Not quite. There's plenty of examples in nature of automatons executing simple behaviors without an overseer that lead to tremendous complexity and efficiency.

However, we only see the successful emergent phenomenon, not the failed ones. Also, to this day we can not make accurate predictions of inherently chaotic systems.

So, while a bottom up sysytem may optimize, exactly whatt it optimizes is impossible to discern.