🎤 Comment Netflix utilise l’IA pour transformer la production de contenus (Elizabeth Stone, Netflix)

In the digital world. Our position and very strong philosophy on AI for production is that it is a tool that creators use to bring their vision to life. So if you think about the diversity of filmmakers we work with, there are of course filmmakers or creators who are not inclined to use AI or newer technologies. That's fine, we work with them every day. There are other filmmakers or creators who are very enthusiastic about experimenting and see generative AI as a tool that allows them to do things that would not have been possible otherwise, and we want to work with them and support them as well.

Hello.

I had never thought of it as a super CTO role, but it's a nice compliment. So let me describe my role that way. There is a set of different functional expertise: product management, experience design, engineering, data science that are all part of a single organization at Netflix. We cover many different areas of the company, such as content, studio production, core user experience, advertising, gaming. So there is a lot of breadth in addition to depth in functional practice.

Our recent priority, which will remain so in the future, is that as we expand our content offering beyond films and series to include live content, games, podcasts and more, we must think of new ways to bring that to members seamlessly, very consistently in terms of user experience, and effortlessly to discover something you will love. One of the things I deeply believe is that entertainment is no longer a single format or one thing. So we want to be able to program to meet that depth of what consumers are looking for. Sometimes watch a movie, sometimes interact with a game, participate in the outcome of a live title, sometimes want to spend time on mobile where you are looking for something entertaining during a commute or at a restaurant rather than just sitting in your living room watching TV. So the product and technology to support that are quite complex. So if it succeeds, it is almost invisible because it works so well, it's so seamless that you don't notice the technology. But an excellent example of one of the things we have increasingly leveraged is Open Connect, which is our content delivery network. It is global. We have servers in thousands of locations that help us deliver not only on-demand video content, so our entire library of movies and series, but are also now essential for live content and to do that seamlessly, as well as for things like cloud gaming. So some of the board games we launched this year or the soccer game, the World Cup, the FIFA, these are things for which this underlying technology was not only innovative at the start but had to continue to develop and innovate to support these new types of content. So that is a major axis. The second is personalization, our way of thinking about recommendations to ensure we show you the right title at the right time for you. And that becomes more difficult as we have more types of content that we could potentially show you at any moment. So we must become more sophisticated in our approach to recommendations and more sophisticated in the user interface. So we updated our TV platform last year. We updated, or are in the process of updating, our mobile platform, and these two things allow us to develop the technology and product experience in a way that truly brings more value to members.

That is absolutely one of the risks when making a major user interface change. What I love at Netflix is that we are not afraid to take that risk if we think the innovation will benefit members. Our new TV platform and the interface we deployed last year are, I think, an excellent example. Even though we had what I consider an excellent interface and user experience that had required more than a decade of optimization, we felt that it would not allow us to innovate as much as we would like, both from an interface and backend perspective. Among the examples of things we changed, not only is the visual different, more high-end, intuitive and responsive for recommendations, but it is actually even more exciting under the hood in what you don't see, namely a more modular base for building the TV platform, which allows us to test and experiment with new types of things, to build the page and experience with much more direction or creativity than before. So it was a risk worth taking because we felt it would be a better foundation for future innovation, and we have found that consumers have reacted very well to it. Of course, there are always varied opinions and perspectives on that, but for the most part, it allows us to present the variety of our offering more attractively than before. And it's the same for the mobile changes we are deploying. It's a new navigation in addition to a vertical mobile video feed that is a very immersive and fun way to discover content that we didn't have before. So again, it's a change worth making. So yes, consumers get used to it, but there is also a pleasure and utility in it that makes us want to keep pushing ourselves to offer something new and great.

Our position and very strong philosophy on AI for production is that it is a tool that creators use to bring their vision to life. So if you think about the diversity of filmmakers we work with, there are of course filmmakers or creators who are not inclined to use AI or newer technologies. That's fine, we work with them every day. There are other filmmakers or creators who are very enthusiastic about experimenting and see generative AI as a tool that allows them to do things that would not have been possible otherwise, and we want to work with them and support them as well. So in a way, AI builds on a reputation and strength we have long had of providing creators with a wide variety of tools so that whatever way they want to bring their story to life, we are able to support that and be a partner with them. What we also firmly believe is that humans will remain at the center of creativity and storytelling. So even if a creator wants to use technologies of different natures, we always work with creators where there is a human with a vision and a story they want to tell, and technology can be useful to help bring that to life. So that's what we focus on.

I think there will be several scenarios, and in a way it's too early to say exactly how it will play out. We certainly see many filmmakers who are enthusiastic about this, not exclusively or even primarily for cost reasons. They are enthusiastic because it allows for better quality content and to do things like reshoot a scene. El Altamira is a good example of being able to build a scene that would have been impossible to do otherwise. And so these are use cases where I think it's too reductive to say, 'You know, it's just to reduce costs or it's just to exclude humans?' Because in a way, it opens up the possibility to do more and better, and that's very exciting for the industry.

So examples: construction. I think another excellent example is actually a recent title where the set of Madrid was built, so that we could have what we call a kind of digital double or something that allowed building the background of the scene and making modifications while humans were still playing their roles. But how to reconstruct a Madrid that is perfectly faithful to the buildings, the cars, the ambiance and atmosphere of the city? And so that type of digital double is something that generative AI can create, that environment, and then the actors remain at the center to tell the story and bring it to life. So I think that's a good example. Another good example is actually in the creative process, using generative AI to pre-visualize or agree on where we are going creatively. What will be the style and mood of this story we are telling? This creates many possibilities earlier, even before starting to shoot. And that has been beneficial for many of our productions where filmmakers say it helps speed up certain steps and give people an idea of the visual behind what they have in mind before the point of bringing actors on set and building the story itself.

I'm sure it will happen. But from Netflix's perspective, humans will be the actors. Humans will be the ones at the center of the story, not only in acting but in the creation of the story itself. That's really the domain of the creators we partner with and focus on.

Builder rather than buyer of technology. That has been true throughout Netflix's history. So for example, what I mentioned earlier about Open Connect, our content delivery network, has been a huge advantage and it's quite unusual to have developed it in-house, but it has been an advantage for us and for subscribers. There are plenty of examples of technologies we built and initiated to innovate. As for LLMs, in that part of your question, we work with many different partners and it depends on the use case, whether it makes sense to develop or build the capability in-house or rather partner with or leverage something someone else offers. There is no one-size-fits-all solution.

Exactly. It's true that every engineer at Netflix has enormous autonomy, independence, and that comes with enormous responsibility. Of course. So yes, we have great trust in the engineers, we have phenomenal talent on the team at Netflix, and they are able to make decisions on the spot to submit code, make changes. Of course, we have good practices regarding testing protocols, safeguards, or things to watch out for, especially for tier 0 or tier 1 services. The last thing we want during a big NFL game or a big boxing match is for a piece of code to cause problems. But we have found that this unusual responsibility is something that engineers appreciate having, and they do an even better job because they know they have that freedom of action.

Fail fast, fix [laughs] even faster. OK. Honestly, the best position we can have is that of course things will go wrong. How quickly can we find them, diagnose what's happening, and fix it? OK.

Yes. So an incredible example of that is our partnership with TF1 that we announced, I think, a year ago to the day, and we will soon deploy an integration of all of TF1's content with Netflix, which creates an opportunity to have an even wider audience from TF1's perspective, and from Netflix's perspective, it's an opportunity to resonate even more with local audiences in France. So it will include video-on-demand content, linear channels, and live. So we consider this partnership as something great for both companies and we are very excited about launching it. We also have a lot of French content that is local. These are local creators and local creative teams who develop, find their audience, and Netflix enables that locally. So I would not see Netflix as an American company with only American content because we also invest very deeply in local content.

My favorite show right now is rewatching Tour de France: Unchained. The documentary about the Tour de France that I watch again every year, just before the start of the race because once the race starts, I wake up around 3 AM California time every day to follow it. So it gets me in the mood to watch that documentary.