Sapere Aude
Month: March 2019
Utah Just Became a Leader in Digital Privacy
Utah legislators voted unanimously to pass landmark legislation in support of a new privacy law that will protect private electronic data stored with third parties like Google or Facebook from free-range government access. The bill stipulates that law enforcement will be required to obtain a warrant before accessing “certain electronic information or data.”
It took me 2 days to get everything extracted, upscaled and put it back together in a way that was pleasing. This resulted only in the first 5 minutes of the episode being done (the episode recap, the opening scene, and the intro). Still pretty good time for a mid-to-high end PC with software that isn’t just available to professionals. The result left me pretty awestruck. It looked better than I had hoped. No weird issues or anything. It looked pretty much like an HD version of DS9. Since (moving) pictures are worth more than 1000 words, here are 2 comparison videos that show off the improvement I was able to get with this machine learning based upscaling technique. Imagine what a real team could do, with more powerful equipment, custom trained neural networks … and access to the original SD files instead of a DVDRip like me.
Swearengen: You ever think, Bullock, of not going straight at a thing? … Bullock: No.
Despite its advanced age, the average American plant has a generating capacity—a measure of the percentage of time a reactor is producing energy—of more than 90%. Plants abroad have an average generating capacity of 75%. “In terms of the ability to reliably generate electricity and safely generate electricity, the US fleet still sets the standard for performance. The place that the US is falling behind in is in the ability to build a new plant at schedule and at a low cost. A combination of solar, wind, and nuclear led to decarbonization at the lowest cost. New York, New Jersey, and Illinois have introduced subsidies to keep their nuclear power plants profitable and running, rather than trying to meet energy demands with renewables alone. But keeping America’s current fleet running is not enough to decarbonize the electricity sector.”
it wouldn’t need saving if people weren’t so dumb and mood-affiliated.
Translating Predictive Coding Into Perceptual Control
some people (often including me) find free energy/predictive coding very difficult to understand, but find perceptual control theory intuitive. If these are basically the same, then someone who wants to understand free energy can learn perceptual control theory and then a glossary of which concepts match to each other, and save themselves the grief of trying to learn free energy/predictive coding just by reading Friston directly.
On Starship progress:
If SpaceX hit the targets for the Starhopper and the Starship then SpaceX will have accelerated rocket development by 4x. This would be 10-20x faster than most of their competition. SpaceX continues to get more ambitious with its rockets and is accelerating its rate of progress. This faster rate of development will mean that the world will get the space program that we have always wanted.
2019-09-09: What the future may hold for starship:
The next 7 years could see space habitation increase 100s of times. We can go from 6 people in space to 100s in a larger rotating one gravity space station and a lunar mining base.
2019-10-15: What moon missions and bases look like:
SpaceX will leave most Starships on Mars or the Moon, when they are flown for long-range missions. SpaceX will need to use ~5 launches of Super Heavy Starships to fully fuel a Super Heavy Starship in orbit. They will then send a fully fueled Super Heavy Starship to Mars. The Super Heavy booster will separate around the orbit of the moon. The booster will return and 37 Raptor engines will be reused. The 6 Raptor engines in the Starship will take the Starship to Mars with ~100 tons of payload.
Starship Moon mission needs a 10-ton lunar lander
2019-11-05:
Robert Zubrin indicates there is a need to stage the SpaceX Starship from low earth orbit or injection orbits for the moon and Mars. Missions to the moon would be far more efficient with a 10-ton lunar lander. This could be a mini-starship.
2021-03-03: History!!! To celebrate, let’s cancel SLS
2021-05-13: Robert Zubrin on the profound potential that Starship represents
Starship won’t just give us the ability to send human explorers to Mars, the moon, and other destinations in the inner solar system, it offers us a 100x increase in overall operational capability to do pretty much anything we want to do in space. That includes not only supporting a muscular program of probes to the outer solar system, and making all sorts of experimental investigations in Earth orbit economical, but enabling the construction of giant space telescopes. Much of our knowledge of physics has come from astronomy. This is so because the universe is the biggest and best lab there is. There is no better place to do astronomy than space. The 2.4-meter Hubble Space Telescope has made extraordinary discoveries. What might we learn once we are able to build 2.4-kilometer telescopes in deep space? The possibilities are literally inconceivable.
2021-10-12: a FT take focusing on competitors complaining, instead of scrapping their obsolete approaches:
SpaceX’s vertically integrated manufacturing approach will also deprive other US suppliers of business, weakening the wider industrial base the country had built up to support its long-term ambitions in space, Amazon and others warn. However, SpaceX’s customers — including those in government — do not seem to share the misgivings. “Before SpaceX we only really had the ULA, so we’re in a better position than we were,” says Phil McAlister, director of Nasa’s commercial space flight division. Diamandis goes further: “The US government is lucky to have a company like SpaceX based here,” he says, since its efficiencies feed through directly into the US space programme. And companies that compete with SpaceX in some markets seem more than happy to use its launch services, despite supporting a rival.
2021-10-30: Another interesting take that argues that Starship will change the entire space industry away from “reduce weight at all cost” towards rapidly producing space worthy hardware at scale:
Starship obliterates the mass constraint and every last vestige of cultural baggage that constraint has gouged into the minds of spacecraft designers. There are still constraints, as always, but their design consequences are, at present, completely unexplored. We need a team of economists to rederive the relative elasticities of various design choices and boil them down to a new set of design heuristics for space system production oriented towards maximizing volume of production. Or, more generally, maximizing some robust utility function assuming saturation of Starship launch capacity. A dollar spent on mass optimization no longer buys a dollar saved on launch cost. It buys nothing. It is time to raise the scope of our ambition and think much bigger. Prior to Starship, heavy machinery for building a Moon base could only come from NASA, because only NASA has the expertise to build a rocket propelled titanium Moon tractor for $1b per unit. After Starship, Caterpillar or Deere or Kamaz can space qualify their existing commodity products with very minimal changes and operate them in space. In all seriousness, some huge Caterpillar mining truck is already extremely rugged and mechanically reliable. McMaster-Carr already stocks 1000s of parts that will work in mines, on oil rigs, and any number of other horrendously corrosive, warranty voiding environments compared to which the vacuum of space is delightfully benign. A space-adapted tractor needs better paint, a vacuum compatible hydraulic power source, vacuum-rated bearings, lubricants, wire insulation, and a redundant remote control sensor kit. I can see NASA partnering with industry to produce and test these parts, but that is no way to service the institutional overhead embodied by a team of 100s of people toiling on a single mission for 10 years. There is a reason that JPL’s business depends on a steady stream of directed flagship missions with $1b price tags. Hordes of PhDs don’t come cheap and need a lot of care and feeding.
2021-11-05: The new 10 year NASA research plan doesn’t yet take vastly better costs into account. The telescopes in particular need to be rethought completely. Perhaps a combination of HavEx and LUVOIR, redesigned to be 10x cheaper, would do the trick.
2 current NASA mission concepts, HabEx (Habitable Exoplanet Imaging Mission) and LUVOIR (Large UV/Optical/IR Surveyor), are aimed at pulling this off. Both would use large, extremely clear mirrored optical telescopes, UV rays, and infrared to hunt for exoplanets with signs of water, oxygen, and ozone. HabEx would use a “starshade” to block out light from stars to reveal the planets surrounding them; LUVOIR would use a very large system of unfolding mirrors. A blend of the 2, though—now that might be just right for a mission that “combines a large, stable telescope with an advanced coronagraph intended to block the light of bright stars,” as the survey states, and is “capable of surveying a 100 or more nearby Sun-like stars to discover their planetary systems and determine their orbits and basic properties. Then for the most exciting ~25 planets, astronomers will use spectroscopy at ultraviolet, visible, and near- infrared wavelengths to identify multiple atmospheric components that could serve as biomarkers.”
At Haenyeo, a new corner restaurant in Park Slope, the menu description “saucy and spicy rice cake fundido” is intriguing, but it cannot entirely prepare a diner for the dish that will arrive: a deeply flavored bowl of ingredients topped with a bubbling sheath of cheese. Somehow, this shared appetizer manages to be Korean, Mexican, Spanish, Swiss, and maybe even a little Japanese, all at once.
I made a little flow chart of mainstream programming languages and how programmers seem to move from one to another.
I previously tested the out-of-the-box version of GPT-2 and couldn’t make it understand rhyme and meter. I wrongly assumed this was a fundamental limitation: “obviously something that has never heard sound can’t derive these complex rhythms just from meaningless strings of letters.” I was wrong; it just didn’t have enough training data. Gwern’s retrained version gets both of these right, and more too. For example:
Thou know’st how Menoetiades the swift
Was dragged, of Hector and the fierce compeers
And Phrygian warriors. So, we will dispatch
Your bodies, then, yourselves to burn the ships
In sacrifice; with torches and with bells
To burn them, and with oxen to replace
Your gallant friends for ever. But I wish
That no man living has so long endured
The onset of his foes, as I have power
To burn or storm; for mighty Hector erst
Was slain, and now returns his safe returnThis is all perfect iambic pentameter. I know AP English students who can’t write iambic pentameter as competently as this.
Does reality drive straight lines on graphs, or do straight lines on graphs drive reality?
Here’s a graph of US air pollution over time:
Some people pointed out that this showed the Clean Air Act didn’t matter. The trend is the same before the Act as after it. This kind of argument is common. If we wanted to be really harsh, we could make a graph like this:
But the same argument that disproves the importance of photolithography disproves the importance of anything else. We’d have to retreat to a 1000-coin-flips model where each factor is so small that it happening or not happening at any given time doesn’t change the graph in a visible way.
Gregor J. Rothfuss 