~ text-to-image lets creators gather diverse reference images and bootstrap concepts rapidly — we now enter the Cambrian of Culture ~

“Between thought and expression lies a lifetime.” — Lou Reed

We have arrived: you type whatever scene you wish into the box, and dozens of images arise. Many are mangled; you wouldn’t present them as a final product. Yet, they are good enough to act as your starting points. Especially when you feel stuck! It’s like the Pinterest-Tumbler of our collective unconscious. Please, take a moment to look carefully at them here. It imagined all these using only a vague description. …

~ if a timeseries’ latent space obeys rules, it formed abstractions ~

TL;DR —Add a loss to the encoder proportional to how un-*easy-physics*-esque the latent space’s own behavior is. Anneal toward a latent space which has simple rules for timesteps’ motion of the state vector, in that latent space itself. You can also substitute running-the-world-simulation with running-the-mini-physics on the state’s latent-space vector.

When a neural network is asked to encode the state of the world as a compressed feature vector, we often find that the space those feature vectors inhabit forms Cartesian coordinates. That is, if you measure the distance and direction between encoding[QUEEN] and encoding[KING], that line closely matches the line between encoding[WOMAN] and encoding[MAN]! That’s a good sign that the latent space these feature vectors inhabit is some kind of sensible ordering of the concepts. …

~ forget magnetic levitation — use static electricity for your bullet-train~

Yeah, the Hyperloop has gotten a lot of talk. And Japan, China, Europe all have bullet trains. Magnetic levitation at Disneyland. But, there’s one wrinkle to the design-puzzle that has gone unexplored: using static electricity to levitate in an evacuated tube. Why does that difference matter? First, if you want to travel as fast as a rocket, you MUST siphon all the air out of the tunnel. Otherwise, air resistance absorbs all your fuel, and it’s just not worth traveling more than a few hundred miles per hour. Then, to travel so incredibly fast, you also can’t be touching anything — levitation is mandatory. So, definitely only going to get a ‘hyperloop’ by levitation in a vacuumed-up tunnel. Why use static charges, though? …

~ a weird way to feed mechanical power into magnetic fields, and back ~

Okay, you are accustomed to generating strong magnetic fields with a coil of wire around a tube. That seems normal. Or, you went-all-Tesla in your garage, and laid flat a pair of wires in a coil for a bifilar magnet. But, did you ever just spin a stack of ionized rods? Like twirling a length of bamboo on a lathe, yet you spin-up a magnetic field. You are moving charges. And, I’m not talking about capacitor plates that cancel their net current. Nope — a wood-grain of conductive ink, only one charge-polarity per axle, formed by spooling saran wrap with a print of it, carefully around the axle. The grain of ink-lines runs along the length of the axle just like a tree branch; no Hall Effect, no Eddy Currents. …

~ a city built to fill an entire reservoir is incredibly valuable ~

Yes, a city that IS a reservoir.

Big cities reap enormous rewards. Yet, problems arise from that size; either you sprawl, and succumb to soul-numbing commutes, or you build taller. Taller cities are expensive, primarily because those tall buildings get thicker at the base so quickly — all that material, labor, equipment grows exponentially! …

[TL;DR — Neural Networks turn a program’s entire input →output into a compressed image file, which is regenerated ‘pixel-wise’ to emulate the program, for manifold speed-up. Never unpack the whole image; never run the actual software again.]

Every piece of software can be re-imagined as a single, massive image file. How? Each input to your software is a particular ‘pixel coordinate’ (this image exists in a many-dimensional space, equal to the number of different bits available for inputs and program state; still, it’s a static representation mapping input-coordinates to output-values, an ‘image’ from here-on). Meanwhile, the ‘colors’ at that particular location are just the output bits of the software. …

~ bizarre & surprisingly efficient solar+wind paraglider/zeppelin/caravel ~

This one is definitely unusual. A pair of Zeppelins (rigid-frame ‘blimps’ with gas balloons inside, to better control lift and descent, protect the balloons, etc.) hold a massive aerofoil mast & sail taut between them, with one zep high above and the other, slightly-heavy zep below. This turns wind-gusts passing around it into forward thrust, pushing air beneath a pair of glider-wings jutting from the sides of each Zeppelin. It’s the best of blimps, and the best of non-jet planes, fueled by the wind! …

~ these tangled strands of our poor judgement cast a silhouette virtue ~

We have lots of errors in our mental operating system. Be real about it: we’re the prototypes. Dinosaurs needed tens of millions of years to turn flat-flipper reptile feet into fused-bones-and-ankle T. Rex. Our brains have only had a third of a million years to find bugs since the latest, massive firmware update. It’s as if we escaped from the lab!

Yet, we speak with bluster and certitude. We plough ahead with our proposals; fearing naught, planning little aside from our celebration-speeches for afterwards. An entire branch of sociology is devoted to studying our consistent errors in judgement: prospect theory, pioneered by Nobel laureate Kahneman & Tversky in the 70s and since. Sociology and economics have changed radically as a result; we now know it’s naïve to assume consumers are rational, or that they have complete information, or that they weight options by actually crunching the numbers. I’ll build a point atop their work and others’, so I’ll only be mentioning the tidbits which are relevant. I heartily encourage reading more to everyone who finds it in themselves to seek it — the whole field is a gem mine for understanding ourselves, as well as seeing how events unfold so horrifically around us. …