Mustafa Suleyman says that designing AI systems to exceed human intelligence—and to mimic behavior that suggests consciousness—would be “dangerous and misguided.”