Wayne Liu Warns of AI Hangover and Token Paradox

Chief Growth Officer and President of Perfect Corp America Wayne Liu warns of an AI hangover, describing a disconnect between the low cost of generating AI content and the high cost of the human judgment required to validate it. Liu argues that generative AI has shifted the professional burden from creation to supervision, forcing humans to review an increasing abundance of machine-generated reports and proposals.

Liu identifies a token paradox where decreasing unit costs for AI inference lead to higher overall consumption and hidden compute costs, a trend exacerbated by the rise of AI agents. He suggests that the industry is reaching a point where the volume of output creates a bottleneck in human review.

To address these issues, Liu proposes three operational disciplines: measuring the complete cost of an outcome including human review, directly tracking the human review burden, and implementing strict budgets and stopping rules for AI workflows. He asserts that these measures are necessary to ensure that intelligence improves judgment rather than increasing operational inflation.