As AI agents become reliable for complex, multi-step tasks, the critical human role will shift from execution to verification. New jobs will emerge focused on overseeing agent processes, analyzing their chain-of-thought, and validating their outputs for accuracy and quality.

AI is not a 'set and forget' solution. An agent's effectiveness directly correlates with the amount of time humans invest in training, iteration, and providing fresh context. Performance will ebb and flow with human oversight, with the best results coming from consistent, hands-on management.

The frontier of AI training is moving beyond humans ranking model outputs (RLHF). Now, high-skilled experts create detailed success criteria (like rubrics or unit tests), which an AI then uses to provide feedback to the main model at scale, a process called RLAIF.

AI models lack access to the rich, contextual signals from physical, real-world interactions. Humans will remain essential because their job is to participate in this world, gather unique context from experiences like customer conversations, and feed it into AI systems, which cannot glean it on their own.

If AI were perfect, it would simply replace tasks. Because it is imperfect and requires nuanced interaction, it creates demand for skilled professionals who can prompt, verify, and creatively apply it. This turns AI's limitations into a tool that requires and rewards human proficiency.

Once models reach human-level performance via supervised learning, they hit a ceiling. The next step to achieve superhuman capabilities is moving to a Reinforcement Learning from Human Feedback (RLHF) paradigm, where humans provide preference rankings ("this is better") rather than creating ground-truth labels from scratch.

It's a common misconception that advancing AI reduces the need for human input. In reality, the probabilistic nature of AI demands increased human interaction and tighter collaboration among product, design, and engineering teams to align goals and navigate uncertainty.

Karpathy argues against the hype of an imminent "year of agents." He believes that while impressive, current AI agents have significant cognitive deficits. Achieving the reliability of a human intern will require a decade of sustained research to solve fundamental problems like continual learning and multimodality.

AI models have absorbed the internet's general knowledge, so the new bottleneck is correcting complex, domain-specific reasoning. This creates a market for specialists (e.g., physicists, accountants) to provide 'post-training' human feedback on subtle errors.