The article, "You Exist in the Long Context," presents an interesting vision of the future of AI that goes beyond simple question answering. In the flood of articles and opinions about LLMs and AI, this one stands out in many ways, particularly due to its compelling predictions about how AI might reshape various aspects of business and society. Here are some of the predictions and speculations I found most intriguing that potentially have significant implications on the economics and business opportunities around AI models with long context.
Human expertise and wisdom encapsulated in an AI as a paid service: As AI evolves, access to AI-powered "wisdom" will become a valuable commodity that people will be willing to pay for. This could revolutionize how experts share their knowledge, creating a new revenue stream for those who currently make a living from books, lectures, and other platforms. Imagine being able to tap into the collective knowledge of the world’s leading economists, historians, or scientists, all through a simple AI interface. This goes beyond the current role-playing prompting techniques. While role-playing relies on the general understanding of the said role from the model's training data, what long context makes possible is a closed context with curated information from the experts themselves. This is essentially the Masterclass revenue model, where experts make money sharing their knowledge, and like the author says in the article, "the AI is not a replacement for your hard-earned expertise; it’s a new distribution medium." On a more personal level, there's also potential for an alternate possibility this opens up, capturing not a person's expertise or knowledge, but them as a whole in spirit, for a sense of presence after they are gone, to fulfill the deep yearning we feel for our lost loved ones. From that perspective, the title of the article could not be more apt.
Long-context AI in organization decision-making: The long-context AI models will soon become an integral part of decision-making processes in all kinds of organizations. They would be able to access and process vast amounts of organizational data, including internal documents, historical records, and market trends. This would enable them to provide insightful feedback on plans, identify recurring patterns, and conduct realistic scenario planning exercises, ultimately leading to better-informed and more strategic decisions. This prediction reminds me of the article on using NotebookLM as a CRM killer where a business dumped decades worth of customer emails and interactions into the model context and used it to make strategic marketing decisions!
New techniques and jobs around content curation: Perhaps the most insightful prediction is that the success of AI in an organization will depend not solely on the model's power, but also on the quality and relevance of the information it’s given access to. This highlights the importance of thoughtful data curation and the role of human expertise in shaping the AI's understanding. It suggests a future where organizations invest in strategies to curate and refine their data to maximize the value they get from their AI systems. This becomes in part necessary also because of the deluge of data that comes with the ease of content creation, also using AI. Rather than largely structured data that we use today, we'll move to a world where unstructured data will rule. Audio recordings, video snippets, images will be thrown at the AI to make sense of the world. It's one of those things where technology introduces a problem we don't have, and creates a job market around solving it.
What's evident though, from all of the above is that just as electricity and compute have become commodities, smartness will become one, too. Everyone will have access to a smart assistant, to organize, to perform the menial activities, build things, or make strategic decisions, which begs two critical questions - *What would people do with their time? And just as importantly, how will they earn money?