Link to First Installment: https://chrisprophet.substack.com/p/spacex-evolution
Startup values
Arguably a company has the greatest creative efficacy during the startup phase, then gradually lose this creative edge as it grows; typically through adding layers of bureaucracy, diffusing ownership and diversifying corporate goals. However, Elon Musk enterprises maintain a high degree of development efficiency by eliminating bureaucracy wherever possible, for example by allowing their personnel near autonomy in their work. This disregard for paperwork can be quite shocking for new hires, at least according to Gwynne Shotwell who joined SpaceX as their vice president of sales back in 2002…
“Accepting Musk’s job offer liberated Shotwell from the constraints of a more traditional aerospace company. During her first day at work, she set about formulating a strategy to sell the Falcon 1 rocket to both the US government as well as small satellite customers. Seated in the cubicle farm at 1310 East Grand, in El Segundo, Shotwell wrote a plan of action for sales. Musk took one look at it and told her that he did not care about plans. Just get on with the job. ‘I was like, Oh, OK, this is refreshing. I don’t have to write up a damn plan,’ Shotwell recalled. Here was her first real taste of Musk’s management style. Don’t talk about doing things, just do things.”1 ~ Wired
Senior management generally set high level requirements such as: build a reusable booster or a new launch vehicle then loose their design engineers on the problem. This approach coupled with their own Enterprise Resource Planning (ERP) software, called WarpDrive, allows even large concerns like SpaceX and Tesla to operate with minimum bureaucracy and maximum efficiency. WarpDrive coordinates most internal operations at both of these companies, giving anyone the ability to track everything from material deliveries, through component manufacture, on to final product deployment. Every design change, material handling procedure and test is coordinated by computer, allowing the complete lifecycle of any individual component or finished product to be tracked and accounted for. While this makes each employee completely responsible for their work, it also encourages professional attitudes due to the near autonomy they enjoy in their respective role. In addition, this software system is designed to facilitate any changes to the production process or product, which enables extraordinarily rapid iteration, sometimes in a matter of hours.
“One of our responsibilities is to build the [WarpDrive] software used by almost everyone at the company to get the vehicle to the pad and ready for launch. That includes supply chain, manufacturing, finance, inventory, etc.”2 ~ Anthony Rose, SpaceX Software Engineering Manager
As previously noted, SpaceX donated their WarpDrive software to Tesla, who use it to perform wonders...
“Sometimes 20 small changes [are made to Tesla models] in a single week… Small changes usually affect production for less than an hour, as they’re applied to a few stations during breaks. Big changes may impact output for a day or two.”3 ~ Elon Musk
By contrast big auto manufacturers usually produce a new model every year to enact a relatively small number of approved changes, because each change would have to be agreed with the component supply chain, then downtime scheduled to alter the production machinery. At Tesla (or SpaceX) any improvements are applied to the vehicle as soon as possible through harnessing agile engineering techniques.4 Lean development teams work in parallel to produce a continuous stream of technical improvements which then flow into vehicles throughout the working day, using automated production software. All other advantages aside, this day vs year difference in technological advancement suggests salient companies like Tesla and SpaceX will steamroll the competition,5 with the promise of many more disruptors, like The Boring Company and Neuralink to follow.
“Basically Tesla is aiming to be the best at manufacturing of any company on Earth. This is the thing that’s actually most important in the long run…Tesla will be – absolutely be head and shoulders ahead of everyone else in manufacturing, that’s our goal.”6 ~ Elon Musk/Battery Day
To some this might sound like common fare from the CEO of a multinational company, but Elon is an engineer first so always strives to be accurate – and takes all necessary steps to achieve Tesla’s ascendency. They say the greatest compliment often comes from one’s rival…
“Future competition with Tesla's new Gigafactory [in Germany] will be brutal. The electric car pioneer sets new standards in car production. For example, a Model 3 is built in 10 hours, more than 3 times as fast as a VW ID.3 (vehicle) in Zwickau. This puts Tesla in another dimension in terms of productivity and profitability.”7 ~ Ralf Brandstätter, CEO of Volkswagen Passenger Cars at Wolfsburg crisis meeting
It seems most conventional companies have been out evolved by these Musk disruptors and simply too slow to respond to this new paradigm for technologic development. As Sandy Munro wryly observes: “It’s not the big that eat the small, it’s the fast that eat the slow.”8
Artificial Intelligence
As previously mentioned, Tesla produces a motherboard which incorporates twin neural processors that perform parallel processes, similar to the twin hemispheres of the brain. This runs their FSD neural network software, which allows Tesla vehicles to drive autonomously in any road conditions, much as a human driver. These neural processors are fairly unique because they are designed to process vision-based information, supplied by internal and external cameras, then make judgment calls over the best course of action.
Of course these powerful neural processors could also be used for more advanced applications such as fully autonomous robots or androids. Overall these chips should provide sufficient hand to eye coordination for quite delicate manipulation, coupled with the deep learning required to perform complex tasks. SpaceX are becoming increasingly invested in robots, both of their own design (Octagrabber robot used to secure booster stages after ASDS landings), off the shelf models from outside suppliers (e.g. Spot the robot dog supplied by Boston Dynamics) and will likely employ more advanced robots supplied by Tesla (i.e. Tesla Bot and wheeled logistics robots).
“Tesla’s Mobile Robotics team designs and builds humanoid bi-pedal robots (Tesla Bot) to automate repetitive tasks and wheeled robots for manufacturing and autonomous logistics.” ~ Tesla Robotics Careers page
No doubt there will be a distinct shortage of labor on Mars for the foreseeable future, hence a machine which could perform complex tasks autonomously in partial atmosphere would be indispensable, across a variety of applications. For example: SpaceX have to prove methalox propellant can be produced in situ on Mars, to fully refuel a return Starship, before they risk sending any personnel. This will likely require a wide range of specialized equipment, capable of operating autonomously using mainly vision based input, over a multi-year duration.
This range of robotic equipment would likely include: -
Starship spacecraft – capable of precise navigation through space including successful landing on Mars, relying mainly on visual information (i.e. star fixes and surface terrain identification respectively).
Mining rovers – to find and extract concentrations of subsurface ice then haul extracted material back to Starship for processing.
In situ propellant plant – (mounted inside Starship) to synthesize methane and oxygen propellant from raw materials, i.e. water mined from Martian regolith and carbon dioxide extracted from the Martian atmosphere.
Mars androids – to perform similar work to personnel, i.e. operate, maintain and repair all equipment with minimum external guidance.9
“The logical thing to do is basically outfit one of the ships as a propellant plant itself, and just land it on the planet as a working propellant plant. And then you just need little miner droids to go dig up ice and bring it back and unfurl the solar panels.”10 ~ Elon Musk/Popular Mechanics
To successfully achieve this level of sophistication implies they will have effectively created thinking machines, capable of fully autonomous operation; the first true AGI (Artificial General Intelligence) robots. If SpaceX (and by inference X.com) possess such technologic marvels, it would grant them enormous commercial advantage, both in space and back here on Earth.
“Tesla AI might play a role in AGI, given that it trains against the outside world, especially with the advent of Optimus (aka Tesla Bot)”11 ~ Elon Musk
These AGI robots will likely prove crucial to creating a functioning space economy. Some foundational technologies required for a true space economy are: asteroid mining, in-space construction, and mass transport – all of which would need to be highly integrated and automated for practical applications. In most cases people are poorly suited to working in vacuum and zero-g, whereas automata work well under these conditions and ideally suited to the empirical nature of these emerging space industries.
The rise of automata alone will likely place X.com in an unassailable position and guarantee their commercial ascendance. For example: Tesla Bot could work 24/7 at no cost, except for energy, which is found in abundance in space. By comparison, if the same work was carried out by a human operator, they might manage a few hours a day in a pressure suit and probably achieve far less overall, due to their limited speed and dexterity. In addition the human operator would likely demand a substantial wage to carry out such hazardous work, overall rendering any non-robotic endeavor a non-starter with regards to competing with X.com. Essentially robots will enable the space economy to exist, because they are the only way to generate the efficiencies required to make it work in practise.
Of course these AGI robots would need to be exquisitely programmed, to even approach the complexity of human functionality. Essentially huge amounts of data will have to be collected for every conceivable task and situation the AGI might encounter. Then this data would need to be processed to create a sophisticated behavioral program and comprehensively tested in a high fidelity simulated environment before it could be uploaded into the AGI’s neural network. No surprise, Tesla are building one of the most powerful computers in the world, using their own custom designed hardware and software. This Dojo computer could be used to train neural networks for any conceivable role due to its monstrous data crunching capability, i.e. 1.1 exaflop (1018 floating point operations per second), and relative ease of operation.12 This raises the real prospect of creating a superintelligent AI at some point in the future, which implies enormous opportunities for SpaceX despite the inherent hazard (see Appendix F: Fermi Paradox). For example, one key restriction for pushing technology is the availability of excellent engineers. Elon bemoans the fact he has to constantly shepherd his small band of ace engineers from one priority project to another, to keep the company on track. No doubt the prospect of manufacturing a superintelligent engineer seems pretty alluring right now, as it would remove the next bottleneck after manual labor becomes ubiquitous.
For example: say SpaceX need to build a rotating space station to mitigate the adverse effects of microgravity (i.e. physical and mental deconditioning).13 A superintelligent AI could complete the station design in less than a day, create the space hardened robotic machinery required in less than a week, then construct the station in orbit in less than a month, ready for human habitation.
Whether they decide to risk creating a superintelligence or not, their Dojo supercomputer offers staggering commercial potential. Once it’s cut its teeth on Full Self Driving, Dojo could then be used to train any semi-sentient application, such as: aerospace traffic control, building management, corporate administration, government finance etc. Assuming no massive breakthrough, say in quantum computing from a competitor, X.com would hold the purse strings to the worlds.
1 https://www.wired.com/story/how-elon-musk-gwynne-shotwell-join-spacex/
2 https://stackoverflow.blog/2021/05/13/building-the-software-that-helps-build-spacex/
https://twitter.com/elonmusk/status/1496195175990845441
5 https://arstechnica.com/science/2021/03/european-leaders-say-an-immediate-response-needed-to-the-rise-of-spacex/
https://mobile.twitter.com/alex_avoigt/status/1443995564295741442
https://twitter.com/elonmusk/status/1428951783615864835
10 https://www.popularmechanics.com/space/moon-mars/a26513651/elon-musk-interview-spacex-mars/
https://twitter.com/elonmusk/status/1483728729545363457
13 https://www.medpagetoday.com/neurology/generalneurology/94951