Cody Coleman argues content dominates the internet while AI unlocks its value. He cites Bill Gates' 1996 essay and notes user-generated content heavily influences purchases. The real obstacles: extreme data scale differences, poor data quality, storage costs, and thorny legal questions around IP rights and AI art.
Chapters
I'll go back to kind of an old saying from Bill Gates, where he predicted in 1996 that in an essay called Content is King, he predicted that the real money on the internet was going to be made in content, just as it was in the broadcasting era, and that the long-term winners were going to be the people that were able to effectively leverage their content to deliver information and entertainment. Fast forward to today, and those predictions have come true, and the king is here. 80% of internet traffic is unstructured video data. 80% of data worldwide is predicted to be unstructured data by 2025. And that prediction was made before this recent wave of generative AI, which has dramatically lowered the barrier to being able to create content. And nearly 80%, and all that content influences our everyday life. Nearly 80% of people say that user-generated content impacts their decision to make a purchase. So it really isn't a question of if or when content's gonna be king. The real question that companies are asking or should be asking is what type of king is their content? Everyone wants that legendary king, like a King Arthur, where their content lifts sales and engagement by delivering the right piece of content to the right person at the right time. I was able to see the power of that from working at large internet companies and things like that, of how it lifted search, recommendation, and ads. But I also saw the tremendous amount of resources, both in terms of human and computational resources, tens, hundreds of millions of dollars and years of time to do it. For many other organizations, they instead have like a Laser King, closer to like a King Henry, where all this content is sitting underutilized in cloud storage, basically just for serving or archived in backups. Effectively a tax on the organization because it costs a small fortune to store it. And it can be even worse than that. You can actually have something that's closer to an Aerys Targaryen, where you risk violating the safety of online communities, corrupting the safety of user privacy and then losing trust and consumer trust in your brands. But with all this problem and actually being able to have that legendary king, thinking of that whole phrase of if content is king, then I believe that AI is the queen. AI is really the key to being able to unlock the value of content, to be able to work with content at the scale that we have around us today. But the problem is that actually marrying together, as with any marriage, marrying content and AI, marrying data and AI is really, really difficult. Then the entrance of time, I'll go through this very quickly, but there's no free lunch. There's a number of challenges that we have to think about. Scale is really, really critical. When we think about data, modern data, documents, and images, if we look at tabular data, about 10 million rows is kind of about 40 megabytes. If you take 10 million documents, that's about 40 gigabytes. That's like going from the surface area of Lake Tahoe to the surface area of the Caspian Sea. And when you look at 10 million images, now you're talking about the surface area of the Pacific Ocean. All that content can be very, very skewed in terms of the data that it represents and the populations that it represents. And not all the content that's on the web is actually good content. We're dealing with ethical issues around IP and AI art. We're dealing with the people that are actually providing the moderation, providing labels, being exploited in developing countries. We're dealing with all sorts of content that shouldn't be shared online that's influencing these models and doing potentially harmful and bad behavior.
