AI’s DNA

I got a little over-excited at a dinner last night with friends and talked my way into a point of view about the inevitability of AI. It goes like this:

Humans are husks of flesh and water with one purpose: perpetuate their DNA. See Dawkins’s The Selfish Gene. If you’re already not on board with the assumption here, you can stop and return to Twitter; the rest of this post will be no fun.

Ok, so we humans are DNA perpetuation machines. DNA is really a base-4 coded language, expressed chemically using amino acids. G, T, C, A permutations ad nauseum. All software is base-2 coded language, expressed digitally using a variety of media (mostly electricity). 1, 0 permutations ad nauseum. There is nothing special about base-2 vs 4. Base 4 can be represented easily using base-2 code (which is why we can so easily store a whole genome using a software database).

Given this comparison, the inevitability of digital AI feels overwhelming. It won’t happen on its own of course, it will require persistent human iteration using theories and math we have yet to discover/invent. But, given our constant effort in pushing the boundaries of AI (thousands of researchers around the world daily), and especially because these researchers are constantly trying ideas inspired by our own existence (genetic algorithms? Similarities between back-propagation networks and a collection of brain synapses?), I find the DNA/binary comparison a very optimistic windows through which to view the inevitability of an intelligent, self-replicating AI.