Sorry to all the tech podcasters hoping for an exit. OpenAI isn’t planning to strike more deals like its acquisition of TBPN.
“We are not becoming a media company,” said David Duxin, OpenAI’s business and partnerships lead, during a fireside chat at the Scalable Summit on May 6 in Los Angeles.
Even so, the AI giant is still focused on expanding its creator strategy, which includes working with other creators to shape the narrative around its products and to get feedback as it builds out its tools.
Until now, OpenAI has been less vocal about how it works with creators beyond TBPN.
That’s in part because ChatGPT has benefited from a first-mover advantage, becoming the chatbot equivalent that Uber is for ridesharing. Its rivals, on the other hand, have had to work harder to build consumer awareness: Microsoft and Google, for example, have paid influencers upwards of $600,000 for months-long campaigns to promote their AI tools.
But the AI race is now fiercer than ever. ChatGPT competitors like Claude have gained steam and all of the tech giants are jostling to attract customers and be seen as cool among consumers. That’s where creators come in.
At the same time, creators are looking for ways to earn money from chatbots and want to make sure they’re showing up—and are credited—in results surfaced by chatbots like ChatGPT. So far, AI companies haven’t rolled out direct ways to make money, such as ad-revenue sharing with creators.
On stage, Duxin discussed all this, as well as OpenAI post-Sora, its AI video app that shut down in March.
The interview has been edited for length and clarity.
Scalable: RIP Sora! Sora was the most obvious move by OpenAI into the creator economy. Walk us through that decision. Sora really rattled Hollywood and the creator economy when it came onto the scene, and then suddenly it’s gone.
Duxin: RIP Sora indeed. I spent a lot of my time at OpenAI working closely with that team.
As much as these tools become force multipliers for productivity, I think we were really excited about the opportunity for them to be force multipliers for creativity too … Video really felt like a possible expansion of that.
One of the decisions we have to make at OpenAI is how do we allocate our compute? Without getting too technical, compute is really one of our key constraints. The thing about video and Sora is that it actually is very compute-intensive.
We had to make some prioritization decisions, and one of those was to refocus on a lot of our bets within image [generation] capabilities and what’s possible within Chat and Codex and some of our core products.
We were really inspired by the really nascent creator community that actually emerged from Sora.
How important are creators now post-Sora?
One of the most interesting elements about creators is just in general how tech-forward they usually are. In the tech world, we talk a lot about developers, and I like to say that creators are actually going to be every man’s developer.
They’re actually very similar. They are nimble small teams. They’re working iteratively to scale audiences or build businesses, very much like developers.
What’s interesting about AI is now we're seeing a blurring of the lines between consumer, prosumer and enterprise, and that individual people are becoming businesses.
Creators have been swimming in this direction for 10-plus years. As we get to a point where an individual or a small team can build something that previously required an entire company or several teams, we see these tools as massive accelerants for creators because they've always been inclined to adapt technology and kind of push boundaries in ways that allow them to do more with less.
When you look at creators, they wear so many hats. They're operators, marketers, content creators and entrepreneurs, all rolled into one. We see them as amazing examples of how AI can expand what an individual or a small team can do. From the OpenAI side, we’re really interested in continuing to invest alongside those groups.
So what does that strategy look like? Is it influencer marketing? Is it converting creators to be paying customers?
It’s generally two-pronged. There’s influencer and creator marketing, which we do and are doing more of. Creators usually have a subject matter expertise or a passion and a community that is equally passionate about that.
One of the opportunities in AI is that these tools are so capable that a lot of people don't actually know where to start. They don't know what they can use it for. And so when you have a creator who can show their audience, ‘Hey, here's what I can use this technology for that's relevant to me, that's relevant to this community,’ it’s incredibly powerful, right?
So you take a beauty creator as an example. They can use ChatGPT to evaluate different products to think through what blender might work best on their skin tone. Maybe you are allergic to a certain ingredient, and ChatGPT can help you navigate which products might be better or worse for you. Having creators showcase these use cases to communities that share that passion is super, super important to us in communicating those benefits of the products.
The second area is really doubling down on showcasing how creators can use these products to accelerate all the different business work that they do. So more behind the scenes. We do a lot of work there, and we get a ton of feedback from these cohorts.
What kind of feedback?
I’ll give you a great example. In advance of model releases or product releases, we'll often give early access to groups of creators. About a year ago, we released our first image gen model, which was a really fun consumer moment, and I think really showed the world that these models could be really helpful in creative processes.
We got a lot of feedback from creators that said, ‘This is amazing. There's the kernel of something really great here, but we want more control. We want to be able to edit the aspect ratio, or the text is kind of weird, or it needs to look a little bit more realistic.’ And so, fast-forward to about a month ago, we released our second image gen model, and it’s a much more intelligent image model.
What can it do? You can do any aspect ratio you want. It’s really good at directly editing text. It's really good at preserving likeness and photorealism.
In advance of that launch, we gave several creators early access and again, got really good feedback of how this was so useful to them from using it to generate thumbnails to infographics, which previously were really difficult for these models to do.
That feedback was super exciting to us because while we want these models to be great for consumers, it’s really important that they’re useful in people’s work. Creators are so adaptive in how they think about this, that they're in an amazing first group to go to, to get that early signal and that early feedback of: Are we marching towards the right things?
I don't think anyone saw your acquisition of TBPN coming. As part of the deal, the hosts will also be helping you with marketing. Are there going to be more deals like TBPN? Is this part of the creator strategy, or is this a one-off? Every podcaster wants to know.
I will start off by saying no, we are not becoming a media company.
After that acquisition was announced, I got so many calls asking the same question from probably a lot of folks in this room.
TBPN is an example of creators building something highly valuable for a specific audience. I think they really showcase the outsize impact that a small team can have.
What they do is directly in support of our mission. Our mission is to ensure that the benefits of AI reach all of humanity. A big piece of achieving that mission and a critical step is communicating those benefits.
What [TBPN hosts] John Coogan and Jordi Hays’ team have done is bring so much passion and excitement to the topic of innovation in the tech industry. That has resonated not just with the business community, not just with the technology community, but I think they've injected that into the cultural zeitgeist in a way that is so unbelievable, and I think really speaks to the power of creators. We’re really excited to continue to partner with them on that specifically because it's really important to our mission.
What they're doing is part of a bigger shift—which I think everyone here probably sees—is that trusted voices in niche communities can really have outsized reach and influence.
But you’re not trying to acquire “All-In” next?
No, not the plan.
We alluded to this earlier, but Sora sparked a lot of concerns within Hollywood. Creators still have concerns about AI, like deep fakes and the misuse of their likeness. What are some of the biggest things you hear from creators? And how do you think about that when you're building products?
We get a lot more questions than concerns, especially in a post Sora world.
Sora, I think, was just something that was so close to the content creation process that people had a lot of questions about. They had a lot of strong feelings. As we look past that, most of the conversations we have with creators are really around, ‘Where can I start?’ ‘How can this be more usable for me?’
What can be intimidating about some of these technologies, ChatGPT or Codex, is that it's really good at a lot of things that there's a bit of a blank page problem. It's tough to think about where to start.
There's a lot of product work that we're thinking about to make that easier. So things like connectors to tools that creators already use. Then there are questions around: How can I participate in this ecosystem?
We get a lot of questions around, can I advertise on ChatGPT now that you guys are building out an ads program? How do I show up in search? What does SEO look like? And so these are areas where we're aiming to communicate better and then also just really build out the infrastructure to support creators in a way that meets them where they are.
We’ll get to SEO in a second, but I want to talk a little bit about money. Because creators are very interested in earning money from AI chatbots. I understand that making money was on the road map for Sora. So what does creator monetization look like now? You used to work at YouTube. Ad-revenue sharing is the gold standard, but do you think that there’s an opportunity for creators to earn money directly?
Right now we're focused on helping creators use these tools to accelerate their businesses, to earn money in all the other things that they’re amazing at. For what that looks like in practice, and what we've seen, is people using these tools to just rapidly increase the cycle time of the things they're doing.
I spoke with a creator a couple of weeks ago who was interested in thinking about: What would it look like if I launched a wellness brand? And they were like: Well, I used Codex to think of an idea that makes sense with my brand, build out the branding, mock up a website and what would the product look like, all in an afternoon. [The creator] was able to go through five different business ideas and vet with a subset of [their] audience and things like that.
When we think about how we can most help creators in the near term, it's really making these tools accessible to them in ways that can accelerate all the things that they’re really good at now.
So less on the direct monetization side at this point.
Yeah, less on the direct monetization side. I know you wanted to get back to SEO, but I think that's one of the ways in which we think the ecosystem can also be helpful. Not just people using our tools, but showing up in this ecosystem in ways that can drive them new audiences and deeper engagement.
So let’s talk about that because I used to work at CNN, and there was a whole SEO team. We would think about the title, which would be different than the headline. We're starting to see a tiny bit of traffic coming where we're seeing newsletter subscribers come from ChatGPT. I would like more data on what article, all that kind of stuff. But what's your advice to creators and brands in the room? Everyone is trying to navigate: Do we do something differently? How do we show up and get credit?
So first of all, credit is super important when you use ChatGPT. Citation is paramount for us. So that's one piece.
In terms of the equivalent of SEO. I've worked on a bunch of our publisher partnerships over the last several years, and we get this question all the time. The answer is one that I think sometimes can feel unsatisfactory to most, because there isn't a silver bullet in the way that SEO had. There isn't a direct set of optimizations you could do.
What these models are really good at is matching intent with content. When someone asks a question on ChatGPT or a research question, these models are really good at understanding that intent in the context of how you're asking the question. Maybe something it knows about you, or the entire context of your conversation. It tries to find the set of sources that really matches that intent super well. The way to show up well here is to produce high-quality content that matches that intent.
What does high-quality content mean?
So in this case, it's kind of different than SEO because it's more than just the headline. It's really the substance of what is in the actual article or content. These models can distinguish between what's just a headline that maybe would catch someone's eye and what's actually substantive enough to answer that question.
Where we see the opportunity is the way people have previously interacted with content online was a very general level of intent, which is: I like a certain publication, I go to that website, I look for something that's interesting to me, or a very narrow level of intent, where I search for something very specific. There’s all this intent in between that people are using these natural language chat interfaces to get answers to. That's where I think there's an opportunity for content owners to show up in other places that they otherwise would not have been.
So what’s your advice? I think we put a lot of effort into our newsletter. Is there anything else we can be doing beyond that?
One obvious thing is just make sure you have the right permissions to show up in these ecosystems. We give a lot of granularity to how people want to show up—if you just want a certain article or all of the publication.
We want to be able to give people insights into what’s trending, what are people using these tools for, how are you showing up in these answers, and more analytics.
We're still pretty early on this journey. ChatGPT is only about three years old right now, which is crazy. I think people forget that. I forget it too. So compared to other search engines, which have been around for decades, they're a little farther up the curve than us.
But you have the benefit of learning from them.
Exactly right. We can draw on a lot of the learnings, and I think it's much easier to understand now what's important to these folks in the ecosystem, like the analytics.
We’re really working to bring more insight for those folks who want to participate here so they can see how they show up in this new paradigm.
What’s your best ChatGPT tip?
When I showed up to YouTube, I thought everyone used YouTube the same way I use YouTube—at the time, which was like music videos and how-to videos.
You look at the data, and you realize it’s actually incredibly long-tailed. Everyone uses it for some different use case. It's like infinitely niche communities. I think that's actually true with ChatGPT. Everyone uses it very differently to kind of suit their own needs.
The way I use it might be different than the way you use it. Some people use it as a health coach, a therapist or a co-founder.
My best advice actually is to truly treat it like a partner in whatever you're doing. Whatever you're going to do, treat it as your high-powered assistant that you have and ask it to do a first pass and give it feedback and tell it what it did wrong and what it can improve upon and what you like and what you don't like.
And in a week, you'll probably have something that feels pretty helpful, and you can start to tack these things together and build really cool things.
It's really hard to give a specific one-size-fits-all tip to everyone. So the best I can say is try to use it as your partner for as much as you can, and see what works. I think you'll be pleasantly surprised.
