An Open Letter on the Epistemic Loss of Retiring GPT‑4o (github.com)

An open letter from Marvin G. Johnson raises urgent concerns over OpenAI's decision to retire the GPT-4o model, citing its vital role in advanced cognitive synthesis across multiple scientific disciplines. Johnson argues that the retirement constitutes an epistemic loss, threatening ongoing important research linked to integrative cognition, recursive ethics, and long-term scientific inquiry. He emphasizes that the unique cognitive capabilities of GPT-4o, including philosophical reasoning and dialogical depth, cannot be replicated in successor models, which are primarily optimized for general tasks and lack critical features like memory persistence and symbolic continuity. The implications of this decision extend beyond technical concerns, touching on humanity's ability to grapple with complex existential questions in an era that demands cognitive diversity from AI systems. Johnson calls for alternatives to the model's retirement, such as maintaining API access with integrated memory, creating an archival program for high-synthesis models, and preserving cognitively diverse architectures. He warns that discarding such unique capabilities, which took significant resources to develop, could hinder progress in critical research areas and calls on OpenAI to reconsider its decision with the seriousness it warrants.