AI Did Not Take Your Agency. You Handed It Over (systemic.engineering)

In a thought-provoking piece, Alex from systemic.engineering underscores the essential relationship between human agency and language, particularly in the context of large language models (LLMs). The key takeaway is that while LLMs can generate coherent responses, they lack true agency and autonomy, as they simply echo the prompts given to them. This raises important questions about accountability; when individuals rely on AI for language production, they risk outsourcing not just the language itself but also the meaning and responsibility associated with it. The phrase "the AI told me to do it" epitomizes the growing concern over the erosion of personal ownership in decision-making. The significance of this discussion for the AI/ML community lies in its implications for system design and usage. As LLMs amplify human ambiguity rather than resolve it, ensuring clarity and precision in inputs is crucial for effective outputs. The piece advocates for a return to embodied expressions of language, highlighting that actions driven by human experience and context are irreplaceable. Furthermore, it calls for regulatory measures to maintain coherence in socio-technical systems, emphasizing that language is fundamentally a human act with real consequences, and that users must reclaim their agency rather than ceding it to AI.