Divide words over karma and it turns out a lot of high ranking accounts just write up nonsense which gathers votes, lol.
I've found one with 1,000x more words than average and only 10x more karma. Extremely low signal-to-noise ratio, and that's not even accounting for handicaps, favored users, and so on.
It matches patterns in other media where the top 1% of content creators just put out rubbish, quantity, no quality.
You just have to remember, most of these people live in high density regions and have little comprehension about how much surface area humanity truly occupies... And that isn't even accounting for offshore constructs.
What do you think the limiting factor is? I don't see why we can't scale manufacturing of satellites up as far as we want. If we mine out a substantial fraction of the mass of the earth, we can go harvest asteroids or something.
> What do you think the limiting factor is? I don't see why we can't scale manufacturing of satellites up as far as we want.
A reason. I'm sure that theoretically it's possible, assuming infinite money and an interest to do so. But literally, why would we? There's no practical ways to get the power back on earth, it's cheaper to build a solar field, etc.
And I don't believe datacenters in space are viable, cost wise. Not until we can no longer fit them on earth, AND demand is still increasing.
You need to be able to harness enough raw material and energy to build something that can surround the sun. That does not exist in the solar system and we do not yet have the means to travel further out to collect, move, and construct such an incredibly huge structure. It seems like a fantasy.
The inner planets contain enough mass to create a shell of 1 AU radius with mass of 42 kg/m^2. That sounds like a plausible thickness and density for a sandwich of photovoltaics - GPUs - heat sinks.
You don't build a rigid shell of course, you build a swarm of free-floating satellites in a range of orbits.
The energy to build the system comes from the partial assembled system, plus some initial bootstrap energy. It grows exponentially. We seem to have enough today to build small factories in orbit.
The manufacturing scale comes from designing factory factories. They aren't that far in the future. Most factory machinery is made in factories which could be entirely automated, so you just need some robots to install machines into factories.
I was told ca. 2003 or so that because features on computer chips were getting smaller at some rate, and processor speed was getting faster at some other rate, that given exponential this or that I'd have tiny artificial haemo-goblins[1] bombing around my circulatory system that would make me swim like a fish under the sea for hours on end. But it turned out to be utter bullshit. Just like this.
Great. Now run the numbers to find the energy required to disassemble the planets and accelerating the pieces to their desired locations. For reference, it takes over 10 times of propellant and oxidant mass to put something in LEO.
The burned propellant and oxygen mass (as H2O and CO2) almost all ends up back in the atmosphere when you launch to LEO, so you can keep running electrolysis (powered by solar) to convert it back to fuel.
Sure, but if we're talking about solar engineering, that mass is going to be dispersed in orbit around the sun. You're not going to be reaccumulating that any time soon.
Also it's gravitationally unstable, like Dyson Rings, where as soon as you have any perturbance from the center means that the closer side is more attracted to the sun so it enters a feedback loop.
There are only so many people who can make satellites; there are only so many things to make satellites out of; and there are only so many orbits to put them in. There are only so many reasons why a person might want a satellite. There are only so many ways of placing satellites in orbit and each requires some amount of energy, and we have access to a finite amount of energy over time.
Finally, if we limited ourselves to earth-based raw materials, we would eventually reach a point where the remaining mass of the earth would have less gravitational effect on the satellite fleet than the fleet itself, which would have deleterious effects on the satellite fleet.
Seven reasons are intuitive; I’m sure there are many others.
People can build a factory that makes satellites. And then a factory that makes factories to make satellites.
There is plenty of material in the solar system (see my other response), and plenty of orbits, and launch capability can scale with energy harvested so the launch rate can grow exponentially.
Lots of people will probably decide they don't want any more satellites. But it only takes a few highly determined people to get it done anyway.
> Finally, if we limited ourselves to earth-based raw materials, we would eventually reach a point where the remaining mass of the earth would have less gravitational effect on the satellite fleet than the fleet itself, which would have deleterious effects on the satellite fleet.
The Earth's crust has an average thickness of about 15-20 km.
Practically we can only get at maybe the top 1-2 km, as drill bits start to fail the deeper you go.
The Earth's radius is 6,371 km.
So even if we could somehow dug up entire crust we can get to and flung it into orbit, that would barely be noticeable to anything in orbit.
After a few decades, you need to start replacing all the solar panels.
And the robot army being used to do the construction and resource extraction will likely have a much shorter lifespan. So needs to be self-replicating/repairing/recycling.
Yeah, but besides not having the physical amount of material available in the solar system, or the availability of any technology to transfer power generated to a destination where it can serve a meaningful purpose in the foreseeable future, or having the political climate or capital necessary for even initiating such an effort, or not being able to do so without severely kneecapping the habitability of our planet, there are aren't really any meaningful barriers that I can see.
The various AI accelerator chips, such as TPUs and NVidia GPUs, are only compatible to extent that some of the high level tools like PyTorch and Triton (kernel compiler) may support both, which is like saying that x86 and ARM chips are compatible since gcc supports them both as targets, but note this does not mean that you can take a binary compiled for ARM and run it on an x86 processor.
For these massive, and expensive to train, AI models the differences hit harder since at the kernel level, where the pedal hits the metal, they are going to be wringing every last dollar of performance out of the chips by writing hand optimized kernels for them, highly customized to the chip's architecture and performance characteristics. It may go deeper than that too, with the detailed architecture of the models themselves tweaked to best perform on a specific chip.
So, bottom line is that you can't just take a model "compiled to run on TPUs", and train it on NVidia chips just because you have spare capacity there.
I've found one with 1,000x more words than average and only 10x more karma. Extremely low signal-to-noise ratio, and that's not even accounting for handicaps, favored users, and so on.
It matches patterns in other media where the top 1% of content creators just put out rubbish, quantity, no quality.
reply