[anigbrowl]

I wonder if part of the problem is that computation is often excessively focused on precision, when what you often need is a good approximation. Precision is very valuable in the right context, of course, but in the wrong context it's like bringing a scalpel to a sword fight.

[colorint]

Computing must be focused on precision, because you need exact addressing to have data structures and branches. You could maybe let data structures slide, but if you don't have branching (or something equivalent) then you don't have "universal" computation.

[anigbrowl]

I don't agree at all. Of course you want to be able to do things precisely but if that's the only way you can do things then your system is too fragile.

There are times when it's very important to distinguish between 0.999998 and 1, and times when it's very important to treat the difference as negligible.

Have you ever played with an analog computer? I think you'll find it interesting. Modular synthesizers are essentially lego for analog computation, and you can find completely analog sequencers and so on. It takes some time to get the hang of this, but learning to build self-sustaining chaotic attractors is both enjoyable and interesting. The demo version of teh software here will get you started if you are not near a place with a decent music store: http://www.nordkeyboards.com/downloads/legacy/nord-modular-g...

Your brain is not a perfectly precise or reliable organ, but it's taken you this far.

[colorint]

This is days old, but I figured I'd mention, what I mean isn't the difference between 0.99999998 and 1, but you need exact addressing. That is, a program always has to branch to the right place, i.e., pointers must be exact; this is what gives digital computers the appearance of acting like finite state automata. The theoretical formulation of so-called universal computation is based on discrete state (be it state machines in Turing's version or "pure" symbols in Church's), and the only way to simulate that with analog computers (i.e., the electronics in the actual device you're using right now) is with high precision.

Analog computation is definitely interesting, but at least in terms of theory and standard practice, it's worlds apart from digital computation. Put another way, though this may be even more obscure, the problem with analog computers is that they're application-specific. "Digital" computers are still analog computers under the hood, and their application is simulating set theory (so the "universality" of computation is contingent on set theory being universal). My point here, really, is that, if you make that not the case, then you're just saying, what if we didn't use digital computers at all? An interesting enough question, but then you really do have to study analog computers, and whatever you come up with will be totally unrecognizable to current computer science.

In any case, my original point is that brains can't be formal systems, as a comment on the supposed "rational brain." Formal systems don't work out in practice for a lot of reasons, and many of them are exposed by the actual experience of computer science. Returning to analog computation would just mean agreeing with me.

[08-15]

> Your brain is not a perfectly precise or reliable organ, but it's taken you this far.

...and it took only four billion years of evolution. (Or maybe 100 million years of mammalian evolution, or 20 of primate evolution, or 200 thousand years of human evolution. Whatever you pick, it's slow.)