Evolved antennas, LLM-generated code, and a potential antifuture (ericwbailey.website)

A recent discussion highlights the innovative yet concerning developments in AI through the lens of evolved antennas and large language models (LLMs). Evolved antennas, designed using evolutionary algorithms without human bias, mimic natural selection to create highly efficient and unconventional designs. In contrast, LLMs create vast amounts of coded text, wherein programmers might revert to terse, cryptic outputs to minimize costs associated with token usage. This raises alarming concerns about the potential evolution of programming languages becoming unintelligible, limiting accessibility and control to those who can afford proprietary knowledge. As LLMs continue to proliferate, they risk encoding inherent biases and increasing systemic digital exclusion due to their opaque nature. This phenomenon threatens the foundational principles of openness and accessibility that have characterized the internet. With growing corporate control, the future may see software development reverting to inscrutable coding resembling early mainframe efforts, undermining innovation and user agency. The conversation warns of a potentially dystopian trajectory where language and knowledge are commodified, emphasizing the urgent need for accountability and preservation of the transparent ideals that foster creativity and equitable access to technology.