Well said! That’d be full on meta-cocoNuts … in a latent space backtracking kind of way (12/11/24: https://arxiv.org/abs/2412.06769)! 8^b

Nowadays though, it seems we’ve acquired an expensive taste for doing exactly this, “automating” tasks that the brain-&-body do naturally, organically, effortlessly, and efficiently. But devices that attempt to replicate and possibly amplify these abilities, in roundabout ways, forcing such organic processes through the needle eyes of strict arithmetic and logic units (originally designed as aids to help our “fuzzy” brains perform painful correct and precise computations when needed), result in artifacts that are wildly expensive, inefficient, enormous, and that fail in new ways that hinder their potential for domestication. Paradoxically, the derived software implements further end up sucking in both arithmetic and logic.

We’ll probably end up in a better place by focusing our verbal comms and arts efforts (genAI ANN Transformers) on analog-space neuromorphic devices instead of ALUs and their derivatives ( https://www.nextplatform.com/2024/04/24/sandia-pushes-the-neuromorphic-ai-envelope-with-hala-point-supercomputer/ ). And for activities that challenge our brains and require rigor and high-precision, including logical reasoning, mathematical computation, chess, and Jeopardy, that we’ve worked on for millenia, ALUs, Deep Blue expert systems and chess chips, Watson’s Prolog, and the likes, are definitely the tools to use. It’s different strokes of the spurs for different horseback riding folks — not the same in countryside trot, as in rodeo …

I wouldn’t saddle-up an abacus to do verbal comms either!