The panels at least radiate *their* heat away from their large rear surface area, but the datacentre itself has to have large amounts of fluid cycling out and back to roughly comparably large radiators.
I wonder if it would help at all to build one modest server into the back of each solar panel, rather than trying to concentrate all of the servers together?
One option would be to just not use orbital servers for jobs where low latency is important. E.g. if you want your server to spend the next four hours optimizing an elaborate programming solution, you can have that done in space, but if you want it to remote-control your Tesla's driving, you'd better use a ground-based server instead.
Realistically, though, latency is the least of the problems with Musk's idea. Other factors will keep this project on the ground.
I actually think it is possible that FSD will become a thing with just cameras.
Well, it's a thing, in that Teslas are driving with it now, but more specifically will it ever be safe enough to be fit for purpose (i.e. to drive actually unsupervised)? To get there, Tesla would need to come up with a way for the cameras to handle bad visibility (such as driving directly into the sun), and it's not clear to me that there is any good solution to that except adding other kinds of sensor for redundancy.
Meantime, a solar farm in the southern US that can deliver 1GW would cost about $1bn, and the battery bank another $1bn
.... amortized over a 30-year operational lifespan, with near-zero risk of catastrophic loss during that period.
$1b plus gone is just the cost of doing business.
That's the cost of doing business in space, sure. But Musk would be competing against companies providing the same service using data centers based on Earth, that don't have to pay that cost because they can just walk over to the malfunctioning server and fix or replace it. Given the rate at which the GPU-farm industry is commoditizing, it's hard to see how Musk wouldn't get underpriced and driven out of the market.
Tomorrow, if the PLA wanted, they could just walk into Primorsky and Krasnoyarsk and annex them. Heck, they could easily overrun and conquer all of Russia from the Urals to the Bering Sea
You don't think Russia's nuclear arsenal would still serve as a deterrent against that sort of thing? (Not that I have a whole lot of confidence in Russia's ICBMs actually working, but China would probably have to assume at least some of them would, and that losing some cities would be a bad thing for them)
Imagine if so many of us drive Teslas to avoid car exhaust emissions looking at the exhaust from hundreds of rocket launches.
Fun fact about the Starship: its fuel is liquid methane and liquid oxygen, which when combusted results in exhaust consisting of water vapor and carbon dioxide. So, not great from a global-warming perspective, but not really polluting in the classic sense.
Falcon 9, OTOH, burns kerosene and oxygen, and emits water vapor, carbon dioxide, and soot. Presumably they will start phasing that out in favor of Starship though, when they can.
Musk has done this before. In 2016, SolarCity was in financial trouble, so Musk had Tesla bail them out by purchasing them. Now he's using the same maneuver to keep xAI afloat using SpaceX's assets.
I respect people who are willing to sink money into ambitious projects like this, but this is about as safe as a hangglider.
Hang glider pilot here -- a hang glider is much safer than this vehicle. A hang glider has zero moving parts, zero electronics, zero propellers, and zero batteries/fuel, so there's much less than can go wrong with it in flight... and even if you do somehow lose control and crash, there's nothing flammable on board, and your speed is typically low enough that your injuries will be minimal.
The Helix is a very cool vehicle, but I'm glad it's other people flying it and not me
takeoff and landing need to be, at best, on a helipad -- which you will have to clear immediately for the next guy coming in whose battery is going flat.
Takeoff and landing can be from anywhere you can find 100 square feet of open space and permission from the landowner.
As for having to clear the landing area immediately for the next guy, there are only 6 of these in the world, so the chances of two of them needing to land at exactly the same spot at about the same time are pretty low.
Wouldn't it be more efficient to provide an electric rail on the tracks? Then the train wouldn't have to carry a battery around.
Doing anything at all helps depression to some extent, since depression is partially about being unable to form a satisfactory self-narrative about your ability to handle your problems.
By taking some steps to combat your depression (regardless of their inherent level of efficacy), you see yourself as someone who is doing something that might lead you to a better future. Just knowing/thinking that can help reduce feelings of hopelessness.
Yes, it's the placebo effect; but for depression, where self-perception is central to the problem, it's particularly relevant.
Once we are at the point when there is plenty of compute to generate 4k (or VR equivalent) video or interactive content based on a user's unstructured input, what else is there? If Moore's law becomes dead soon does it matter?
You'll always have bitcoins to mine -- your computer will never be fast enough unless it's faster than the other guy's, and vice-versa.
Just stick in some batteries!
If only it were that easy! What you really mean is, "just throw away every one of the billions of ICE-powered vehicles in the world and replace it with the closest battery-powered equivalent"
In the meantime, what can we do about all of those ICE-powered vehicles that are (and will be) in daily use? If we can figure out how to run them without continuing to transfer underground CO2 into the atmosphere, that's also a worthwhile objective.
He's going to keep on re-studying the data until it damn well provides him with the conclusion he wants to reach!
1 Angstrom: measure of computer anxiety = 1000 nail-bytes
