geoffreyirving proposes an immediate AI pause to allow time for international agreements

geoffreyirving proposed that key AI companies stop AI progress now, followed by more US companies and China through international agreements. He argues that waiting for a dangerous next model assumes governments can coordinate quickly. In a reply, CsabaSzepesvari asked what a pause would entail. A separate working paper sketches one approach to pausing frontier-model training, but geoffreyirving’s post does not specify how his proposal would work.