Speed yes, but cost? Can it make better comments in 100 tokens input and output, than humans in 15 seconds?
(I still haven't used 4 properly, so possibly? And given the hypothetical situation under discussion, could it have if it had not been directed not to?)
It ended up being so complex, that none of the implementations were up to spec, and hardly any 2 implementations talked to each other out of the box. i.e. the exact opposite it was meant to do.
Yeah, no night, no atmosphere, constant power output, lot's of real estate, and panels in space do not need heavy superstructure to withstand weather or gravity, thus they could potentially be extremely light for the area (micrometers to tens of micrometers thick at most needed for light absorption). And in far future, if they are built of materials taken from moon or meteorites, you could also bypass most of the lifting cost for even that. Of course, that would need huge scale to justify the R&D to pull it off.
Anyway, space based solar power is the end game. Nothing on earth will ever provide the quantities of power (not even nuclear, fusion or fission) that capturing solar energy can.
Worth noting this isn't as much of a benefit as it's made out to be.
If you were designing a perfect power source, it would match demand, so produce more in winter in polar regions, and more in summer for regions with lots of AC. Similarly, you'd generally want more power during the day than at night.
This is part of the reason a mix of solar and wind that varies by latitude is an ideal mix.
Space power might get more bang for buck if it could target its power to different regions e.g. swapping from north to south as the seasons change, and/or following the day/night cycle and/or weather to maximise energy price.
The satellites would be in geostationary, with phased-array transmitters, focused by a reference signal from the ground target. They actually could be repointed to various receivers. The ground stations would be relatively cheap so it wouldn't be all that important to maximize their utilization.
> They could potentially be extremely light for the area (micrometers to tens of micrometers thick at most needed for light absorption).
This is really the key, if you can make a solar panel that’s as light and thin as say mylar, and then unfold it when you get to space, we could put up several kilometers of solar panels without requiring much mass at all. It’s not like there’s wind or rain up there to wear it down.
No night? Where are you putting these satellites? The only orbit that doesn't eclipse is sun synchronous, an already crowded orbit. Even GEO satellites experience eclipse.
> Glaser had noticed that a satellite in geosynchronous Earth orbit (GEO), 35,786 kilometers above the equator, would pass through Earth's shadow for only a few minutes each year.
The article addresses this. To my mind 4 minutes per year is equivalent to the parent’s layspeak “no night” comment.
4 minutes??? Not even close. Each eclipse varies from a few seconds for the first one to practically an hour in the middle, then gets shorter again, over roughly 30 eclipses each season (2 per year)
So, let's cut down the pedantry a bit. Even in the most pessimal reasonable orbits, up to an hour of outage in the middle of the night for 3-4 months per year is a very different beast from getting just several hours of sun per day.
Why?
- You still get better duty cycle from the panels
- The time the power is missing is very small
- The need for storage, etc, is low because it's a short period of relatively low demand that is missing.
They address this in the article: Only twice per year are the panels in shadow, at the equinoxes. Presumably all the rest of the time they're slightly above/below the earth with respect to its orbit. It's the same reason the moon is lit nearly all the time.
Respectfully, I believe you are inaccurate on both accounts, sun synchronous orbits (SSO) only have no eclipse if their orbital plane is within a very degrees of the terminator, this of course widens with altitude but there exists far more possible SSOs with eclipses than without.
Secondly with GEOs as with any very high altitude orbit the eclipse time trends towards zero so effectively at a GEO the eclipse time is minimal compared to the illuminated time. Of note the length for the eclipse time of GEO varies throughout the year.
Respectfully, you are not correct that eclipse in Geo is insignificant. It drives many engineering constraints for spacecraft systems. Look I get why people are excited about space space solar power, but I've been in this industry for 15 years and when you dig into the numbers and understand the realities of spacecraft engineering, SBSP just seems like a fool's errand. It may not always be that way, but the technical challenges are extreme and costs are still nowhere near what would be required to make it work.
You can chalk my comments up to a grumpy engineer tired of the cyclical SBSP pushes that never go anywhere.
Hmm, well I totally understand how it drives spacecraft design and I don’t know off the top of my head the length of greatest eclipse but it cannot be greater than a small fraction of the period (upwards of 1/12). This is significant for a spacecraft itself but as far as a regionally integrated power grid is concerned it is a clear improvement over current solar PV systems.
I get that there maybe some fatigue here with the idea and blue sky optimism that comes with SBSP, I think there are valid criticisms to level at it orbit selection and corresponding ground tracks and eclipses are not one.
> It drives many engineering constraints for spacecraft systems.
Here the only thing it's going to drive are thermal requirements-- which are admittedly significant problems. There's no need to continue to use large amounts of power for comms, etc, like on most GEO birds.
And, of course, the grid needs to deal with the power disappearing for an hour in the middle of the night for short periods of the year.
Why in this case would it cause significant problems? Are you referring to the heating (that would make sense with such a large surface area of panels)?
Concerning cooling- with such short times in eclipse I can’t imagine that it would have enough time to have cooling issues beyond flexing of the superstructure (if made from metals). Be interested to hear though if I’m missing something.
As to power disappearing with an adequately geographically integrated grid I don’t forsee that as really too much of a problem. Currently the grid deals with short term outs fairly well especially if they are planned for months in advance.
> Concerning cooling- with such short times in eclipse I can’t imagine that it would have enough time to have cooling issues
Lots of thin structure with 70 minutes to radiate, with the only thing shining on it the earth's albedo subtending a tiny angle. I'd imagine it creates rather significant demands on structure and electrical connections.
I've not run the numbers on a GEO solar spacecraft, but the smallsat group that I'm mentoring that would be "thicker" than a lot of the GEO craft... gets down to -30C without heaters during its 40 minutes in eclipse while much closer to Earth.
> As to power disappearing with an adequately geographically integrated grid I don’t forsee that as really too much of a problem. Currently the grid deals with short term outs fairly well especially if they are planned for months in advance.
Yup, that's the point I'm making. A space based solar power craft has smaller problems from eclipse than a typical comsat. Batteries, etc, are not nearly as much of a concern. It's mostly the thermals that are left.
There's no real intrinsic electrical or battery system requirements for a giant solar power satellite. The energy use is comparably trivial: no big transponders to run in eclipse like a comsat.
But you do need everything to survive the cold and thermal cycling.
Not just thermal cycling but a giant solar collector is literally a giant solar sail. There would be significant force applied to a giant collector.
You'd need propulsion to maintain position and orientation. You'd also need a number of propulsion units to balance solar pressure gradients as the collector entered and existed the Earth's shadow as well as the thermal expansion/contraction of the structure.
It's likely not a lot of power but a non-trivial amount of fuel.
Ideally sail effect would be used for position keeping, and to reduce loads needed to keep the assembly together. I suppose that would reduce amount of area available for soaking the sun, but might still be lighter than trying to build stiff enough large structures, plus you wouldn't constantly need to ferry more fuel.
You probably maintain orientation with a control moment gyro and periodically desaturate it with thrust. You also use thrust to reboost and stay in orbital slot.
High-impulse ion engines, etc, are a good match for this task.
You'd need a rigid structure to keep the gyro from ripping itself loose from the structure. Even constructed as a giant space frame, that's a lot of mass to deal with the torsion of the structure rotating.
A structure 100m on a side would be just at the bounds of current technology (the ISS's control moment gyros). With 30% efficient panels that's only about 4MW before conversion and path losses.
The high impulse ion engines to desaturate the gyros would still need to be refueled regularly. I think you're hand waving a lot of complexity that even if completely solved still leaves a solution that's orders of magnitude costlier than solar panels on the ground.
This is exactly how the NOAA GOES sats work. Every day at a proscribed time we used to desaturate the reaction wheels, so they could more or less keep running constantly and keep the satellite pointing where it should be.
Yup. This would be big enough that a control moment gyro would be "worth it," too-- and could store a whole lot of momentum and allow less frequent desaturation burns.
If someone has compiled a cost comparison of transcontinental HVDC links compared with the alternative spaceborne solution I'd certainly love to see it.
The space based solution is possibly actually more reliable, as there are actually less components involved that could fail.
One isn't a death ray; but if you only have one, then you need a global power grid anyway.
If you have enough satellites to not need the distribution grid, and they're all in geostationary orbit, then many are over the horizon at the same time and they can (in principle) be combined on the same place.
If they're in a low enough orbit that you only get a few over the horizon at any given moment, you get a substantial penalty from Earth's shadow.
On Mars this would be a great thing for colonies; get past the global dust storms, and it won't matter if you have only a handful of sites; on Earth… pick which failure mode you prefer.
Not for a single device with a fixed wavelength and size; unless someone figures out a way to cheat quantum mechanics and the diffraction limit, how well you can aim light is directly related to the size of your aperture (/antenna) in wavelengths.
Convincing governments you've not cheated with a gigawatt optical laser on your satellites (optical wavelengths being smaller than microwaves makes them easier to focus with smaller parts), that's a separate question. I assume an Iranian one of these would get destroyed by Israel for the same reason they attack their neighbour's nuclear reactors.
But was it parsed and reformatted specifically in the "chat format" (i.e. the same as inputs later fed to the model when used as a chatbot)? It can make a surprisingly big difference.
Remember that these models generate one token at a time. They do not "think ahead" much more than maybe a few tokens in beam search. So if the problem requires search - actual comparison of approaches, and going back-and-forth between draft and thinking through the implications - the model can't do it (except in a limited sense, if you prompt it to give it's "train of thought"). So it's comparable of you being in front of whiteboard, hit with a question, and you would have to start answering immediately without thinking more than you can while talking through your answer at the same time. Doable if you know the material well. If it's a new problem, that approach is doomed.
Given that, I think the language models do remarkably well. A little bit of search, and maybe trying to generate the answer in different order (like short draft -> more detailed draft -> more detailed draft... etc.) will improve things a lot.
With discrete logic you'd be committed to the model for whatever its lifetime is, which might be acceptable depending on the application. With FPGA you would still be able to update it after it's deployed.
