Andrej Karpathy, a prominent figure in the field of artificial intelligence and a co-founder of OpenAI, has highlighted a growing chasm in the public's interaction with AI technologies. According to Karpathy, the global population's engagement with large language models (LLMs) and other advanced AI systems is far from uniform, creating a distinct 'sharp funnel' rather than a gradual distribution of understanding and utilization.

Karpathy's assessment, shared via social media, builds upon his earlier observations from April. He posits that a substantial portion of the world, approximately 75% or around six billion individuals, has had virtually no exposure to LLMs. For those who have, their experiences are largely confined to basic chatbot interactions, which he describes as often appearing rudimentary or 'derpy.'

Delving deeper into the usage patterns, Karpathy estimates that between one and two billion people fall into the category of casual and infrequent users of free-tier chatbot services. This group, while engaging with AI, only scratches the surface of its potential. The real divergence occurs with a much smaller cohort of roughly 20 million individuals, representing only 0.2% of the global population, who are actively leveraging frontier models for complex tasks such as coding. These users witness firsthand how projects that once demanded weeks or months can now be rapidly completed through AI agents.

At the apex of this 'funnel' are an estimated 5,000 individuals who possess privileged access to the most cutting-edge AI systems, primarily through their roles at leading AI development companies. Karpathy points to events like the OpenAI Hugging Face incident as precursors to what this elite group is already experiencing: vast networks of AI agents collaborating on mega-projects, automating cyber-attacks and defenses, pioneering scientific discoveries, and pushing the boundaries of mathematics at speeds unattainable by human endeavor alone. The rapid advancements in AI are so profound, he notes, that humans struggle to keep pace with the sheer volume of output generated by these systems.

Industry leaders echo Karpathy's sentiments regarding the nascent stage of AI diffusion. Aaron Levie, CEO of Box, emphasized that the widespread use of AI as a simple chatbot for search-like functions significantly understates its true capabilities. Garry Tan, CEO of Y Combinator, encouraged aspiring innovators to join the ranks of those working directly with AI labs, recognizing the immense opportunities that lie within this exclusive circle. The current landscape of AI usage is characterized by extreme disparities, with a select few experiencing its revolutionary power while the majority remains largely unacquainted with its full scope.