what are the orchestration layers again, how do we store memory and so on. Those are all the opportunities that that's emerging to build infrastructure. So that's why we raised the 1.7 billion round. And on top of all of that we're still in the very early innings of developing new research, new algorithms to develop even more powerful models. So all of that again will require a large amount of capital. So we want to be able to support all the founders building in the research and engineering space.
- [17:49]
Max
Rapid change. Absolutely. And what percentage would you say is is existing infrastructure that needs upgrades or net new that like you said there's so many new developments aren't things that necessarily existed before with AI?
- [18:03]
Jennifer Lee
Yeah, a lot of these two goes hand in hand. I think again this is my personal view. Like unlike consumer space, we invent completely new paradigm new interactions Infrastructure tend to be layered. They're built upon each other and it's layered upon old existing infrastructure with new layers. Like we always had transactional databases, but the vector storage was new to store embeddings. We have always again had memory and file systems and now agents are using them. So how to access these tools Like MCP is completely new layer but API has been around forever. So these are just like newer layers stack on top of each other to serve like either different Persona or different paradigm of like how tools are being plugged in together and utilized by either human developers or agents.
- [18:57]
Max
And you've raised the 1.7 billion and you're deploying on the infrastructure. And are there specific companies or areas that you're particularly interested in?
- [19:07]
Jennifer Lee
Yeah, we define infrastructure in the pretty broad way but there are certainly very focused areas we tend to spend time time on definitely foundation research of AI models. We have been longtime backers of OpenAI and thinking machines SSI. A lot of creative models from World Labs to 11 Labs to BFL and Ideogram. These are all the first party model developers starting from pre training building sort of either full stack application or launching the model itself as an API and so on. We spent a lot of time in core AI infrastructure and that includes again all the things I talk about of like systems storage data pipeline. I invest in a company called Redacto. Like what they do is really using vision language models to turn PDF documents into LLM ready data structure like that itself is a core piece of system that required for building agents and automating knowledge work. We still spend a ton of time in developer tools again because is the developer Persona is changing. It's not just humans writing code anymore. Now probably like more than 90% of the code are being written by agents and we need a new set of tooling for code review for like cicd, for pretty much the whole toolchain of software development. Cursor of course is one of the very prominent investment of ours on the IDE side and we continue to double down in this killer use case. I think there's a lot of interesting opportunities and after that their security like it is how to how do we secure both the software parameter but also the team and the people organizational parameter. We're seeing a lot of both sort of fear mongering scare of like what AI hackers will do. By the same time we can also secure that realm much, much better. And now we have these really capable coding agents. The software we're writing and launching hopefully will be much more secure too.
- [21:07]
Max
Yeah.
- [21:08]
Jennifer Lee
And then we'll look at you know, application stack as well as sort of traditional like data pipeline and data systems. Like these again are just like fundamental building blocks that's required for building great applications and software.
- [21:22]
Max
Incredible. And one of the companies that you've invested in you mentioned is 11 labs. You I believe led their series B in 2024.
- [21:30]
Jennifer Lee
We actually led Series A, B and C and participate in the D. Okay, okay.
- [21:36]
Max
So You've been with 11 labs since their series A?
- [21:39]
Jennifer Lee
Yes.
- [21:39]
Max
Through to their series D. Yeah. What did you see early in the voice AI space? Because that was a little bit prior to voice was as widely recognized.
- [21:47]
Jennifer Lee
Yeah. So there are two things really are the drivers behind on the thesis side investment? Of course, you know, it's just such a compelling team and founders that you know on that alone I think would have written that check myself. But we've been tracking the first synthetic voice space for many years and it's just never crossed a uncanny valley like you can literally just hear this is robotic sound. You know, it can automate some stuff to like, you know, narrate paper. But nobody will be like paying attention to that narration for a long period of time just because again, it's not engaging. It's not like we're having this conversation more human like and having all the intonation and emotions embedded in it. So when we first saw the 11 Labs demo, I remember they were like a gunned off voice to just like narrate a book or like holy shit, this is really alike. And also has all the right like pauses stressing intonation that it just like it's super engaging. And you can imagine how that's being used. For all the creative use cases we had a thesis around like yes, language model is really important. It's like driving all the intelligence. But all the media and creative models is actually where AI was having the best run and the best use cases just because there's no accuracy related to it.
- [23:07]
Max
Right, right.
- [23:07]
Jennifer Lee
Like you're not really judging these models by of course we'll see like six fingers, five fingers. But sometimes even the imperfection is like creation is creativity and we want really to double down and investing in flourishing creative ecosystem. And voice model is just such an important pillar because we need that in a video. We need that tool like you know, for podcasting, for again creative expressions and so on. So betting on sort of the voice model was also one of the thesis driving behind and we what didn't really came into picture or like we didn't really have the foresight to render. Right. Is again all these voice agents. Yeah. Like in 2022 it was still very early when we're still just like prompting the box like creating images and videos and generating like 30 seconds or like a minute episodes of like voice interactions. But I'd say voice agent, it was one of the first few agents really took off because all the deslex workers, customer service, front desk, like a lot of these use cases are just like really easily automatable because it's first, repetitive second, it is natural human language. It doesn't really involve a lot of like jargons and so on. So you can easily bring sort of the synthetic voice into the picture and having like a fluid conversation. And the language models was enough, good enough to complete many of those tasks being the driving force behind. So we're, we were seeing like this voice agent took off in like 2014 and we backed also Decagon in the customer support space and many others that are all using eleven's voice. So that was again just another unlock of how big the space and the market could be when you really nailed the accuracy and also the fidelity of the model itself.
- [24:52]
Max
Absolutely. And you've been with 11 labs amongst many other companies throughout multiple raises and stages of their journey. So you observe a lot on the go to Market side of how they're building. Curious what you're seeing in the infrastructure space, what's working in Go to market?
- [25:07]
Jennifer Lee
Yeah, it's such a great question because I've always felt I'm like an enterprise person through and through. Like in 2022 was the first time it turned into more of like a consumer die hard because again, all of these tools are first being picked up by the consumer and prosumers. And that was a great learning is like there's not really a clean line between like, what are your side hustle, hobby use cases and which ones are really like, you know, workforce, enterprise ready. Back then it was like a lot of the activities were happening just for fun and people are tinkering with the tools but really quickly. And this was the biggest difference from all the prior technical evolutions or like paradigm changes is like these products gets into the team enterprise, these work scenarios really fast because you just see the ROI really quickly to like gain either productivity, improve efficiency or just like, you know, creating things that's never been done before. So we think there's a lot of potential to build like vertical integrated products. And that's what eleven has done. Like they have a killer developer API, but they also have a really comprehensive product suite since day one, from the Creative Studio to now the agent platform. So they can tailor for both of the audiences. But we also have companies like Fall where they have always been developer driven, but really quickly they're able to go after the enterprise customers with sort of the workflow product that's built on top of it. So there's just a lot of opportunity to own a market and own the interface and Persona really quickly. And that sort of either you can call it kingmaking or like a brand effect was really prominent in AI too. Like, like people tend to go to a default product when they think about certain functionalities. Like it's pretty much synonymous of 11 labs versus voice models and fall versus general media or like video image models. Like this is a phenomenon that hasn't really happened before because it's always a bit of like oligopoly for enterprise space. There's number one, number two, but they're not far distances. But in this AI landscape, again because everything happens so quickly. Like how to get your brand and developer or like consumer recognition really fast to become that default choice is what I think all the go to market challenges and opportunities are. If you found the open space, just go as quickly as you can to become again the name that everybody knows and default to. And that's, you know, just a great advantage. And talking about moat really defensible position to be in.
- [27:39]
Max
Yeah, everybody's striving to have that Kleenex phenomenon Now on the B2B side and the speed to. To market on the go to market side and speed to own that category now is something that we haven't really seen before. So.
- [27:50]
Jennifer Lee
Totally.
- [27:51]
Max
Yeah, it's. It's very interesting. And one thing that you mentioned earlier about eleven Labs was the founder caliber. You said I would have written them a check based on that alone. And I'm paraphrasing a little bit here.
- [28:02]
Jennifer Lee
Yes.
- [28:02]
Max
But for any founders listening, what is that trait or traits that makes an investor feel like that?
- [28:10]
Jennifer Lee
I think if I simplify, there are a lot of like, you know, really amazing qualities now that I've worked as Maddie and Peter over a few. But I think if I dumb it down to like the beginning and what are the most outstanding qualities. It's really, I think the passion, conviction to the problem. It can come from anywhere. But like to them is a very personal story that they have just seen these like dubbed movies.
- [28:31]
Max
Yeah.
- [28:32]
Jennifer Lee
That are really boring with like only one monotone voice narrating through the whole movie that like really takes away all the emotions and excitement from it. Like they want to solve that problem. But second, they're like not only technically very talented, but also have a great product line side. I think that's again something always being overlooked by especially in these sort of technical driven evolutions is like if you have the best tech, we'll win. But that's only like half or maybe 60% of the equation. Like you still need to package it into a product that's easy to consume, easy to understand because these models are really capable. But sometimes, you know, they still need some guardrails and guidance to really bring the best quality out to the consumer and users too. And 11 labs even talk about this in one of our podcasts in length of like, how do they advance on the research side but also use product to and product functionalities to patch some of the imperfections of the research until it's ready? Let's say like we know in three months or six months this research will be ready, but now it's not, but it is something users are really craving for. How do we turn it into a product functionality that sort of feels a bit of the shortcoming of the model but can still deliver the premise not in the best way or in the most perfect way, but again like bring the future forward a bit. And then when the model is ready, you have the model replace that. Like that kind of understanding of like where a product can shine, where the technical and research can shine is a really important trade integration on both sides. And the last part is, you know, they took go to market really seriously. Not just winning consumers but also going after enterprises with you know, the API product, with the voice agents and agent platforms and their certainly winning in that space now too.
- [30:20]
Max
Definitely. That's great advice for founders too. And now as you look out across 2026, which we're almost midway through, which is crazy.
- [30:28]
Jennifer Lee
It is crazy.
- [30:30]
Max
What are the biggest changes that you anticipate seeing in the landscape? Because like you said, everything is moving so, so quickly. So if you had to put a prediction out almost perhaps 2027 at this point, what does the landscape look like?
- [30:41]
Jennifer Lee
I think we'll continue to see acceleration on the frontier Soda model. But the best news for the ecosystem is open source is catching up really closely and really fast. And I'm very excited about that because that's again how a lot of startups, companies can sort of combine the different level of intelligence, also like characteristics, economical value of them into again their system and compose a workflow instead of just one model run through all the way which is both slow and expensive. Very excited about a couple new modalities that are becoming more prominent. One is work model models, the second is vision language models. A lot of use cases are really just emerging from them. Whether it's you know, robotics or like real time intelligence. These are again our new unlocks that we have not had in the past because the model either was not good enough or not economical to run at large scale. So those are the ones I'm, I'm super excited about.
- [31:37]
Max
Very exciting, well, exciting times ahead. And you're investing out of, I mean $1.7 billion infrastructure and AI fund, but at the same time you've got some strong perspectives on human creativity. So I'm curious to hear what does the future look like with so many advancements on AI and human creativity. Can they coexist?
- [31:55]
Jennifer Lee
I have always been amateur but huge fan of like using all the creative tools from product design to now. All the like video image models sometimes maybe just for a simple like birthday cars or like you know, sharing with the team. Like something you have like visually in your head but you don't know how to make it into a meme. Now it's so easy, like it's at your hand. Again these are like very small use cases but again I think about how you know someone really creative have a lot of ideas but they are limited by the tools and the people and resources they have now they can literally have like a one to two people studio but making a full movie like with back a few of those creators and we're seeing their creative workflows and it's just really incredible to observe how the AI tools have given them power and given them the opportunity to tell the story, show their ideas. Actually just came from from the pitch this morning. Yeah, another really exciting piece is like the model quality have been improving in the last three years but it's finally gotten to a place that is ready for professionals. Like it's been great for consumer consumers, it's been great for entertainment advertising but now we're thinking about movie making. We're thinking about again like really premium storytelling for like luxury brands and so on. Like the models now have gotten to a place that can can really maintain brand consistency having like really complex workflow tools that can compose both image, audio, video models and work models to create something that again used to take teams of like tens of people month of work to come about.
- [33:26]
Max
Right.
- [33:27]
Jennifer Lee
And lastly I'll just tell like sort of a personal sort of dream list is I've always been a big fan of the one piece manga and they have this like live action movie on that Netflix. It's only every couple years for eight episodes, right?
- [33:41]
Sophie Buon Assisi
Yeah.
- [33:42]
Jennifer Lee
You have all the materials and even like an animated version like how can we make it faster like to. And that's a super long manga. I don't know if you know like the, the whole book can like wrap the earth and so on. Like how can we tell more of the audience like how amazing the story is through sort of live action engaging movie storytelling. Like now again we have have these really capable tools that we don't have to wait for years and years until like we can get to see the full episode. So that's what I'm looking forward to.
- [34:14]
Max
Good. That is the goal. Somebody's got to change that. And I mean it's a very inspiring prosperous future from what we've talked about. But there's often always blockers to that too. What's going to stop us from getting there?
- [34:27]
Jennifer Lee
I think you Know the technical advancement will keep happening, it's going to go faster. Like I'm not worried about that. Of course. Like I'm not saying any of these models are perfect. They still require a ton of gut railing direction to get to the point of like being able to be professionalized. I think more so of the like attitude and the adaptation of you know, the creative industry. The individuals who are still like having a bit of like insecurity around what do we, where does our job go and what happens when you know, these models are becoming so powerful. I still think inherently creativity is human expression and it is always going to be, you know, our nature and own owned, the origins being owned by humans. These are just great, great tools. Like it's not going to be able to replace, you know, a director or like a book author, someone that just have these great stories in their mind. Like the models will never be able to replace that in my opinion. And I would love to see more creatives just embracing these tools and making it part of their daily workflow and like becoming way more productive in being able to explore, express what is in their mind, what's in their heart and tell that to a much more broader audience.
- [35:39]
Max
Definitely. I always think about this quote I heard a long time ago. But it's the best ideas live in the graveyard. And I think this is a time that suddenly we've hit an unlock where people can be creative, they can shift. It's never been easier to build and ship and actually express. Like you said, bring out what's in in the mind to reality. So yeah, very, very exciting times.
- [35:59]
Jennifer Lee
No 100%. I think like every single human is creative inherently, but not every person have the best tools for them. Now it's not only like, you know, all the S tier and a tier, you know, creative storytellers can use the tools to become even better than where they are today. But a lot of, you know, us can use these tools to tell stories or like show really creative pretty things to our family and friends. Like that alone is just making everything better.
- [36:29]
Max
Definitely, definitely democratizing access to it. And if you could leave founders leaders is listening just one message, whether it's about the future or just an inspirational tidbit. Personally, what's the thing that's been in kind of the back of your mind that maybe you haven't brought to the world yet that you'd want to tell people?
- [36:46]
Jennifer Lee
I would say just make it habit. Make these products and tools your friends. Like it's taken us some years to like just get used to and now so dependent on our smartphone smart devices. Yeah. Now not only we have these really powerful machines, we also have such intelligent models that runs inside them. I think it's such a blessing that we're able to live in such a time and again with any power you kind of have to learn to harness it. And as much as you know as founders as operators can spend time and just like living and breathing it, I think you know a lot of good ideas will come out of it.
- [37:27]
Max
I love it. Jennifer, this has been fantastic. Appreciate the time. Thank you for joining.
- [37:31]
Jennifer Lee
Thank you so much. Thank you for the questions.
- [37:33]
Max
Absolutely.
- [37:36]
Sophie Buon Assisi
Thanks for listening to this episode of the A16Z podcast. If you like this episode, be sure to like, comment, subscribe, leave us a rating or review and share it with your friends and family. For more episodes go to YouTube, Apple Podcasts and Spotify. Follow us on X16Z and subscribe to our substack@A16Z substack.com thanks again for listening and I'll see you in the next episode. Episode this information is for educational purposes only and is not a recommendation to buy, hold or sell any investment or financial product. This podcast has been produced by a third party and may include paid promotional advertisements, other company references, and individuals unaffiliated with A16Z. Such advertisements, companies and individuals are not endorsed by AH Capital Management, LLC, A16Z or any of its affiliates. Information is from sources deemed reliable on the date of publication, but A16Z does not guarantee its accuracy.