If dtype is not specified, NumPy defaults to float (depends on implementation/platform but usually float64).