Sequoia’s Sonya Huang: The generative AI hype is 'absolutely justified'

She’s bullish on generative AI given the “superpowers” it gives humans who work with it.

Sonya Huang

Sonya Huang, a partner at Sequoia, shared her thoughts about generative AI's future.

Photo: Sequoia

Generative AI can create more than just text and images — it’s clearly generated a hype cycle around AI companies and rabid investor interest in the space.

In a bright spot for an otherwise lackluster funding environment, two “gen tech” companies became unicorns this week after Stability AI, the company behind the wildly popular image generator Stable Diffusion, and Jasper, which makes an AI-powered system that writes marketing copy, both announced funding rounds. The launch party for Stability AI drew people like Sergey Brin, Naval Ravikant, and Ron Conway into San Francisco for “a coming-out bash for the entire field of generative A.I.,” as The New York Times called it.

The buzz spilled over onto Twitter when a market map of companies in the space went viral after Sequoia partner Sonya Huang laid out the companies (including Stability AI and Jasper) that are building in generative AI, solidifying the space as a sector and not just a handful of companies making fanciful images.

A firm believer in the “superpowers” humans can gain by working with machines — she’s used the language model GPT-3 to help write a blog post on the future of generative AI — Huang talked with Protocol about what sectors founders are building in, the looming ethical concerns, and whether the hype in AI startups is overblown or justified.

This conversation was lightly edited for length and clarity.

I don't think I've ever seen a market map go quite so viral on Twitter. Why do you think it generated — forgive the pun — so much attention when you posted it?

It’s speculative, but I think a few things. One, I would say I think the world is in dark times right now. People are looking for something to latch on to that is hope, and generative AI appears to be that. So I think part of it is just the emotional [impact] where we see so many terrible things happening in the world and [think], “Wow, what an exciting time to be in technology if you look at what's now possible.”

Then the second thing is there's been a lot of really exciting progress that's been shared publicly in terms of models getting bigger, better, and amazing images that are flying around Twitter. But it was probably new to think about this as a model versus application layer. People were thinking, “OK, these models are not only going to be super interesting things that we can play with to make a fun image, but they're actually going to drive the future of how we work and the next big application companies.” That was probably a new framing that people hadn't really thought about in that way. Those are probably a couple of the reasons it took off, but it’s speculative. I don't know what makes things go viral on Twitter. I'm not good at Twitter. [Laughs]

For me, it was interesting because I wrote a few weeks ago about how VCs were exploring whether this was a toy or the next big thing, and then there’s been this huge shift to VCs saying it's definitely going to be the next big thing. Why do you think there has been a shift? Was it just companies announcing fundraises this week?

I think it’s market maturity. We’ve been following pretty closely these large models for the last several years, and if you look at what's possible, it is pretty mind-blowing just the rate of progress. There is some benchmark, which is human-level performance, and now that these models are just in the last couple of years starting to exceed that, only then can you have AI that really, really augments how we work. Because if it's not as good, the technology's not ready. So the first thing I'd say is, the technology is finally getting ready.

Then the second thing is just that access to these models is now available. Obviously Stability has made an incredible amount of progress and developer interest because they've just made their models completely open. And OpenAI has also made their models more publicly available as well. I think GPT-3 was in closed beta until last year, if I'm remembering correctly. So these models are finally available for people to play with, which they weren't before. Once technology becomes available for people to play with, it is very natural for step one to be very cool demos of what's possible. I almost compare it to when the iPhone came out: It was a bunch of really gimmicky stuff that came out at first, but then you've got people who are really thinking about the business applications in deep ways, and we're starting to see that.

This is obviously still a pretty nascent area, but where are you seeing most concentration of activity so far?

Image generation is a big one. Images speak to us so viscerally, and so they're a lot more fun to share on Twitter than whatever GPT-3 could spit out for me. By nature, images are a lot more viral, so we're seeing a lot there. People are moving on beyond one-shot image generation to a few different branches, whether it's “Let's make this really good for the process of interior design or product design” or “Let's make it a really, really good image generator that you can continue innovating with the machine on” or “We want to take all these images that people are doing and the community we've built, and build the next big social media platform.” People are going different directions from image generation, but I would say that's been an area of just incredible interest from both founders and from users because you need user interest for these things to work.

Then the other big category where there has been a lot has been in the text space. And in the text space, who needs to write all the time? Marketers. So there’s a lot of these marketing Gen AI companies, and some of them are really working. We're seeing it evolve, as well, where people started from shorter-form generations and now we're getting really, really long form. We're getting creative writing, we're getting scripts and novels. It’s pretty good.

And then code. Code is one that OpenAI has cultivated for a while, and I think GitHub Copilot is incredible. The stat — [that] they’re responsible for 40% of their users' code — is just mind-blowing to me. I've seen the demos from Replit — it's pretty extraordinary. And so code is the other effort where we're seeing a lot of both exciting founder development and then also user interest.

The other categories — the boxes on our landscape that are relatively sparse now — I don't think they're gonna be sparse for long. When I first put out our blog post about how the tech and how the different pieces of technology are becoming ready, I thought 3D, video, bio, they were going to take longer based on some conversations. Basically everyone wrote in to me like, “You're wrong, this stuff is happening way faster than you think it is.” And they were right. I think similar to what you saw with text and image happen where the models were a couple years back, I think you'll start to see the application space start to flourish for these other modalities as well.

Are there any areas on the map that you left off that you're excited to see happen someday? Any other to-be-discovered areas?

We organized the map by modality, which I thought was most relevant just because it's the enabling technology that is creating the application within each box. I do think that a lot of the most interesting companies will own the end user, but they will be multimodality. I couldn’t fit that in a neat space on my map, but that would be one.

Another would be this concept of an AI companion or AI copilot. It doesn't really fit neatly in the map anymore, but you have something intelligent — browse the web, stitch together all your different tools to do things for you — versus currently the map is very much focused on tools that you work with for a very specific test task.

What do you think about the hype cycle around AI right now?

Well, I would be alarmed if the hype was really high and the results weren't there. The hype is high, and I think part of that is, again, people emotionally want to attach onto something that gives them hope and optimism, but the results are there as well. You see what these models are capable of doing. You see the fundamental technology. You see the applications that people are starting to build. I’ll go back to “writing 40% of people's code” example — that is phenomenal both technological progress as well as economic value delivery. So absolutely the hype is high. I think it's absolutely justified given the results that we're seeing. My hope is actually that by putting out this landscape, we plant that seed and a lot of future founders that have been trying to figure out what to build next, I think it's wonderful to draw them to this. If I was a founder in [Y Combinator] right now, I would 100% be pointing my guns at one of these models and seeing what I can do.

What about some of the ethical concerns? I spoke with Khosla’s Kanu Gulati about this a few weeks ago and there are some real concerns around copyright and other shortcomings in the space. What do you think the role of investors should be in addressing some of these?

Absolutely. I agree with the ethical concerns. The copyright concerns are a problem that's really important and hasn't been solved yet. There isn’t a legal framework for this stuff. Ideally, this is a constructive dialogue between all folks on all sides of the table because until there is that clarity, I think it's very hard to make progress. I think our role would be helping ensure that whatever rules are crafted are clear so that once you have that clarity, everybody can innovate on all sides of the spectrum, and hopefully everybody feels good.

If I was a founder in [Y Combinator] right now, I would 100% be pointing my guns at one of these models and seeing what I can do.

One of my favorite responses to your map was if it was truly generative AI, it'd be able to generate a market map itself. Are we gonna get to that point anytime soon?

[Laughs] Actually, I’ll tell you how I created the map in the first place: I went into GPT-3.

Before even the map, we put out this blog post of what was going to happen. It wasn't clear in my head even how to define generative AI. There was some stuff on the internet that wasn't that good, and so I literally put it in OpenAI, “the difference between classical AI and generative AI,” and it started spitting out amazing stuff. So that became a lot of the base for our article. It wasn't just a joke that the article was co-written with GPT-3; it actually was. And then, I'm not the most creative person, and I was having trouble brainstorming what the applications will be, so I typed into GPT-3, “the potential applications will be …” and then it started spitting out things, and I was like, “and then 10 more of them will be …” and it just kept on going, which is amazing. And then I'd be like, “Specifically for image generation, you can think of it as ….” That human-machine iteration loop I hadn't experienced before, and it was very much how we created both the blog post and landscape.

It’s cool to see how the point of generative AI is that it can generate things that you don't think about. It’s a new frontier for a lot of tech companies.

What I would have loved is if we have eight companies in a bucket on the map somewhere, I would have loved to have a natural way for having a machine that would browse the internet and find companies that sound similar and suggest them for my map. There isn't a great product encapsulation for that yet, but as we dream about how this might play out, I would guess it’s probably not that far out.

I'm very curious if I can train a model on my newsletter and have it write my newsletter and then submit it to my editor to see if my editor can tell the difference of who wrote it. Although maybe I shouldn't do that because I'd be out of a job one day.

I’ve had it write some investment memos for me and I swear it was as good as what I can write. [Laughs] To your point about being out of a job, I realize it was said in jest, but there's the knowledge and the craft of being able to work with the machine and I think that is a new skill that we need to learn. But I think once you master it, you're better than before, right? You're more productive, you're more creative, whatever it is, if you can really really embrace the machine. I don't think there's a world where the machine fully replaces us. We have to train how we work with the machines, but I think the result really is we are superpower humans as a result of being able to work with these machines.

Image: Sequoia


Judge Zia Faruqui is trying to teach you crypto, one ‘SNL’ reference at a time

His decisions on major cryptocurrency cases have quoted "The Big Lebowski," "SNL," and "Dr. Strangelove." That’s because he wants you — yes, you — to read them.

The ways Zia Faruqui (right) has weighed on cases that have come before him can give lawyers clues as to what legal frameworks will pass muster.

Photo: Carolyn Van Houten/The Washington Post via Getty Images

“Cryptocurrency and related software analytics tools are ‘The wave of the future, Dude. One hundred percent electronic.’”

That’s not a quote from "The Big Lebowski" — at least, not directly. It’s a quote from a Washington, D.C., district court memorandum opinion on the role cryptocurrency analytics tools can play in government investigations. The author is Magistrate Judge Zia Faruqui.

Keep ReadingShow less
Veronica Irwin

Veronica Irwin (@vronirwin) is a San Francisco-based reporter at Protocol covering fintech. Previously she was at the San Francisco Examiner, covering tech from a hyper-local angle. Before that, her byline was featured in SF Weekly, The Nation, Techworker, Ms. Magazine and The Frisc.

The financial technology transformation is driving competition, creating consumer choice, and shaping the future of finance. Hear from seven fintech leaders who are reshaping the future of finance, and join the inaugural Financial Technology Association Fintech Summit to learn more.

Keep ReadingShow less
The Financial Technology Association (FTA) represents industry leaders shaping the future of finance. We champion the power of technology-centered financial services and advocate for the modernization of financial regulation to support inclusion and responsible innovation.

AWS CEO: The cloud isn’t just about technology

As AWS preps for its annual re:Invent conference, Adam Selipsky talks product strategy, support for hybrid environments, and the value of the cloud in uncertain economic times.

Photo: Noah Berger/Getty Images for Amazon Web Services

AWS is gearing up for re:Invent, its annual cloud computing conference where announcements this year are expected to focus on its end-to-end data strategy and delivering new industry-specific services.

It will be the second re:Invent with CEO Adam Selipsky as leader of the industry’s largest cloud provider after his return last year to AWS from data visualization company Tableau Software.

Keep ReadingShow less
Donna Goodison

Donna Goodison (@dgoodison) is Protocol's senior reporter focusing on enterprise infrastructure technology, from the 'Big 3' cloud computing providers to data centers. She previously covered the public cloud at CRN after 15 years as a business reporter for the Boston Herald. Based in Massachusetts, she also has worked as a Boston Globe freelancer, business reporter at the Boston Business Journal and real estate reporter at Banker & Tradesman after toiling at weekly newspapers.

Image: Protocol

We launched Protocol in February 2020 to cover the evolving power center of tech. It is with deep sadness that just under three years later, we are winding down the publication.

As of today, we will not publish any more stories. All of our newsletters, apart from our flagship, Source Code, will no longer be sent. Source Code will be published and sent for the next few weeks, but it will also close down in December.

Keep ReadingShow less
Bennett Richardson

Bennett Richardson ( @bennettrich) is the president of Protocol. Prior to joining Protocol in 2019, Bennett was executive director of global strategic partnerships at POLITICO, where he led strategic growth efforts including POLITICO's European expansion in Brussels and POLITICO's creative agency POLITICO Focus during his six years with the company. Prior to POLITICO, Bennett was co-founder and CMO of Hinge, the mobile dating company recently acquired by Match Group. Bennett began his career in digital and social brand marketing working with major brands across tech, energy, and health care at leading marketing and communications agencies including Edelman and GMMB. Bennett is originally from Portland, Maine, and received his bachelor's degree from Colgate University.


Why large enterprises struggle to find suitable platforms for MLops

As companies expand their use of AI beyond running just a few machine learning models, and as larger enterprises go from deploying hundreds of models to thousands and even millions of models, ML practitioners say that they have yet to find what they need from prepackaged MLops systems.

As companies expand their use of AI beyond running just a few machine learning models, ML practitioners say that they have yet to find what they need from prepackaged MLops systems.

Photo: artpartner-images via Getty Images

On any given day, Lily AI runs hundreds of machine learning models using computer vision and natural language processing that are customized for its retail and ecommerce clients to make website product recommendations, forecast demand, and plan merchandising. But this spring when the company was in the market for a machine learning operations platform to manage its expanding model roster, it wasn’t easy to find a suitable off-the-shelf system that could handle such a large number of models in deployment while also meeting other criteria.

Some MLops platforms are not well-suited for maintaining even more than 10 machine learning models when it comes to keeping track of data, navigating their user interfaces, or reporting capabilities, Matthew Nokleby, machine learning manager for Lily AI’s product intelligence team, told Protocol earlier this year. “The duct tape starts to show,” he said.

Keep ReadingShow less
Kate Kaye

Kate Kaye is an award-winning multimedia reporter digging deep and telling print, digital and audio stories. She covers AI and data for Protocol. Her reporting on AI and tech ethics issues has been published in OneZero, Fast Company, MIT Technology Review, CityLab, Ad Age and Digiday and heard on NPR. Kate is the creator of and is the author of "Campaign '08: A Turning Point for Digital Media," a book about how the 2008 presidential campaigns used digital media and data.

Latest Stories