Don't give up on carbon-based so easily - we have a solid lock on self-replication technology, and we're so cheap!

Develop your neural network in carbon over decades, then later "distil" its outputs into silicon-on-insulator for the long haul. We already have the technologies for that, we just don't call them by the right names.

What we don't have and should develop is the "technological toolkit" to colonise icy objects outside habitable zones. Probably needs to be fission-based since we cant seem to make fusion work. Building a society inside a KBO and putting it into an extrasolar trajectory is our only hope of teaching other stars without new physics.