Rohit Krishnan challenges the idea that greater model capability automatically produces a straight path to catastrophe. He argues that current systems combine extraordinary performance on some formal problems with sharp practical limits, so claims about existential danger still need a concrete causal chain.
The discussion distinguishes long-range catastrophic scenarios from nearer-term disruption. AI agentsAn AI agent is a system that observes context, decides what to do, and takes actions through tools to pursue a goal. may increase claims, paperwork, cyberattacks and other automated pressure on slow institutions, but those institutions can adapt with automated triage, access controls and new defensive systemsAI-enabled cyberdefense uses AI systems to help detect, analyze, prioritize or respond to authorized security activity..
Rohit Krishnan compares multi-agent failuresA multi-agent system contains multiple AI agents that interact, coordinate, divide work, or influence one another while pursuing tasks. with financial flash crashes. MonitoringSecurity monitoring collects and analyzes system activity to detect suspicious behavior, control failures and emerging threats., circuit breakersA circuit breaker is a reliability control that temporarily blocks calls to a failing dependency after errors cross a defined threshold. and defensive agents could interrupt harmful behavior as it emerges, while markets, companies and governments provide imperfect but useful signals about whether damage is becoming systemic.
The final section considers public pessimism about AI. The conversation suggests that practical demonstrations of useful work, rather than abstract reassurance alone, may help younger people see AI as a source of agency while still taking its risks seriously.
Watch on YouTube
