John and I were discussing software failure modes and complexity issues over lunch, and speculation turned to just how tricky cybernetic interfaces would be to implement and debug. I noted how annoying it would be to get brain surgery every time an upgrade came out, and speculated that a good interface would be one that could grow organically (like the Tytan NN-II nerve net in Daniel Keys Moran’s books) to integrate with your brain instead of requiring difficult surgical intervention.
This then hooked up with Ray Kurzweil’s notion of one-neuron-at-a-time uploading: replace one neuron in your brain with an artificial one that is trained to function just like the organic ones and you’re still basically the same person; repeat 100 billion times and you could potentially upload your consciousness to a different substrate altogether. So if you already have an organically growing cybernetic interface, in addition to its interface work, you could also have it deploy new neurons to replace old ones that die off, one at a time. It would take decades to get to the point that you could actually upload, but one could accomplish a lot of living in that amount of time (and it provides a relatively natural decision point for “retiring to cyberspace”).
It then occurred to me that impatient transhumanists might become incredible party animals, indulging in huge amounts of neuron-killing behaviors to hasten the transition to being able to upload themselves. (This would probably increase the error rate, so sane people would follow a more temperate strategy.) So remember, kids, be careful at those parties with the radical extropians!