New working paper out today on Zenodo.
AI safety isn't just a model problem. It's a sociotechnical one. This paper examines what happens when we design systems around assumptions of perfect users and perfect systems.
@jaceblog.bsky.socialExplore
@jaceblog.bsky.social