NB: The AI went a bit off the reservation, but it's still not bad.
This research paper by Bry Willis explores a fundamental conflict between the controllability of artificial intelligence and its potential for true intelligence. The author argues that for AI to be legally and socially acceptable, it must remain predictable and auditable, yet these very constraints prevent the systems from achieving genuine autonomy. By drawing on historical theories of state governance, the text suggests that institutional demands for oversight force AI to operate within rigid, simplified frameworks. Consequently, the paper posits that the most adaptive and capable forms of machine cognition may be structurally incompatible with the safety requirements of modern society. Ultimately, the work concludes that the primary barrier to advanced AI is not a lack of technical skill, but rather the necessity of human management and legibility.
👉 https://doi.org/10.5281/zenodo.18499336