You know what really grinds my gears?
Working with people who don't understand how a neural computer works.
Be it some mass ratio optimizing payload engineer, a logistics officer frustrated with the difficulties caused by our team's solutions or just our boss looking for reasons to fire us because they thought our initial cost estimate was "unrealistically high" and are now sorely disappointed at reality, these people are miserable to deal with. On the surface, their complaints make sense; we are seemingly doing a much worse job than everyone else is and anything we come up with creates lots of problems for them. Satisfying all their demands, however, is impossible. With this post I intend to educate my audience on
so that my blog's engineer-heavy audience may understand the inevitable troubles those in my field seemingly summon out of thin air and so that you people will hopefully not bother us quite as much anymore.
First of all, neural matter is extremely resource heavy. Not by mass, mind you; a BNC of 2 kilograms requires only a few dozen grams of whatever standardized or specialized mix of sustenance is preferred in a single martian day. (I'm not going to bother converting that.) The inconvenient part is the sheer variety in the things they need and the waste products they create.
This is just a shortened list, but already it causes problems. If you want to create a self contained system to avoid having to refuel constantly, you will need a lot of mass and a lot of complexity. This is what a typical sustenance diagram for such a system looks like:
(Keep in mind, this diagram doesn't even have electricity drawn in.)
Typically these systems are even more complicated, with redundancies and extra steps. In any case, this is complicated, energy expensive and a nightmare to maintenance crew. I mean, just keeping the bacterial microbiome alive is a lot of effort!
Second of all, neural matter is extremely vulnerable. Most power plant and rocket designers just round away all temperature changes less than 100 K, but neural matter will outright die if its temperature is just a few kelvin off of the typical value. The same goes for a lot of other things - you'll need some serious temperature regulation, shock absorption, radiation shielding (damn it I wish we had access to the same stuff as those madmen in the JMR) and on top of all of that, you need to consider mental instability!
That last one is kind of the biggest pain in the ass for these things - we need to give them a damn game to play whenever they don't have any real work to deal with or they degrade and start to go insane. (Don't worry, I'm not stupid, I know these things aren't actually sentient, I'm just saying that to illustrate the way they work.) It can't even be the same game - you need to design one based on what the NC is designed to do! (Game is a misleading term by the way; it's not like a traditional video game. No graphics - just a set of variables, functions and parameters on a simple circuit board that the NC can influence.)
And lastly, neural computers are complicated. Dear Olympus are they complicated. There are so so many ways to build them, and the process of deriving which one to use is extremely difficult. You can't blame the NC team for an inappropriate computer if the damn specifications keep changing every week!
There's the always-on, calculation-heavy, simple and slow Pennington circuits, the iconic Gobbs cycle (Bloody love that thing!), the Anesuki thinknet and its derivatives, the Klenowicz for those insane venusians and so so many more frameworks for both ANCs and BNCs. Oh yeah, by the way, the acronyms ANC and BNC actually don't stand for Advanced and Basic Neural Computer respectively. They stand for Type A Neural Computer and Type B Neural Computer. It comes from that revolutionary paper written by Anesuki.