6 ms·
Computation.
by amitu 3y ago
Computation.
- jackmott42 3y agoThat is a good point! GPT5 in space, beam down the model, not the energy.
- Faint 3y agoWith the speed GPT4 is replying, you could easily move inference to GEO as well, and no-one would notice.
- CorrectHorseBat 3y agoCooling is going to be a big problem for that
- mlyle 3y agoAny use of energy in space is going to have pretty big cooling problems unless it's a fundamentally "hot" process that radiates away heat easily. Because any usage of energy in space all becomes heat.
- stevenhuang 3y agoSolvable with a large enough thermal mass. And once we're able to start mining the moon or asteroids for mass to be cheap enough, heat would no longer be a problem in space: use conduction and convection to move heat away quickly, have a holding area for hot matter for radiative cooling to do its thing, then keep adding additional mass until equilibrium temperature of the system is below requirements
- mlyle 3y agoThermal mass doesn't change the equilibrium temperature at all. Only increasing surface area does (well, and changing its emissivity) .
- stevenhuang 3y ago> have a holding area for hot matter for radiative cooling to do its thing The implication of course is that surface area can be increased trivially by adding matter. And with a larger thermal mass, more energy will be required to heat the system to the same temperature, giving more slack in the system for radiative cooling to work.