Why Netflix Collapsed

Netflix is dying, but not because their shows aren't popular. Squid Game was by far the biggest show of 2021, but Netflix stock has lost $200 billion in value over just the last few months. So what's going on? Well, the simple answer is that Netflix has stopped growing. But digging into the details tells a much more complicated story. See, while it's true that Netflix isn't gaining new subscribers anymore, the reasons for the slowdown are what we really need to analyze here. And interestingly, I recently found a young artificial intelligence researcher who's building a new technology that could completely transform Netflix as we know it and turn them back into the tech powerhouse that they once were. So let's figure out why Netflix has lost two thirds of its value in just six months and create a plan for how they can get back on top. Basically, everyone has a Netflix account now, and that's actually part of the problem here. The last two years have forced millions of people to spend more time inside. And Netflix was a great option to stay sane when you can't go out and do normal activities. Activities. This led to incredible growth for Netflix. Throughout 2020 and 2021, the stock nearly doubled as everyone spent more time streaming content from their couches. But this actually accelerated one of the biggest problems for market saturation. Anyone who could afford Netflix and likes streaming content basically signed up over the last two years. Initially, most analysts who looked at the massive growth in Netflix subscribers figured that these people were subscribing just a bit earlier than expected. But the reality was much worse. These new subscribers weren't merely pulled forward. They were essentially all of the remaining people who would ever subscribe. The team at Netflix began realizing that they had essentially converted every possible customer and growth had to slow down. Their chief financial officer even acknowledged this during their end of year review.

Overall, the business was healthy. Retention was strong. Churn was down, viewing was up. But on the margin, we just. We didn't grow acquisition quite as fast as. As we would have liked to see.

Then, just a few months later, Netflix announced that they actually lost subscribers for the first time in over 10 years. Obviously, it's difficult to grow your subscriber base when pretty much everyone already subscribes. But there are a few other problems they're dealing with right now. First, people share Netflix accounts like crazy. This wasn't a problem back when Netflix was sending DVDs through the mail. But now that you can just log in with a friend's Netflix password and watch whatever you want, why would you pay? Now Netflix does have a plan to deal with password sharers, but we'll get to that later. A potentially even bigger problem for Netflix is direct competition. For years, Netflix has been the king of streaming. And most big media companies have really struggled to transition away from old school TV programming. But Disney completely changed that perception. They launched at the end of 2019 and already have 130 million subscribers. That's more than half of Netflix. This is a big problem because streaming content doesn't have a network effect like Facebook or Twitter. If a Netflix customer cancels their subscription and switches to Disney plus, it's not like they're leaving a bunch of social connections behind on Netflix. They can just move over in an instant. Netflix does have one major advantage though. They're a global company, it helps to look at a map here to understand what's really going on. See, even though Netflix lost subscribers overall, they actually grew in the Asia Pacific region. That growth wasn't enough to offset losses in America, Europe, Africa and the Middle east, though. The global nature of Netflix is a double edged sword. On the one hand, hit shows like Squid Game can become international sensations and drive growth for the entire platform. But on the other hand, many countries don't have as much disposable income to spend on streaming services. India is a great example here. Average American incomes are roughly 20 times higher than average Indian incomes, but India has 4 times the population. It's clearly a big opportunity, but Netflix has to develop new pricing strategies if they want to start growing significantly in India. The team at Netflix will clearly need to rethink their business model significantly if they want to start growing again. But pricing is just a small part of the puzzle when it comes to winning over international customers. The vast majority of Netflix content is still created for English speaking audiences. And that's a big problem.

If you speak English, you're talking to the minority. 20% of the world speaks English. 80% does not know what we're talking about.

NAS Daily has generated billions of views precisely because the content is designed to be internationalized from the start.

So every video we make on nas, every course we put on NAS Academy, every video we put on NAS Daily, we've built software to take it and put subtitles in a different language.

Independent content creators have already started capitalizing on international opportunities, but Netflix is falling behind. Sure, they have subtitles for most shows, but considering the incredible production value of every show they publish, it could be a lot better. And that's where that AI hacker I mentioned earlier comes in. His name is Rich Pavlovsky. And he's building a startup that automatically translates video content into different languages. That alone wouldn't be worth mentioning. After all, Squid game did have English dubs, even if they were terrible. But Rich is planning to take video translation several steps further, using cutting edge deep learning algorithms. Instead of just layering a voice recording over existing footage like most dubs do, Rich's company, Blanc, essentially uses deep fakes to realign the mouth movements to match what the new audio track sounds like. This solves that problem where the mouth movements don't match up and creates a much more natural and immersive experience. But how'd he get here? Rich has a fascinating backstory that motivates everything he's working on now. He grew up in Moscow, Russia, and overcame a remarkable set of obstacles to get to where he is today. Rich had a natural talent for math as a young kid, and his parents knew that he could go on to do great things if he just had the right resources. There was a problem, though. Higher level exam prep and English tutors aren't cheap, and his parents didn't want to see Rich fall behind. So they sold their family car to pay for his studies and and make sure that he had the opportunity to apply his skills. That gamble paid off, and Rich got accepted to go study math and physics at the University of Toronto. This was a huge break for Rich. He'd taken some Stanford machine learning courses online, but now he was in the heart of one of the best AI universities in the world. But while he was in college, he realized something that would permanently change his view of the world. Sure, Rich clearly had an innate affinity for solving technical problems, but if he hadn't learned English, he would have never been able to understand what was going on in those Stanford online classes. He'd gotten lucky, but for billions of people around the world, online education was still out of reach. He knew that he could change this, and AI was the perfect tool for the job.

Machine learning is actually, like, I would say, the closest thing to magic that we have in our life. It's like if you can imagine it, you probably can build it.

The goal was clear in his mind. Tons of valuable videos are already out there. They just need to be translated into every possible language. It was a cool idea, but he was still working towards his college degree, so it had to stay a side project. That was the plan, at least until economic factors threw a wrench in all of that. His parents could no longer afford to help him with tuition, so he had to leave university and get started immediately. But getting a startup off the ground is never easy, and Rich knew this. He'd built small side projects in the past, like a 3.3D printed prosthetic arm. But scaling video translation would be an entirely new challenge. He had to find an advisor who is not just an expert in machine learning, but also a successful entrepreneur. Luckily, he found one in Aidan Gomez. Aidan is the founder of Cohere AI, which is one of the fastest growing artificial intelligence companies in the world. Now, Aidan is clearly a talented entrepreneur. He's raised over $170 million to date. But more important importantly, he's a world class expert in deep learning and had been educating the world about the power of neural networks for years.

A neural network is just a program consisting of a set of neurons which are wired together. At one end of the network will feed in inputs. At the other end of the network we'll receive outputs.

Rich had assembled a talented team to help make his dream a reality. So now it was time to build. But getting a functional product out the door would be no easy task. In order to deliver fully automated translations with proper lip synching, Rich and his team would essentially need to build five systems at once for any video they received. They would first need to transcribe the audio so that they knew exactly what was being said. Then they'd need to translate that text into the target language. Once they had an accurate translation, they'd use machine learning to paraphrase the script so the words would match the original duration. See, some sentences just take longer to say when directly translated, so they'd have to account for this. Then, once they had a usable translation, they'd need to generate natural sounding speech. And lastly, they need to actually change the video frames so that the lips match up with the new audio. Here's an example of how that works in practice.

The translation is something I wanted to ask you about. And first, just to give a shout out to the supercomputer. La traducciones alcos sobre loqueria Preguntarte y primero sono paradarun saludo la super comprehensive.

Now, the results are clearly still a bit rough around the edges, but Rich and his team are making improvements every day. And most importantly, since they're using a deep learning approach, the system will get better as it's fed more and more data. With a working system in place, Rich and his team went into user acquisition mode. But no one is going to give money to a company without a name, so they had to come up with something that fit the company. They were building. They settled on the name Blanc after the famous voiceover artist Mel Blanc, who who had voiced Bugs Bunny, Daffy Duck and Porky Pig throughout the early 20th century. Hollywood has come a long way since Mel Blanc brought Looney Tunes to life with his voice acting skills. And it's about to get a lot more fun for everyone. We've seen visual effects artists create remarkably lifelike facial animations in films like Star wars and the Irishman. But this work is painstaking and time intensive. The results are amazing. But the traditional approach just won't scale to meet the volume of video content out there. Alexander Wang is the founder of Scale AI and has a really good rationale for thinking about when to apply machine learning models.

If you think about what tasks humans are good at versus what tasks algorithms are good at, generally that more or less boils down to data availability, which is that where there are large pools of digital data an algorithm can learn from and those pools of digital data either have been collected in the past or easy to collect in the future. Those are going to be the problems which algorithms can do effectively and can learn to do effectively and then areas where there does not exist digital data and is expensive to collect this digital data, those are going to be the last things to be automated.

And we're already seeing deepfakes have an impact on filmmaking. Visual effects studios are starting to use deepfakes to help make actors look younger or even bring them back from the dead. Rich is working on making those AI enabled superpowers available to everyone at ridiculously low cost.

Our mission is to solve language barrier worldwide. If you want to do this, you need to. You need your technology to be used by a lot of people. It needs to reach a lot of people. And the best way to do it is to sell to the platforms. We don't want to compete in terms of marketing.

His company is still just starting to scale and they aren't going to be able to deliver Hollywood level effects at the click of a button just yet. But by taking an AI first approach and and integrating into the large video platforms, they'll be able to translate millions of hours of video footage extremely quickly. So how does all of this affect Netflix? Well, they should clearly start investing in AI technologies to help further internationalize their content library. Every movie and TV show on Netflix should be available in any language with perfect voiceover and lip movements that actually match the dialogue. It's crystal clear that this technology is coming. The only question is if they can find a way to implement it. Reliably at scale. Implementing the technology that Rich and his team have built clearly won't be enough, though. They're going to have to figure out how to deal with people that share passwords, and they will probably need to launch a free tier that has advertisements. Getting households to upgrade their plans if they really want to share accounts with extended friends and family is pretty straightforward, but launching an ads business will be another challenge entirely. To be fair, the size of the prize here is massive. The global TV ad market is $141 billion and YouTube generated nearly $30 billion just last year. If Netflix gets it right, they could easily grow their revenue significantly. But Netflix has always had a premium brand and differentiated itself from traditional TV precisely because it didn't have ads. Even though Netflix would obviously still offer ad free streaming to paying customers, the mere existence of a free tier would make the whole experience feel different. Instead of paying for exclusive content, you're now just paying to avoid ads. And that isn't exactly in line with the brand they've spent so many years building. Developing the software required to deliver relevant ads anywhere in the world will also be a big engineering effort. So the risk here is that Netflix winds up only thinking of the short term and misses the importance of AI. They clearly have talented engineers on staff, but if they have their hands tied up on ads for the next few years, they might need to work with a startup to solve their translation problem, and that creates an opportunity. Netflix clearly has a long road ahead, but with a shrewd business strategy and the latest AI technology, I'm optimistic that they can turn it around. And as for Rich, his company is still thousands of times smaller than Netflix, so it's a real David and Goliath situation. But he doesn't actually have to defeat any big companies to be successful here. As long as he stays ahead of the curve, everyone will want to use Blanc for video translation. And if you want to hear another story of a talented founder using AI to build an incredible company, just go watch my video about how one of the co founders of Twitch built a self driving car company worth billions. Thanks for watching.