AI safety expert Connor Leahy warns that superintelligent AI could arrive sooner than expected, urging stronger safeguards to prevent unpredictable and potentially dangerous outcomes.