safetyopinionbearishWe are nowhere close to solving critical AI alignment issuesGary Marcus26 Jul 2026https://x.com/GaryMarcus/status/2081108158815342638