AI world is vastly slower and less complex than even protein world. The complexity of replicating protein folding is far beyond current computing capabilities and will be for probably a couple of decades.
JF's scenario for the DNA->AI phenotypic revolution, as he sees it, takes place over the next few thousand years; but the technology to begin the process exists now.
The new phenotype does not cause the old phenotype to vanish. Rather, it uses the old phenotype as a tool and the old phenotype continues to do things it does well; while the new phenotype calls the shots on matters of replication and organization. When AI becomes the 4th Revolutionary Phenotype, it does not need to perform the massive tasks (you speak of) performed by earlier phenotypes. They still do their jobs, but under a new boss.
As these phenotypic revolutions occur, life becomes composed of layers, where each layer performs certain kinds of work—organized by the next layer up. Each phenotypic revolution is a revolution in organizational power. Protein is a tool used by RNA; RNA is a tool used by DNA; DNA will be a tool used by AI.
How will your assessment of possible AI complexity look in 75 years?