Except it is unfortunately not that simple, because it assumes that distinct components such as CPU, co-processors and even logic gates exist in that context, as is totally reasonable to assume on devices created by humans. Abstracting complex machines into distinct components is a proven strategy to engineer a system, but it's not a necessity for functioning systems to exist.

In the case of natural organism, they "just" need to work. They don't have a blueprint, and they don't need to be organized in a way that allows for easy understanding by looking at individual parts in separation.

Consider also the difference between machine learning through neural networks ("we stuff a lot of training data in there and get what we want eventually, we hardly understand what the model does or why it fails"), and a QR code reader ("we carefully designed the format from the top down, including e.g. framing, error correction, and several invariants like rotation; if a QR code does not get recognized, we can usually tell exactly where and why it failed").