Andrej Karpathy Warns of Growing Divide Among AI Users

Andrej Karpathy and His Vision for AI

Andrej Karpathy, former director of artificial intelligence at Tesla, recently shared his thoughts on the evolution of AI. On the platform X, he emphasized that artificial intelligence is particularly "sharp" when it comes to technical tasks. According to him, there is a deaf dialogue between advanced AI users and those who remain skeptical. This divergence is especially visible between free users of ChatGPT and those who leverage the latest models, each having a distinct perception of AI's capabilities.

AI and Its Technical Characteristics

AI stands out due to its technical characteristics, which can sometimes deter non-technical users. Karpathy observes that super-users, who master these technologies, and those who reject them, are gradually drifting apart. This divide could be explained by the complexity and specificity of the tasks that AI can accomplish, making its adoption more challenging for the general public.

Karpathy's Influence in the AI Field

Karpathy is often seen as an influential thinker in the AI domain. A former founding member of OpenAI, he is known for his in-depth analyses of trends in artificial intelligence. Among his notable contributions is the term "vibe coding," which has found its place in modern lexicon. In his recent writings, he discusses a "growing gap in understanding AI capabilities," dividing users into two distinct categories.

The Two Categories of Users

According to Karpathy, on one side, there is the everyday user who has tested the free version of ChatGPT, often limited to the earlier versions of AI chatbots. On the other side is the advanced user, who invests in the latest models for professional applications. The first group tends to mock the errors and hallucinations of AI, as illustrated by a viral video where OpenAI's voice mode responded inappropriately to a simple question.

Limitations of Free Models

Karpathy emphasizes that free and outdated models do not accurately represent the capabilities of the most advanced AI agents of the year. Paying users, on the other hand, leverage tools like Claude Code and Codex, placing significant importance on the capabilities of these technologies, particularly in programming, mathematics, and research. These models are particularly effective in technical fields, which are easier to train and potentially more profitable.

Limited Progress in Certain Areas

Karpathy notes that typical queries related to research, writing, or advice have not seen the most spectacular advancements in terms of capabilities. This could explain why some users remain skeptical about the usefulness of AI in these areas.

The Impact of Interactions with Recent Models

A friend of Karpathy suggested that interest in moments like OpenClaw has grown because it was often the first interaction for many users with more recent models. This underscores the importance of exposure to advanced technologies in changing perceptions.

Challenges of AI Adoption

More broadly, there exists a gap in AI adoption within the general population. Sam Altman, CEO of OpenAI, acknowledged that AI is not yet very popular, citing concerns about data centers and the use of technology by some CEOs to justify layoffs. This situation sometimes creates tensions between ardent AI supporters and those who avoid it, a dynamic that Business Insider has referred to as a "Claude-gap relationship."

Future Perspectives

As more people begin to experiment with tools like Claude Code, OpenAI's Codex, or Cursor, the situation could evolve. For now, Karpathy observes that the two user groups continue to "talk past each other," illustrating the ongoing challenge of integrating AI into society.